| { | |
| "best_metric": 7.5543536040825, | |
| "best_model_checkpoint": "whisper-it-medium-multids-augmented/checkpoint-7500", | |
| "epoch": 1.3639999999999999, | |
| "global_step": 7500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.9000000000000005e-06, | |
| "loss": 2.2754, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.900000000000002e-06, | |
| "loss": 0.5525, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.49e-05, | |
| "loss": 0.3671, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9900000000000003e-05, | |
| "loss": 0.34, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4900000000000002e-05, | |
| "loss": 0.3427, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.9900000000000002e-05, | |
| "loss": 0.3189, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.49e-05, | |
| "loss": 0.3161, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 3.99e-05, | |
| "loss": 0.3195, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.49e-05, | |
| "loss": 0.3199, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.99e-05, | |
| "loss": 0.3156, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.965e-05, | |
| "loss": 0.3276, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.929285714285715e-05, | |
| "loss": 0.3707, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.893571428571429e-05, | |
| "loss": 0.3778, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.857857142857143e-05, | |
| "loss": 0.3276, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.8221428571428574e-05, | |
| "loss": 0.3292, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.7864285714285714e-05, | |
| "loss": 0.3173, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.750714285714286e-05, | |
| "loss": 0.3031, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.715e-05, | |
| "loss": 0.3026, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.679285714285714e-05, | |
| "loss": 0.2948, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.643571428571429e-05, | |
| "loss": 0.3088, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.6078571428571434e-05, | |
| "loss": 0.3146, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.5721428571428574e-05, | |
| "loss": 0.3035, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.5364285714285714e-05, | |
| "loss": 0.2946, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.500714285714286e-05, | |
| "loss": 0.2852, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.465e-05, | |
| "loss": 0.2938, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.429285714285715e-05, | |
| "loss": 0.2955, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.393571428571429e-05, | |
| "loss": 0.2906, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.357857142857143e-05, | |
| "loss": 0.2794, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.3221428571428575e-05, | |
| "loss": 0.2689, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.2864285714285715e-05, | |
| "loss": 0.2627, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.2507142857142855e-05, | |
| "loss": 0.2498, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.215e-05, | |
| "loss": 0.2587, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.179285714285715e-05, | |
| "loss": 0.252, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.143571428571429e-05, | |
| "loss": 0.2515, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.1078571428571435e-05, | |
| "loss": 0.2172, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.0721428571428575e-05, | |
| "loss": 0.2312, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.0364285714285715e-05, | |
| "loss": 0.236, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.000714285714286e-05, | |
| "loss": 0.236, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 3.965e-05, | |
| "loss": 0.2186, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 3.929285714285714e-05, | |
| "loss": 0.2131, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 3.893571428571429e-05, | |
| "loss": 0.235, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 3.857857142857143e-05, | |
| "loss": 0.2259, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.8221428571428575e-05, | |
| "loss": 0.2192, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.786428571428572e-05, | |
| "loss": 0.2368, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.7507142857142855e-05, | |
| "loss": 0.2238, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.715e-05, | |
| "loss": 0.2084, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.679285714285715e-05, | |
| "loss": 0.2144, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.643571428571429e-05, | |
| "loss": 0.1985, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.607857142857143e-05, | |
| "loss": 0.2006, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.5721428571428575e-05, | |
| "loss": 0.226, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_loss": 0.27792105078697205, | |
| "eval_runtime": 8108.2258, | |
| "eval_samples_per_second": 1.85, | |
| "eval_steps_per_second": 0.058, | |
| "eval_wer": 14.664177120986604, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.5364285714285715e-05, | |
| "loss": 0.1799, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.500714285714286e-05, | |
| "loss": 0.1948, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.465e-05, | |
| "loss": 0.1982, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.429285714285714e-05, | |
| "loss": 0.1909, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.393571428571429e-05, | |
| "loss": 0.1717, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.357857142857143e-05, | |
| "loss": 0.1809, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.3221428571428575e-05, | |
| "loss": 0.191, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.2864285714285715e-05, | |
| "loss": 0.1716, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.2507142857142855e-05, | |
| "loss": 0.1617, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.215e-05, | |
| "loss": 0.1768, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.179285714285715e-05, | |
| "loss": 0.1879, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.143571428571428e-05, | |
| "loss": 0.1771, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.107857142857143e-05, | |
| "loss": 0.1552, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0721428571428576e-05, | |
| "loss": 0.1744, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0364285714285716e-05, | |
| "loss": 0.1773, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.0007142857142856e-05, | |
| "loss": 0.188, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.965e-05, | |
| "loss": 0.1788, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.9292857142857146e-05, | |
| "loss": 0.1759, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.8942857142857144e-05, | |
| "loss": 0.1901, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.8585714285714287e-05, | |
| "loss": 0.1832, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.822857142857143e-05, | |
| "loss": 0.1829, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.787142857142857e-05, | |
| "loss": 0.1765, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.7514285714285714e-05, | |
| "loss": 0.1829, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.7157142857142858e-05, | |
| "loss": 0.1828, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.6800000000000004e-05, | |
| "loss": 0.1882, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.644285714285714e-05, | |
| "loss": 0.1957, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.6085714285714284e-05, | |
| "loss": 0.1954, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.572857142857143e-05, | |
| "loss": 0.1725, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.5371428571428574e-05, | |
| "loss": 0.1523, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.5014285714285718e-05, | |
| "loss": 0.1402, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.4657142857142858e-05, | |
| "loss": 0.141, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.43e-05, | |
| "loss": 0.1389, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.3942857142857144e-05, | |
| "loss": 0.1541, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.3585714285714284e-05, | |
| "loss": 0.1739, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.322857142857143e-05, | |
| "loss": 0.1598, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.287142857142857e-05, | |
| "loss": 0.148, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 2.2514285714285715e-05, | |
| "loss": 0.1329, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.2157142857142858e-05, | |
| "loss": 0.1549, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 2.18e-05, | |
| "loss": 0.1379, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 2.1442857142857145e-05, | |
| "loss": 0.1533, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 2.1085714285714288e-05, | |
| "loss": 0.1412, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 2.072857142857143e-05, | |
| "loss": 0.1382, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 2.037142857142857e-05, | |
| "loss": 0.1493, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 2.0014285714285715e-05, | |
| "loss": 0.166, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9657142857142858e-05, | |
| "loss": 0.1681, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.93e-05, | |
| "loss": 0.15, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.894285714285714e-05, | |
| "loss": 0.1226, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.858571428571429e-05, | |
| "loss": 0.1164, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.8228571428571428e-05, | |
| "loss": 0.1114, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.787142857142857e-05, | |
| "loss": 0.1278, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "eval_loss": 0.18177375197410583, | |
| "eval_runtime": 8353.1084, | |
| "eval_samples_per_second": 1.796, | |
| "eval_steps_per_second": 0.056, | |
| "eval_wer": 10.204922389963853, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7514285714285715e-05, | |
| "loss": 0.1153, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.715714285714286e-05, | |
| "loss": 0.0981, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.6800000000000002e-05, | |
| "loss": 0.0855, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.6442857142857142e-05, | |
| "loss": 0.0808, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.608571428571429e-05, | |
| "loss": 0.066, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.572857142857143e-05, | |
| "loss": 0.0653, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.5371428571428572e-05, | |
| "loss": 0.0796, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.5014285714285714e-05, | |
| "loss": 0.0752, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.4657142857142859e-05, | |
| "loss": 0.0767, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.43e-05, | |
| "loss": 0.0659, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.3942857142857144e-05, | |
| "loss": 0.0616, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.3585714285714287e-05, | |
| "loss": 0.0585, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.3228571428571429e-05, | |
| "loss": 0.0578, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.2871428571428574e-05, | |
| "loss": 0.0614, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.2514285714285714e-05, | |
| "loss": 0.0578, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.2157142857142857e-05, | |
| "loss": 0.0646, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.18e-05, | |
| "loss": 0.0559, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.1442857142857144e-05, | |
| "loss": 0.0611, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.1085714285714287e-05, | |
| "loss": 0.06, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.0728571428571429e-05, | |
| "loss": 0.0553, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.0371428571428572e-05, | |
| "loss": 0.0559, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.0014285714285716e-05, | |
| "loss": 0.0543, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 9.657142857142857e-06, | |
| "loss": 0.052, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 9.3e-06, | |
| "loss": 0.0459, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.942857142857142e-06, | |
| "loss": 0.0443, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 8.585714285714286e-06, | |
| "loss": 0.0462, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 8.22857142857143e-06, | |
| "loss": 0.0429, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 7.87142857142857e-06, | |
| "loss": 0.0469, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 7.514285714285714e-06, | |
| "loss": 0.0477, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 7.1571428571428584e-06, | |
| "loss": 0.043, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 6.800000000000001e-06, | |
| "loss": 0.0429, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 6.4428571428571435e-06, | |
| "loss": 0.0425, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 6.085714285714286e-06, | |
| "loss": 0.0392, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 5.728571428571429e-06, | |
| "loss": 0.0421, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 5.371428571428572e-06, | |
| "loss": 0.0402, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 5.0142857142857144e-06, | |
| "loss": 0.0426, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 4.657142857142857e-06, | |
| "loss": 0.0376, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 4.2999999999999995e-06, | |
| "loss": 0.0396, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.942857142857143e-06, | |
| "loss": 0.039, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 3.585714285714286e-06, | |
| "loss": 0.0398, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 3.2285714285714288e-06, | |
| "loss": 0.0342, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 2.8714285714285713e-06, | |
| "loss": 0.0411, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.5142857142857147e-06, | |
| "loss": 0.0361, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 2.157142857142857e-06, | |
| "loss": 0.0334, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.8e-06, | |
| "loss": 0.0335, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.4428571428571429e-06, | |
| "loss": 0.0342, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.0857142857142858e-06, | |
| "loss": 0.0335, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 7.285714285714287e-07, | |
| "loss": 0.037, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 3.7142857142857145e-07, | |
| "loss": 0.0287, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.4285714285714288e-08, | |
| "loss": 0.0304, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "eval_loss": 0.13884508609771729, | |
| "eval_runtime": 8184.9748, | |
| "eval_samples_per_second": 1.833, | |
| "eval_steps_per_second": 0.057, | |
| "eval_wer": 7.5543536040825, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "step": 7500, | |
| "total_flos": 2.4491047723794432e+20, | |
| "train_loss": 0.18038643388748168, | |
| "train_runtime": 278160.1615, | |
| "train_samples_per_second": 1.726, | |
| "train_steps_per_second": 0.027 | |
| } | |
| ], | |
| "max_steps": 7500, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 2.4491047723794432e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |