| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "eval_steps": 100, | |
| "global_step": 1294, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 552.1678833007812, | |
| "epoch": 0.0038639876352395673, | |
| "grad_norm": 1.1229428052902222, | |
| "kl": 0.0001697540283203125, | |
| "learning_rate": 1.9230769230769234e-07, | |
| "loss": 0.0, | |
| "reward": 0.9830357611179352, | |
| "reward_std": 0.36204318702220917, | |
| "rewards/accuracy_reward": 0.7892857611179351, | |
| "rewards/format_reward": 0.1937500096857548, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 561.3339538574219, | |
| "epoch": 0.0077279752704791345, | |
| "grad_norm": 0.6426151990890503, | |
| "kl": 0.0002849578857421875, | |
| "learning_rate": 3.846153846153847e-07, | |
| "loss": 0.0, | |
| "reward": 0.9151786208152771, | |
| "reward_std": 0.413673534989357, | |
| "rewards/accuracy_reward": 0.7642857551574707, | |
| "rewards/format_reward": 0.15089286640286445, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 571.7214538574219, | |
| "epoch": 0.011591962905718702, | |
| "grad_norm": 1.2495057582855225, | |
| "kl": 0.00039796829223632814, | |
| "learning_rate": 5.76923076923077e-07, | |
| "loss": 0.0, | |
| "reward": 0.9214286088943482, | |
| "reward_std": 0.366368842124939, | |
| "rewards/accuracy_reward": 0.7767857491970063, | |
| "rewards/format_reward": 0.14464286342263222, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 575.4223388671875, | |
| "epoch": 0.015455950540958269, | |
| "grad_norm": 0.5862982869148254, | |
| "kl": 0.0009944915771484375, | |
| "learning_rate": 7.692307692307694e-07, | |
| "loss": 0.0, | |
| "reward": 0.8276786148548126, | |
| "reward_std": 0.350461420416832, | |
| "rewards/accuracy_reward": 0.6660714685916901, | |
| "rewards/format_reward": 0.16160715073347093, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 523.0955596923828, | |
| "epoch": 0.019319938176197836, | |
| "grad_norm": 0.7838996052742004, | |
| "kl": 0.004752349853515625, | |
| "learning_rate": 9.615384615384617e-07, | |
| "loss": 0.0002, | |
| "reward": 0.9107143223285675, | |
| "reward_std": 0.38466152399778364, | |
| "rewards/accuracy_reward": 0.6642857491970062, | |
| "rewards/format_reward": 0.24642858058214187, | |
| "step": 25 | |
| }, | |
| { | |
| "completion_length": 520.2955688476562, | |
| "epoch": 0.023183925811437404, | |
| "grad_norm": 0.7546085119247437, | |
| "kl": 0.002777099609375, | |
| "learning_rate": 1.153846153846154e-06, | |
| "loss": 0.0001, | |
| "reward": 0.979464328289032, | |
| "reward_std": 0.3285092800855637, | |
| "rewards/accuracy_reward": 0.7267857491970062, | |
| "rewards/format_reward": 0.2526785835623741, | |
| "step": 30 | |
| }, | |
| { | |
| "completion_length": 564.7143005371094, | |
| "epoch": 0.02704791344667697, | |
| "grad_norm": 0.8112598657608032, | |
| "kl": 0.00201263427734375, | |
| "learning_rate": 1.3461538461538462e-06, | |
| "loss": 0.0001, | |
| "reward": 1.008035773038864, | |
| "reward_std": 0.3334413096308708, | |
| "rewards/accuracy_reward": 0.8553571760654449, | |
| "rewards/format_reward": 0.15267857983708383, | |
| "step": 35 | |
| }, | |
| { | |
| "completion_length": 553.3607482910156, | |
| "epoch": 0.030911901081916538, | |
| "grad_norm": 0.7396882176399231, | |
| "kl": 0.004928970336914062, | |
| "learning_rate": 1.5384615384615387e-06, | |
| "loss": 0.0002, | |
| "reward": 1.0133929133415223, | |
| "reward_std": 0.337762188911438, | |
| "rewards/accuracy_reward": 0.8500000417232514, | |
| "rewards/format_reward": 0.16339286509901285, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 549.3768127441406, | |
| "epoch": 0.0347758887171561, | |
| "grad_norm": 0.8082935214042664, | |
| "kl": 0.00592498779296875, | |
| "learning_rate": 1.7307692307692308e-06, | |
| "loss": 0.0002, | |
| "reward": 1.0535714626312256, | |
| "reward_std": 0.3635849982500076, | |
| "rewards/accuracy_reward": 0.905357176065445, | |
| "rewards/format_reward": 0.14821429401636124, | |
| "step": 45 | |
| }, | |
| { | |
| "completion_length": 545.2357391357422, | |
| "epoch": 0.03863987635239567, | |
| "grad_norm": 0.467140793800354, | |
| "kl": 0.007464599609375, | |
| "learning_rate": 1.9230769230769234e-06, | |
| "loss": 0.0003, | |
| "reward": 0.92857146859169, | |
| "reward_std": 0.3401346325874329, | |
| "rewards/accuracy_reward": 0.6875000327825547, | |
| "rewards/format_reward": 0.24107143878936768, | |
| "step": 50 | |
| }, | |
| { | |
| "completion_length": 559.1250305175781, | |
| "epoch": 0.04250386398763524, | |
| "grad_norm": 0.6232195496559143, | |
| "kl": 0.00318756103515625, | |
| "learning_rate": 2.1153846153846155e-06, | |
| "loss": 0.0001, | |
| "reward": 1.0241072118282317, | |
| "reward_std": 0.43311036825180055, | |
| "rewards/accuracy_reward": 0.8714286267757416, | |
| "rewards/format_reward": 0.15267857685685157, | |
| "step": 55 | |
| }, | |
| { | |
| "completion_length": 532.1384124755859, | |
| "epoch": 0.04636785162287481, | |
| "grad_norm": 0.6029495000839233, | |
| "kl": 0.009820556640625, | |
| "learning_rate": 2.307692307692308e-06, | |
| "loss": 0.0004, | |
| "reward": 1.0446429073810577, | |
| "reward_std": 0.37980674505233764, | |
| "rewards/accuracy_reward": 0.8410714626312256, | |
| "rewards/format_reward": 0.2035714380443096, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 496.79644775390625, | |
| "epoch": 0.05023183925811438, | |
| "grad_norm": 0.5804664492607117, | |
| "kl": 0.011712646484375, | |
| "learning_rate": 2.5e-06, | |
| "loss": 0.0005, | |
| "reward": 1.1000000298023225, | |
| "reward_std": 0.3491546869277954, | |
| "rewards/accuracy_reward": 0.8071428835391998, | |
| "rewards/format_reward": 0.2928571552038193, | |
| "step": 65 | |
| }, | |
| { | |
| "completion_length": 522.5732391357421, | |
| "epoch": 0.05409582689335394, | |
| "grad_norm": 0.29278990626335144, | |
| "kl": 0.0121734619140625, | |
| "learning_rate": 2.6923076923076923e-06, | |
| "loss": 0.0005, | |
| "reward": 1.1071429133415223, | |
| "reward_std": 0.3307413190603256, | |
| "rewards/accuracy_reward": 0.9053571879863739, | |
| "rewards/format_reward": 0.20178572386503218, | |
| "step": 70 | |
| }, | |
| { | |
| "completion_length": 559.1973510742188, | |
| "epoch": 0.05795981452859351, | |
| "grad_norm": 0.3587290644645691, | |
| "kl": 0.0087127685546875, | |
| "learning_rate": 2.8846153846153845e-06, | |
| "loss": 0.0003, | |
| "reward": 1.0446429073810577, | |
| "reward_std": 0.37309455275535586, | |
| "rewards/accuracy_reward": 0.867857176065445, | |
| "rewards/format_reward": 0.17678572237491608, | |
| "step": 75 | |
| }, | |
| { | |
| "completion_length": 563.2482452392578, | |
| "epoch": 0.061823802163833076, | |
| "grad_norm": 0.38649681210517883, | |
| "kl": 0.01405792236328125, | |
| "learning_rate": 3.0769230769230774e-06, | |
| "loss": 0.0006, | |
| "reward": 1.081250047683716, | |
| "reward_std": 0.34369638115167617, | |
| "rewards/accuracy_reward": 0.8875000417232514, | |
| "rewards/format_reward": 0.1937500111758709, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 513.6526977539063, | |
| "epoch": 0.06568778979907264, | |
| "grad_norm": 0.3370773494243622, | |
| "kl": 0.02208251953125, | |
| "learning_rate": 3.2692307692307696e-06, | |
| "loss": 0.0009, | |
| "reward": 1.0651786148548126, | |
| "reward_std": 0.4572626382112503, | |
| "rewards/accuracy_reward": 0.7392857491970062, | |
| "rewards/format_reward": 0.32589287161827085, | |
| "step": 85 | |
| }, | |
| { | |
| "completion_length": 483.4035949707031, | |
| "epoch": 0.0695517774343122, | |
| "grad_norm": 0.6648739576339722, | |
| "kl": 0.02352294921875, | |
| "learning_rate": 3.4615384615384617e-06, | |
| "loss": 0.0009, | |
| "reward": 1.1357143223285675, | |
| "reward_std": 0.3591114327311516, | |
| "rewards/accuracy_reward": 0.7410714626312256, | |
| "rewards/format_reward": 0.3946428716182709, | |
| "step": 90 | |
| }, | |
| { | |
| "completion_length": 460.51698608398436, | |
| "epoch": 0.07341576506955177, | |
| "grad_norm": 0.29094594717025757, | |
| "kl": 0.018267822265625, | |
| "learning_rate": 3.653846153846154e-06, | |
| "loss": 0.0007, | |
| "reward": 1.2437500596046447, | |
| "reward_std": 0.32878804206848145, | |
| "rewards/accuracy_reward": 0.8767857551574707, | |
| "rewards/format_reward": 0.3669643014669418, | |
| "step": 95 | |
| }, | |
| { | |
| "completion_length": 508.53038635253904, | |
| "epoch": 0.07727975270479134, | |
| "grad_norm": 3.374635696411133, | |
| "kl": 0.07086181640625, | |
| "learning_rate": 3.846153846153847e-06, | |
| "loss": 0.0028, | |
| "reward": 1.0633928954601288, | |
| "reward_std": 0.44108102321624754, | |
| "rewards/accuracy_reward": 0.7892857432365418, | |
| "rewards/format_reward": 0.27410715222358706, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.07727975270479134, | |
| "eval_completion_length": 472.33049926757815, | |
| "eval_kl": 0.022151692708333334, | |
| "eval_loss": 0.0008794094319455326, | |
| "eval_reward": 1.147619108359019, | |
| "eval_reward_std": 0.36104824443658196, | |
| "eval_rewards/accuracy_reward": 0.7714286088943482, | |
| "eval_rewards/format_reward": 0.3761904915173849, | |
| "eval_runtime": 91.4389, | |
| "eval_samples_per_second": 1.083, | |
| "eval_steps_per_second": 0.044, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 457.9589447021484, | |
| "epoch": 0.08114374034003091, | |
| "grad_norm": 0.35772719979286194, | |
| "kl": 0.07862548828125, | |
| "learning_rate": 4.0384615384615385e-06, | |
| "loss": 0.0031, | |
| "reward": 1.0741071939468383, | |
| "reward_std": 0.36222362220287324, | |
| "rewards/accuracy_reward": 0.6642857491970062, | |
| "rewards/format_reward": 0.4098214417695999, | |
| "step": 105 | |
| }, | |
| { | |
| "completion_length": 409.0098358154297, | |
| "epoch": 0.08500772797527048, | |
| "grad_norm": 0.4931844472885132, | |
| "kl": 0.0488525390625, | |
| "learning_rate": 4.230769230769231e-06, | |
| "loss": 0.002, | |
| "reward": 1.1589286088943482, | |
| "reward_std": 0.4164972364902496, | |
| "rewards/accuracy_reward": 0.6375000357627869, | |
| "rewards/format_reward": 0.5214285969734191, | |
| "step": 110 | |
| }, | |
| { | |
| "completion_length": 437.12233276367186, | |
| "epoch": 0.08887171561051005, | |
| "grad_norm": 0.31475040316581726, | |
| "kl": 0.0398681640625, | |
| "learning_rate": 4.423076923076924e-06, | |
| "loss": 0.0016, | |
| "reward": 1.1946429133415222, | |
| "reward_std": 0.40721654444932937, | |
| "rewards/accuracy_reward": 0.673214316368103, | |
| "rewards/format_reward": 0.5214285910129547, | |
| "step": 115 | |
| }, | |
| { | |
| "completion_length": 468.64287719726565, | |
| "epoch": 0.09273570324574962, | |
| "grad_norm": 0.3709864318370819, | |
| "kl": 0.0320556640625, | |
| "learning_rate": 4.615384615384616e-06, | |
| "loss": 0.0013, | |
| "reward": 1.2848214983940125, | |
| "reward_std": 0.40001600831747053, | |
| "rewards/accuracy_reward": 0.7464286148548126, | |
| "rewards/format_reward": 0.5383928835391998, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 465.69109497070315, | |
| "epoch": 0.09659969088098919, | |
| "grad_norm": 0.35266467928886414, | |
| "kl": 0.0361328125, | |
| "learning_rate": 4.807692307692308e-06, | |
| "loss": 0.0014, | |
| "reward": 1.304464340209961, | |
| "reward_std": 0.3532786279916763, | |
| "rewards/accuracy_reward": 0.7678571701049804, | |
| "rewards/format_reward": 0.5366071611642838, | |
| "step": 125 | |
| }, | |
| { | |
| "completion_length": 417.57412109375, | |
| "epoch": 0.10046367851622875, | |
| "grad_norm": 0.5035552382469177, | |
| "kl": 0.0549560546875, | |
| "learning_rate": 5e-06, | |
| "loss": 0.0022, | |
| "reward": 1.3687500596046447, | |
| "reward_std": 0.41467164605855944, | |
| "rewards/accuracy_reward": 0.7267857432365418, | |
| "rewards/format_reward": 0.641964316368103, | |
| "step": 130 | |
| }, | |
| { | |
| "completion_length": 449.3339447021484, | |
| "epoch": 0.10432766615146831, | |
| "grad_norm": 0.34655991196632385, | |
| "kl": 0.0698974609375, | |
| "learning_rate": 4.99977236595506e-06, | |
| "loss": 0.0028, | |
| "reward": 1.5678572416305543, | |
| "reward_std": 0.5014733135700226, | |
| "rewards/accuracy_reward": 0.8892857551574707, | |
| "rewards/format_reward": 0.6785714566707611, | |
| "step": 135 | |
| }, | |
| { | |
| "completion_length": 497.0714508056641, | |
| "epoch": 0.10819165378670788, | |
| "grad_norm": 0.6115472316741943, | |
| "kl": 0.090966796875, | |
| "learning_rate": 4.999089505274044e-06, | |
| "loss": 0.0036, | |
| "reward": 1.796428620815277, | |
| "reward_std": 0.5465091168880463, | |
| "rewards/accuracy_reward": 1.0767857611179352, | |
| "rewards/format_reward": 0.7196428894996643, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 506.6348541259766, | |
| "epoch": 0.11205564142194745, | |
| "grad_norm": 0.2844839096069336, | |
| "kl": 0.08369140625, | |
| "learning_rate": 4.9979515423108255e-06, | |
| "loss": 0.0033, | |
| "reward": 1.8116072416305542, | |
| "reward_std": 0.5752836406230927, | |
| "rewards/accuracy_reward": 1.0535714745521545, | |
| "rewards/format_reward": 0.7580357551574707, | |
| "step": 145 | |
| }, | |
| { | |
| "completion_length": 514.6750274658203, | |
| "epoch": 0.11591962905718702, | |
| "grad_norm": 0.32185375690460205, | |
| "kl": 0.0857666015625, | |
| "learning_rate": 4.9963586842966925e-06, | |
| "loss": 0.0034, | |
| "reward": 1.715178644657135, | |
| "reward_std": 0.6052441537380219, | |
| "rewards/accuracy_reward": 1.0142857611179352, | |
| "rewards/format_reward": 0.7008929014205932, | |
| "step": 150 | |
| }, | |
| { | |
| "completion_length": 498.2187835693359, | |
| "epoch": 0.11978361669242658, | |
| "grad_norm": 0.30082836747169495, | |
| "kl": 0.07783203125, | |
| "learning_rate": 4.994311221302617e-06, | |
| "loss": 0.0031, | |
| "reward": 1.80714293718338, | |
| "reward_std": 0.5749644249677658, | |
| "rewards/accuracy_reward": 0.9964285969734192, | |
| "rewards/format_reward": 0.810714328289032, | |
| "step": 155 | |
| }, | |
| { | |
| "completion_length": 458.05091552734376, | |
| "epoch": 0.12364760432766615, | |
| "grad_norm": 0.35644015669822693, | |
| "kl": 0.078955078125, | |
| "learning_rate": 4.991809526186424e-06, | |
| "loss": 0.0032, | |
| "reward": 1.7375000715255737, | |
| "reward_std": 0.6288636207580567, | |
| "rewards/accuracy_reward": 0.9339286148548126, | |
| "rewards/format_reward": 0.80357146859169, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 428.93572998046875, | |
| "epoch": 0.1275115919629057, | |
| "grad_norm": 1.0398685932159424, | |
| "kl": 0.097021484375, | |
| "learning_rate": 4.988854054524897e-06, | |
| "loss": 0.0039, | |
| "reward": 1.769642949104309, | |
| "reward_std": 0.6159187257289886, | |
| "rewards/accuracy_reward": 0.9625000476837158, | |
| "rewards/format_reward": 0.8071428954601287, | |
| "step": 165 | |
| }, | |
| { | |
| "completion_length": 447.8910888671875, | |
| "epoch": 0.13137557959814528, | |
| "grad_norm": 0.8748766779899597, | |
| "kl": 0.141357421875, | |
| "learning_rate": 4.985445344530811e-06, | |
| "loss": 0.0057, | |
| "reward": 1.7616072177886963, | |
| "reward_std": 0.6195760637521743, | |
| "rewards/accuracy_reward": 0.9589286148548126, | |
| "rewards/format_reward": 0.8026786029338837, | |
| "step": 170 | |
| }, | |
| { | |
| "completion_length": 443.47234497070315, | |
| "epoch": 0.13523956723338484, | |
| "grad_norm": 3.474867582321167, | |
| "kl": 0.24501953125, | |
| "learning_rate": 4.9815840169549216e-06, | |
| "loss": 0.0098, | |
| "reward": 1.7017857909202576, | |
| "reward_std": 0.6981926560401917, | |
| "rewards/accuracy_reward": 0.8839286029338836, | |
| "rewards/format_reward": 0.8178571879863739, | |
| "step": 175 | |
| }, | |
| { | |
| "completion_length": 463.7705505371094, | |
| "epoch": 0.1391035548686244, | |
| "grad_norm": 20.703824996948242, | |
| "kl": 0.319140625, | |
| "learning_rate": 4.9772707749729205e-06, | |
| "loss": 0.0128, | |
| "reward": 1.5928572058677672, | |
| "reward_std": 0.7676066577434539, | |
| "rewards/accuracy_reward": 0.8089286029338837, | |
| "rewards/format_reward": 0.7839286029338837, | |
| "step": 180 | |
| }, | |
| { | |
| "completion_length": 364.6160858154297, | |
| "epoch": 0.14296754250386398, | |
| "grad_norm": 7.634427070617676, | |
| "kl": 2.84404296875, | |
| "learning_rate": 4.9725064040573824e-06, | |
| "loss": 0.114, | |
| "reward": 1.6312500715255738, | |
| "reward_std": 0.7480017244815826, | |
| "rewards/accuracy_reward": 0.8339286148548126, | |
| "rewards/format_reward": 0.7973214685916901, | |
| "step": 185 | |
| }, | |
| { | |
| "completion_length": 301.68394470214844, | |
| "epoch": 0.14683153013910355, | |
| "grad_norm": 119.83277130126953, | |
| "kl": 7.9796875, | |
| "learning_rate": 4.967291771834727e-06, | |
| "loss": 0.3199, | |
| "reward": 1.4017857670783997, | |
| "reward_std": 0.9306474328041077, | |
| "rewards/accuracy_reward": 0.7446428894996643, | |
| "rewards/format_reward": 0.6571428954601288, | |
| "step": 190 | |
| }, | |
| { | |
| "completion_length": 398.2705535888672, | |
| "epoch": 0.15069551777434312, | |
| "grad_norm": 7.846933841705322, | |
| "kl": 2.8125, | |
| "learning_rate": 4.961627827927214e-06, | |
| "loss": 0.1125, | |
| "reward": 1.2562500476837157, | |
| "reward_std": 0.8410393178462983, | |
| "rewards/accuracy_reward": 0.6660714626312256, | |
| "rewards/format_reward": 0.5901785910129547, | |
| "step": 195 | |
| }, | |
| { | |
| "completion_length": 452.414306640625, | |
| "epoch": 0.1545595054095827, | |
| "grad_norm": 13.303293228149414, | |
| "kl": 3.492578125, | |
| "learning_rate": 4.955515603780013e-06, | |
| "loss": 0.1395, | |
| "reward": 1.2098214864730834, | |
| "reward_std": 0.91947420835495, | |
| "rewards/accuracy_reward": 0.7017857491970062, | |
| "rewards/format_reward": 0.508035734295845, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.1545595054095827, | |
| "eval_completion_length": 502.390342203776, | |
| "eval_kl": 2.55625, | |
| "eval_loss": 0.1018948182463646, | |
| "eval_reward": 1.0833333810170491, | |
| "eval_reward_std": 0.8227311591307322, | |
| "eval_rewards/accuracy_reward": 0.7095238486925761, | |
| "eval_rewards/format_reward": 0.3738095432519913, | |
| "eval_runtime": 97.2526, | |
| "eval_samples_per_second": 1.018, | |
| "eval_steps_per_second": 0.041, | |
| "step": 200 | |
| }, | |
| { | |
| "completion_length": 485.92323913574216, | |
| "epoch": 0.15842349304482226, | |
| "grad_norm": 1.919914722442627, | |
| "kl": 3892.1974609375, | |
| "learning_rate": 4.948956212473371e-06, | |
| "loss": 156.0877, | |
| "reward": 1.029464328289032, | |
| "reward_std": 0.8151026546955109, | |
| "rewards/accuracy_reward": 0.6910714685916901, | |
| "rewards/format_reward": 0.33839287161827086, | |
| "step": 205 | |
| }, | |
| { | |
| "completion_length": 452.31519470214846, | |
| "epoch": 0.16228748068006182, | |
| "grad_norm": 3.2309248447418213, | |
| "kl": 2.9755859375, | |
| "learning_rate": 4.9419508485199045e-06, | |
| "loss": 0.1187, | |
| "reward": 1.3482143521308898, | |
| "reward_std": 0.8349850118160248, | |
| "rewards/accuracy_reward": 0.7946428954601288, | |
| "rewards/format_reward": 0.5535714477300644, | |
| "step": 210 | |
| }, | |
| { | |
| "completion_length": 420.15180053710935, | |
| "epoch": 0.1661514683153014, | |
| "grad_norm": 0.746425211429596, | |
| "kl": 0.511328125, | |
| "learning_rate": 4.934500787647083e-06, | |
| "loss": 0.0205, | |
| "reward": 1.7473215103149413, | |
| "reward_std": 0.6757471442222596, | |
| "rewards/accuracy_reward": 0.998214328289032, | |
| "rewards/format_reward": 0.7491071820259094, | |
| "step": 215 | |
| }, | |
| { | |
| "completion_length": 430.4910919189453, | |
| "epoch": 0.17001545595054096, | |
| "grad_norm": 1.5358840227127075, | |
| "kl": 0.2576171875, | |
| "learning_rate": 4.926607386564898e-06, | |
| "loss": 0.0103, | |
| "reward": 1.8616072297096253, | |
| "reward_std": 0.5690989345312119, | |
| "rewards/accuracy_reward": 1.0482143342494965, | |
| "rewards/format_reward": 0.8133928894996643, | |
| "step": 220 | |
| }, | |
| { | |
| "completion_length": 480.8393035888672, | |
| "epoch": 0.17387944358578053, | |
| "grad_norm": 0.8902988433837891, | |
| "kl": 736.20126953125, | |
| "learning_rate": 4.918272082718805e-06, | |
| "loss": 29.354, | |
| "reward": 1.9142858028411864, | |
| "reward_std": 0.5961195319890976, | |
| "rewards/accuracy_reward": 1.0714286267757416, | |
| "rewards/format_reward": 0.8428571879863739, | |
| "step": 225 | |
| }, | |
| { | |
| "completion_length": 521.7080596923828, | |
| "epoch": 0.1777434312210201, | |
| "grad_norm": 0.7753103971481323, | |
| "kl": 0.2408203125, | |
| "learning_rate": 4.909496394027945e-06, | |
| "loss": 0.0096, | |
| "reward": 1.6517857789993287, | |
| "reward_std": 0.6544238030910492, | |
| "rewards/accuracy_reward": 0.9446429014205933, | |
| "rewards/format_reward": 0.7071428954601288, | |
| "step": 230 | |
| }, | |
| { | |
| "completion_length": 471.0625274658203, | |
| "epoch": 0.18160741885625967, | |
| "grad_norm": 0.9643970131874084, | |
| "kl": 1.3505859375, | |
| "learning_rate": 4.900281918608732e-06, | |
| "loss": 0.054, | |
| "reward": 1.738392949104309, | |
| "reward_std": 0.6436418563127517, | |
| "rewards/accuracy_reward": 1.0196429133415221, | |
| "rewards/format_reward": 0.7187500357627868, | |
| "step": 235 | |
| }, | |
| { | |
| "completion_length": 397.68840942382815, | |
| "epoch": 0.18547140649149924, | |
| "grad_norm": 1.6753307580947876, | |
| "kl": 0.4291015625, | |
| "learning_rate": 4.890630334483814e-06, | |
| "loss": 0.0172, | |
| "reward": 1.8008929371833802, | |
| "reward_std": 0.610142993927002, | |
| "rewards/accuracy_reward": 0.996428620815277, | |
| "rewards/format_reward": 0.8044643223285675, | |
| "step": 240 | |
| }, | |
| { | |
| "completion_length": 399.38930358886716, | |
| "epoch": 0.1893353941267388, | |
| "grad_norm": 2.494441509246826, | |
| "kl": 17.448046875, | |
| "learning_rate": 4.880543399276499e-06, | |
| "loss": 0.7002, | |
| "reward": 1.5107143640518188, | |
| "reward_std": 0.76023770570755, | |
| "rewards/accuracy_reward": 0.7892857491970062, | |
| "rewards/format_reward": 0.7214286029338837, | |
| "step": 245 | |
| }, | |
| { | |
| "completion_length": 375.9785858154297, | |
| "epoch": 0.19319938176197837, | |
| "grad_norm": 2.4901890754699707, | |
| "kl": 1.29296875, | |
| "learning_rate": 4.870022949890676e-06, | |
| "loss": 0.0517, | |
| "reward": 1.6321429491043091, | |
| "reward_std": 0.6653113335371017, | |
| "rewards/accuracy_reward": 0.8446428894996643, | |
| "rewards/format_reward": 0.7875000357627868, | |
| "step": 250 | |
| }, | |
| { | |
| "completion_length": 368.9598388671875, | |
| "epoch": 0.19706336939721794, | |
| "grad_norm": 2.3111751079559326, | |
| "kl": 0.7330078125, | |
| "learning_rate": 4.859070902176305e-06, | |
| "loss": 0.0293, | |
| "reward": 1.8500001072883605, | |
| "reward_std": 0.5273477554321289, | |
| "rewards/accuracy_reward": 0.9571428954601288, | |
| "rewards/format_reward": 0.8928571820259095, | |
| "step": 255 | |
| }, | |
| { | |
| "completion_length": 392.04912719726565, | |
| "epoch": 0.2009273570324575, | |
| "grad_norm": 1.6753560304641724, | |
| "kl": 3.5484375, | |
| "learning_rate": 4.8476892505805224e-06, | |
| "loss": 0.1425, | |
| "reward": 1.8562500715255736, | |
| "reward_std": 0.5729426324367524, | |
| "rewards/accuracy_reward": 1.0071429133415222, | |
| "rewards/format_reward": 0.849107164144516, | |
| "step": 260 | |
| }, | |
| { | |
| "completion_length": 461.39019470214845, | |
| "epoch": 0.20479134466769705, | |
| "grad_norm": 7.992901802062988, | |
| "kl": 0.6904296875, | |
| "learning_rate": 4.835880067784441e-06, | |
| "loss": 0.0276, | |
| "reward": 1.7035714983940125, | |
| "reward_std": 0.814025753736496, | |
| "rewards/accuracy_reward": 0.9500000417232514, | |
| "rewards/format_reward": 0.7535714566707611, | |
| "step": 265 | |
| }, | |
| { | |
| "completion_length": 400.0518035888672, | |
| "epoch": 0.20865533230293662, | |
| "grad_norm": 16.02567481994629, | |
| "kl": 0.78994140625, | |
| "learning_rate": 4.823645504325699e-06, | |
| "loss": 0.0317, | |
| "reward": 1.8741072297096253, | |
| "reward_std": 0.6473642587661743, | |
| "rewards/accuracy_reward": 1.0446428954601288, | |
| "rewards/format_reward": 0.8294643223285675, | |
| "step": 270 | |
| }, | |
| { | |
| "completion_length": 366.62322998046875, | |
| "epoch": 0.2125193199381762, | |
| "grad_norm": 3.842503547668457, | |
| "kl": 1.327734375, | |
| "learning_rate": 4.81098778820683e-06, | |
| "loss": 0.0531, | |
| "reward": 1.9205357909202576, | |
| "reward_std": 0.5114573985338211, | |
| "rewards/accuracy_reward": 1.0375000476837157, | |
| "rewards/format_reward": 0.8830357611179351, | |
| "step": 275 | |
| }, | |
| { | |
| "completion_length": 419.6143005371094, | |
| "epoch": 0.21638330757341576, | |
| "grad_norm": 1.8317447900772095, | |
| "kl": 1.608203125, | |
| "learning_rate": 4.797909224489531e-06, | |
| "loss": 0.0644, | |
| "reward": 1.725892925262451, | |
| "reward_std": 0.7076287031173706, | |
| "rewards/accuracy_reward": 0.9500000417232514, | |
| "rewards/format_reward": 0.7758929014205933, | |
| "step": 280 | |
| }, | |
| { | |
| "completion_length": 462.8062713623047, | |
| "epoch": 0.22024729520865532, | |
| "grad_norm": 3.1712355613708496, | |
| "kl": 2.66015625, | |
| "learning_rate": 4.7844121948748904e-06, | |
| "loss": 0.1064, | |
| "reward": 1.6767858028411866, | |
| "reward_std": 0.6928595781326294, | |
| "rewards/accuracy_reward": 0.9517857491970062, | |
| "rewards/format_reward": 0.7250000417232514, | |
| "step": 285 | |
| }, | |
| { | |
| "completion_length": 419.5089477539062, | |
| "epoch": 0.2241112828438949, | |
| "grad_norm": 4.4721269607543945, | |
| "kl": 2.098828125, | |
| "learning_rate": 4.770499157269664e-06, | |
| "loss": 0.084, | |
| "reward": 1.8848215460777282, | |
| "reward_std": 0.6089197903871536, | |
| "rewards/accuracy_reward": 1.0571428894996644, | |
| "rewards/format_reward": 0.8276786029338836, | |
| "step": 290 | |
| }, | |
| { | |
| "completion_length": 382.21519470214844, | |
| "epoch": 0.22797527047913446, | |
| "grad_norm": 3.2568917274475098, | |
| "kl": 1.37666015625, | |
| "learning_rate": 4.756172645338675e-06, | |
| "loss": 0.0551, | |
| "reward": 1.9062500953674317, | |
| "reward_std": 0.5363048523664474, | |
| "rewards/accuracy_reward": 1.0035714626312255, | |
| "rewards/format_reward": 0.9026785969734192, | |
| "step": 295 | |
| }, | |
| { | |
| "completion_length": 400.94912109375, | |
| "epoch": 0.23183925811437403, | |
| "grad_norm": 12.386054992675781, | |
| "kl": 2.53203125, | |
| "learning_rate": 4.741435268043412e-06, | |
| "loss": 0.1013, | |
| "reward": 1.702678656578064, | |
| "reward_std": 0.7214795827865601, | |
| "rewards/accuracy_reward": 0.941071480512619, | |
| "rewards/format_reward": 0.7616071760654449, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.23183925811437403, | |
| "eval_completion_length": 454.9096964518229, | |
| "eval_kl": 4.954166666666667, | |
| "eval_loss": 0.19525596499443054, | |
| "eval_reward": 1.3000000516573589, | |
| "eval_reward_std": 0.8129177749156952, | |
| "eval_rewards/accuracy_reward": 0.7142857472101848, | |
| "eval_rewards/format_reward": 0.585714316368103, | |
| "eval_runtime": 98.4065, | |
| "eval_samples_per_second": 1.006, | |
| "eval_steps_per_second": 0.041, | |
| "step": 300 | |
| }, | |
| { | |
| "completion_length": 470.2223419189453, | |
| "epoch": 0.2357032457496136, | |
| "grad_norm": 3.8275136947631836, | |
| "kl": 2.878125, | |
| "learning_rate": 4.7262897091669195e-06, | |
| "loss": 0.1152, | |
| "reward": 1.290178620815277, | |
| "reward_std": 0.830661517381668, | |
| "rewards/accuracy_reward": 0.6928571701049805, | |
| "rewards/format_reward": 0.5973214536905289, | |
| "step": 305 | |
| }, | |
| { | |
| "completion_length": 436.9973388671875, | |
| "epoch": 0.23956723338485317, | |
| "grad_norm": 2.0655641555786133, | |
| "kl": 2.27578125, | |
| "learning_rate": 4.710738726825059e-06, | |
| "loss": 0.091, | |
| "reward": 1.608928644657135, | |
| "reward_std": 0.8151212155818939, | |
| "rewards/accuracy_reward": 0.8553571701049805, | |
| "rewards/format_reward": 0.7535714507102966, | |
| "step": 310 | |
| }, | |
| { | |
| "completion_length": 363.5821594238281, | |
| "epoch": 0.24343122102009274, | |
| "grad_norm": 1.809662103652954, | |
| "kl": 1.327734375, | |
| "learning_rate": 4.694785152964244e-06, | |
| "loss": 0.0531, | |
| "reward": 1.8464286565780639, | |
| "reward_std": 0.5842416912317276, | |
| "rewards/accuracy_reward": 0.9767857551574707, | |
| "rewards/format_reward": 0.8696429014205933, | |
| "step": 315 | |
| }, | |
| { | |
| "completion_length": 341.5821563720703, | |
| "epoch": 0.2472952086553323, | |
| "grad_norm": 2.9773175716400146, | |
| "kl": 2.2734375, | |
| "learning_rate": 4.678431892845714e-06, | |
| "loss": 0.0908, | |
| "reward": 1.8232143759727477, | |
| "reward_std": 0.6686225771903992, | |
| "rewards/accuracy_reward": 0.9571428954601288, | |
| "rewards/format_reward": 0.8660714745521545, | |
| "step": 320 | |
| }, | |
| { | |
| "completion_length": 362.4410827636719, | |
| "epoch": 0.2511591962905719, | |
| "grad_norm": 1.7058671712875366, | |
| "kl": 1.4408203125, | |
| "learning_rate": 4.661681924516466e-06, | |
| "loss": 0.0577, | |
| "reward": 1.9660715341567994, | |
| "reward_std": 0.589709809422493, | |
| "rewards/accuracy_reward": 1.0446429073810577, | |
| "rewards/format_reward": 0.9214286148548126, | |
| "step": 325 | |
| }, | |
| { | |
| "completion_length": 380.65447692871095, | |
| "epoch": 0.2550231839258114, | |
| "grad_norm": 0.6990386843681335, | |
| "kl": 0.7595703125, | |
| "learning_rate": 4.6445382982669365e-06, | |
| "loss": 0.0304, | |
| "reward": 2.0437501072883606, | |
| "reward_std": 0.5305197536945343, | |
| "rewards/accuracy_reward": 1.1089286267757417, | |
| "rewards/format_reward": 0.9348214745521546, | |
| "step": 330 | |
| }, | |
| { | |
| "completion_length": 376.0223419189453, | |
| "epoch": 0.258887171561051, | |
| "grad_norm": 182.0006866455078, | |
| "kl": 1.23125, | |
| "learning_rate": 4.627004136075514e-06, | |
| "loss": 0.0492, | |
| "reward": 2.0428572416305544, | |
| "reward_std": 0.44865317046642306, | |
| "rewards/accuracy_reward": 1.0910714745521546, | |
| "rewards/format_reward": 0.9517857491970062, | |
| "step": 335 | |
| }, | |
| { | |
| "completion_length": 385.6750122070313, | |
| "epoch": 0.26275115919629055, | |
| "grad_norm": 0.5571011900901794, | |
| "kl": 0.4580078125, | |
| "learning_rate": 4.609082631040012e-06, | |
| "loss": 0.0183, | |
| "reward": 2.083928680419922, | |
| "reward_std": 0.48743111491203306, | |
| "rewards/accuracy_reward": 1.1214286267757416, | |
| "rewards/format_reward": 0.9625000357627869, | |
| "step": 340 | |
| }, | |
| { | |
| "completion_length": 376.5723388671875, | |
| "epoch": 0.26661514683153015, | |
| "grad_norm": 1.1355966329574585, | |
| "kl": 0.5251953125, | |
| "learning_rate": 4.5907770467961755e-06, | |
| "loss": 0.021, | |
| "reward": 1.9705358266830444, | |
| "reward_std": 0.5101535975933075, | |
| "rewards/accuracy_reward": 1.0375000417232514, | |
| "rewards/format_reward": 0.9330357611179352, | |
| "step": 345 | |
| }, | |
| { | |
| "completion_length": 462.158056640625, | |
| "epoch": 0.2704791344667697, | |
| "grad_norm": 1.7409440279006958, | |
| "kl": 1.1546875, | |
| "learning_rate": 4.572090716923354e-06, | |
| "loss": 0.0462, | |
| "reward": 1.6875000715255737, | |
| "reward_std": 0.7979255437850952, | |
| "rewards/accuracy_reward": 0.8857143223285675, | |
| "rewards/format_reward": 0.8017857491970062, | |
| "step": 350 | |
| }, | |
| { | |
| "completion_length": 482.0419860839844, | |
| "epoch": 0.2743431221020093, | |
| "grad_norm": 4.6754655838012695, | |
| "kl": 1.208203125, | |
| "learning_rate": 4.5530270443374305e-06, | |
| "loss": 0.0484, | |
| "reward": 1.471428632736206, | |
| "reward_std": 0.8514433205127716, | |
| "rewards/accuracy_reward": 0.7285714626312256, | |
| "rewards/format_reward": 0.7428571701049804, | |
| "step": 355 | |
| }, | |
| { | |
| "completion_length": 484.2116333007813, | |
| "epoch": 0.2782071097372488, | |
| "grad_norm": 11.937480926513672, | |
| "kl": 5.473828125, | |
| "learning_rate": 4.533589500671126e-06, | |
| "loss": 0.219, | |
| "reward": 1.4812500596046447, | |
| "reward_std": 0.8435086131095886, | |
| "rewards/accuracy_reward": 0.7446428775787354, | |
| "rewards/format_reward": 0.7366071760654449, | |
| "step": 360 | |
| }, | |
| { | |
| "completion_length": 538.0169860839844, | |
| "epoch": 0.2820710973724884, | |
| "grad_norm": 1.5927709341049194, | |
| "kl": 2.071875, | |
| "learning_rate": 4.513781625641793e-06, | |
| "loss": 0.0829, | |
| "reward": 1.3562500596046447, | |
| "reward_std": 0.9622864723205566, | |
| "rewards/accuracy_reward": 0.7089285969734191, | |
| "rewards/format_reward": 0.6473214566707611, | |
| "step": 365 | |
| }, | |
| { | |
| "completion_length": 498.48663330078125, | |
| "epoch": 0.28593508500772796, | |
| "grad_norm": 2.080902338027954, | |
| "kl": 3.178125, | |
| "learning_rate": 4.493607026406802e-06, | |
| "loss": 0.1272, | |
| "reward": 1.2741071939468385, | |
| "reward_std": 0.8825342118740082, | |
| "rewards/accuracy_reward": 0.6517857432365417, | |
| "rewards/format_reward": 0.6223214626312256, | |
| "step": 370 | |
| }, | |
| { | |
| "completion_length": 489.34288330078124, | |
| "epoch": 0.28979907264296756, | |
| "grad_norm": 43.23033142089844, | |
| "kl": 8.0921875, | |
| "learning_rate": 4.473069376906657e-06, | |
| "loss": 0.3234, | |
| "reward": 1.111607199907303, | |
| "reward_std": 0.8139643251895905, | |
| "rewards/accuracy_reward": 0.5375000178813935, | |
| "rewards/format_reward": 0.5741071790456772, | |
| "step": 375 | |
| }, | |
| { | |
| "completion_length": 547.9125305175781, | |
| "epoch": 0.2936630602782071, | |
| "grad_norm": 3.59335994720459, | |
| "kl": 2.996875, | |
| "learning_rate": 4.4521724171959404e-06, | |
| "loss": 0.1198, | |
| "reward": 0.965178620815277, | |
| "reward_std": 0.776919960975647, | |
| "rewards/accuracy_reward": 0.47142858505249025, | |
| "rewards/format_reward": 0.4937500238418579, | |
| "step": 380 | |
| }, | |
| { | |
| "completion_length": 447.9116271972656, | |
| "epoch": 0.2975270479134467, | |
| "grad_norm": 10.616156578063965, | |
| "kl": 2.9203125, | |
| "learning_rate": 4.430919952762226e-06, | |
| "loss": 0.1167, | |
| "reward": 1.7098215222358704, | |
| "reward_std": 0.7615076899528503, | |
| "rewards/accuracy_reward": 0.9089286088943481, | |
| "rewards/format_reward": 0.8008928894996643, | |
| "step": 385 | |
| }, | |
| { | |
| "completion_length": 408.0750213623047, | |
| "epoch": 0.30139103554868624, | |
| "grad_norm": 2.474897623062134, | |
| "kl": 1.240625, | |
| "learning_rate": 4.409315853833068e-06, | |
| "loss": 0.0495, | |
| "reward": 1.898214375972748, | |
| "reward_std": 0.6294975191354751, | |
| "rewards/accuracy_reward": 1.0250000357627869, | |
| "rewards/format_reward": 0.873214328289032, | |
| "step": 390 | |
| }, | |
| { | |
| "completion_length": 432.92323608398436, | |
| "epoch": 0.30525502318392583, | |
| "grad_norm": 1.0525386333465576, | |
| "kl": 0.984765625, | |
| "learning_rate": 4.387364054671208e-06, | |
| "loss": 0.0394, | |
| "reward": 1.804464375972748, | |
| "reward_std": 0.7036986917257309, | |
| "rewards/accuracy_reward": 0.9642857491970063, | |
| "rewards/format_reward": 0.8401786148548126, | |
| "step": 395 | |
| }, | |
| { | |
| "completion_length": 426.0428771972656, | |
| "epoch": 0.3091190108191654, | |
| "grad_norm": 28.755624771118164, | |
| "kl": 0.57890625, | |
| "learning_rate": 4.365068552858116e-06, | |
| "loss": 0.0232, | |
| "reward": 1.8732143640518188, | |
| "reward_std": 0.5758820742368698, | |
| "rewards/accuracy_reward": 1.021428632736206, | |
| "rewards/format_reward": 0.8517857551574707, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.3091190108191654, | |
| "eval_completion_length": 421.4955749511719, | |
| "eval_kl": 0.2569661458333333, | |
| "eval_loss": 0.01039376575499773, | |
| "eval_reward": 1.8190476894378662, | |
| "eval_reward_std": 0.5760457158088684, | |
| "eval_rewards/accuracy_reward": 0.9333333849906922, | |
| "eval_rewards/format_reward": 0.885714328289032, | |
| "eval_runtime": 92.0979, | |
| "eval_samples_per_second": 1.075, | |
| "eval_steps_per_second": 0.043, | |
| "step": 400 | |
| }, | |
| { | |
| "completion_length": 431.314306640625, | |
| "epoch": 0.31298299845440497, | |
| "grad_norm": 1.367416262626648, | |
| "kl": 0.28056640625, | |
| "learning_rate": 4.342433408566e-06, | |
| "loss": 0.0112, | |
| "reward": 1.8312501072883607, | |
| "reward_std": 0.6301711022853851, | |
| "rewards/accuracy_reward": 0.9732143342494964, | |
| "rewards/format_reward": 0.8580357611179352, | |
| "step": 405 | |
| }, | |
| { | |
| "completion_length": 430.4732360839844, | |
| "epoch": 0.3168469860896445, | |
| "grad_norm": 0.7723910212516785, | |
| "kl": 0.4091796875, | |
| "learning_rate": 4.3194627438184235e-06, | |
| "loss": 0.0164, | |
| "reward": 1.748214340209961, | |
| "reward_std": 0.6911887288093567, | |
| "rewards/accuracy_reward": 0.9267857611179352, | |
| "rewards/format_reward": 0.8214286088943481, | |
| "step": 410 | |
| }, | |
| { | |
| "completion_length": 524.7294921875, | |
| "epoch": 0.3207109737248841, | |
| "grad_norm": 1.0905834436416626, | |
| "kl": 0.517578125, | |
| "learning_rate": 4.296160741739652e-06, | |
| "loss": 0.0207, | |
| "reward": 1.440178632736206, | |
| "reward_std": 0.7554588854312897, | |
| "rewards/accuracy_reward": 0.7535714626312255, | |
| "rewards/format_reward": 0.686607176065445, | |
| "step": 415 | |
| }, | |
| { | |
| "completion_length": 449.8026947021484, | |
| "epoch": 0.32457496136012365, | |
| "grad_norm": 0.6751212477684021, | |
| "kl": 0.26748046875, | |
| "learning_rate": 4.272531645792876e-06, | |
| "loss": 0.0107, | |
| "reward": 1.554464375972748, | |
| "reward_std": 0.7342777848243713, | |
| "rewards/accuracy_reward": 0.7571428775787353, | |
| "rewards/format_reward": 0.7973214685916901, | |
| "step": 420 | |
| }, | |
| { | |
| "completion_length": 425.3678771972656, | |
| "epoch": 0.3284389489953632, | |
| "grad_norm": 0.857072651386261, | |
| "kl": 0.311328125, | |
| "learning_rate": 4.2485797590074465e-06, | |
| "loss": 0.0124, | |
| "reward": 1.7098215222358704, | |
| "reward_std": 0.6917063415050506, | |
| "rewards/accuracy_reward": 0.8803571879863739, | |
| "rewards/format_reward": 0.829464316368103, | |
| "step": 425 | |
| }, | |
| { | |
| "completion_length": 368.4768035888672, | |
| "epoch": 0.3323029366306028, | |
| "grad_norm": 1.9811636209487915, | |
| "kl": 0.26328125, | |
| "learning_rate": 4.224309443195261e-06, | |
| "loss": 0.0105, | |
| "reward": 1.8812500953674316, | |
| "reward_std": 0.5524741142988205, | |
| "rewards/accuracy_reward": 0.9732143342494964, | |
| "rewards/format_reward": 0.9080357611179352, | |
| "step": 430 | |
| }, | |
| { | |
| "completion_length": 443.5598419189453, | |
| "epoch": 0.3361669242658423, | |
| "grad_norm": 1.1125954389572144, | |
| "kl": 0.38134765625, | |
| "learning_rate": 4.199725118156448e-06, | |
| "loss": 0.0152, | |
| "reward": 1.6392857909202576, | |
| "reward_std": 0.662892284989357, | |
| "rewards/accuracy_reward": 0.8517857611179351, | |
| "rewards/format_reward": 0.7875000298023224, | |
| "step": 435 | |
| }, | |
| { | |
| "completion_length": 445.95270080566405, | |
| "epoch": 0.3400309119010819, | |
| "grad_norm": 1.947538137435913, | |
| "kl": 0.4873046875, | |
| "learning_rate": 4.174831260874489e-06, | |
| "loss": 0.0195, | |
| "reward": 1.7642858147621154, | |
| "reward_std": 0.8089473009109497, | |
| "rewards/accuracy_reward": 0.9946429014205933, | |
| "rewards/format_reward": 0.7696428954601288, | |
| "step": 440 | |
| }, | |
| { | |
| "completion_length": 463.2759094238281, | |
| "epoch": 0.34389489953632146, | |
| "grad_norm": 0.44283971190452576, | |
| "kl": 0.41533203125, | |
| "learning_rate": 4.149632404700925e-06, | |
| "loss": 0.0166, | |
| "reward": 1.7160715103149413, | |
| "reward_std": 0.7051611065864563, | |
| "rewards/accuracy_reward": 0.9357143342494965, | |
| "rewards/format_reward": 0.780357176065445, | |
| "step": 445 | |
| }, | |
| { | |
| "completion_length": 396.64288024902345, | |
| "epoch": 0.34775888717156106, | |
| "grad_norm": 0.5284359455108643, | |
| "kl": 0.1916015625, | |
| "learning_rate": 4.124133138529804e-06, | |
| "loss": 0.0077, | |
| "reward": 1.9616072297096252, | |
| "reward_std": 0.5300792083144188, | |
| "rewards/accuracy_reward": 1.046428620815277, | |
| "rewards/format_reward": 0.9151786208152771, | |
| "step": 450 | |
| }, | |
| { | |
| "completion_length": 359.92144165039065, | |
| "epoch": 0.3516228748068006, | |
| "grad_norm": 0.6384701728820801, | |
| "kl": 0.18466796875, | |
| "learning_rate": 4.098338105962004e-06, | |
| "loss": 0.0074, | |
| "reward": 2.0705358505249025, | |
| "reward_std": 0.48216700553894043, | |
| "rewards/accuracy_reward": 1.1160714685916902, | |
| "rewards/format_reward": 0.9544643223285675, | |
| "step": 455 | |
| }, | |
| { | |
| "completion_length": 353.79465942382814, | |
| "epoch": 0.3554868624420402, | |
| "grad_norm": 2.704573154449463, | |
| "kl": 0.2443359375, | |
| "learning_rate": 4.072252004459612e-06, | |
| "loss": 0.0098, | |
| "reward": 1.977678644657135, | |
| "reward_std": 0.511484894156456, | |
| "rewards/accuracy_reward": 1.0285714864730835, | |
| "rewards/format_reward": 0.9491071760654449, | |
| "step": 460 | |
| }, | |
| { | |
| "completion_length": 352.4866180419922, | |
| "epoch": 0.35935085007727974, | |
| "grad_norm": 1.346289038658142, | |
| "kl": 0.3484375, | |
| "learning_rate": 4.045879584490466e-06, | |
| "loss": 0.0139, | |
| "reward": 1.8348215341567993, | |
| "reward_std": 0.6451143264770508, | |
| "rewards/accuracy_reward": 0.917857187986374, | |
| "rewards/format_reward": 0.916964328289032, | |
| "step": 465 | |
| }, | |
| { | |
| "completion_length": 454.8107391357422, | |
| "epoch": 0.36321483771251933, | |
| "grad_norm": 0.8649754524230957, | |
| "kl": 0.369140625, | |
| "learning_rate": 4.019225648663073e-06, | |
| "loss": 0.0148, | |
| "reward": 1.560714340209961, | |
| "reward_std": 0.811360216140747, | |
| "rewards/accuracy_reward": 0.7607143223285675, | |
| "rewards/format_reward": 0.8000000298023224, | |
| "step": 470 | |
| }, | |
| { | |
| "completion_length": 382.1169830322266, | |
| "epoch": 0.3670788253477589, | |
| "grad_norm": 2.951951742172241, | |
| "kl": 0.2439453125, | |
| "learning_rate": 3.992295050852013e-06, | |
| "loss": 0.0098, | |
| "reward": 1.8357143878936768, | |
| "reward_std": 0.6485567986965179, | |
| "rewards/accuracy_reward": 0.9232143223285675, | |
| "rewards/format_reward": 0.9125000476837158, | |
| "step": 475 | |
| }, | |
| { | |
| "completion_length": 377.8848358154297, | |
| "epoch": 0.37094281298299847, | |
| "grad_norm": 3.4858837127685547, | |
| "kl": 0.23447265625, | |
| "learning_rate": 3.965092695314018e-06, | |
| "loss": 0.0094, | |
| "reward": 1.8285714983940125, | |
| "reward_std": 0.5620756894350052, | |
| "rewards/accuracy_reward": 0.8892857611179352, | |
| "rewards/format_reward": 0.9392857670783996, | |
| "step": 480 | |
| }, | |
| { | |
| "completion_length": 420.1151947021484, | |
| "epoch": 0.374806800618238, | |
| "grad_norm": 0.7943024635314941, | |
| "kl": 0.3015625, | |
| "learning_rate": 3.937623535794864e-06, | |
| "loss": 0.0121, | |
| "reward": 1.8723214864730835, | |
| "reward_std": 0.5844464153051376, | |
| "rewards/accuracy_reward": 0.9607143461704254, | |
| "rewards/format_reward": 0.911607176065445, | |
| "step": 485 | |
| }, | |
| { | |
| "completion_length": 476.45448608398436, | |
| "epoch": 0.3786707882534776, | |
| "grad_norm": 0.619339108467102, | |
| "kl": 0.3705078125, | |
| "learning_rate": 3.909892574627267e-06, | |
| "loss": 0.0148, | |
| "reward": 1.7160714983940124, | |
| "reward_std": 0.6855497658252716, | |
| "rewards/accuracy_reward": 0.8964286088943482, | |
| "rewards/format_reward": 0.8196428835391998, | |
| "step": 490 | |
| }, | |
| { | |
| "completion_length": 449.3410919189453, | |
| "epoch": 0.38253477588871715, | |
| "grad_norm": 0.9153338074684143, | |
| "kl": 0.2828125, | |
| "learning_rate": 3.881904861819914e-06, | |
| "loss": 0.0113, | |
| "reward": 1.7375000834465026, | |
| "reward_std": 0.773430997133255, | |
| "rewards/accuracy_reward": 0.9160714864730835, | |
| "rewards/format_reward": 0.8214286029338836, | |
| "step": 495 | |
| }, | |
| { | |
| "completion_length": 424.5857330322266, | |
| "epoch": 0.38639876352395675, | |
| "grad_norm": 2.4771955013275146, | |
| "kl": 0.30927734375, | |
| "learning_rate": 3.853665494137825e-06, | |
| "loss": 0.0124, | |
| "reward": 1.7491072177886964, | |
| "reward_std": 0.6954550087451935, | |
| "rewards/accuracy_reward": 0.8785714685916901, | |
| "rewards/format_reward": 0.8705357670783996, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.38639876352395675, | |
| "eval_completion_length": 375.0712870279948, | |
| "eval_kl": 0.3849609375, | |
| "eval_loss": 0.015607084147632122, | |
| "eval_reward": 1.8738096157709758, | |
| "eval_reward_std": 0.5338563899199168, | |
| "eval_rewards/accuracy_reward": 0.9523809909820556, | |
| "eval_rewards/format_reward": 0.9214285969734192, | |
| "eval_runtime": 89.8262, | |
| "eval_samples_per_second": 1.102, | |
| "eval_steps_per_second": 0.045, | |
| "step": 500 | |
| }, | |
| { | |
| "completion_length": 388.8553741455078, | |
| "epoch": 0.3902627511591963, | |
| "grad_norm": 0.7688225507736206, | |
| "kl": 0.4513671875, | |
| "learning_rate": 3.825179614174195e-06, | |
| "loss": 0.0181, | |
| "reward": 1.8348215222358704, | |
| "reward_std": 0.6801443308591842, | |
| "rewards/accuracy_reward": 0.9339286029338837, | |
| "rewards/format_reward": 0.9008928954601287, | |
| "step": 505 | |
| }, | |
| { | |
| "completion_length": 369.0491241455078, | |
| "epoch": 0.3941267387944359, | |
| "grad_norm": 0.883755624294281, | |
| "kl": 0.4251953125, | |
| "learning_rate": 3.796452409413887e-06, | |
| "loss": 0.017, | |
| "reward": 1.913392972946167, | |
| "reward_std": 0.598648875951767, | |
| "rewards/accuracy_reward": 0.9767857670783997, | |
| "rewards/format_reward": 0.9366071820259094, | |
| "step": 510 | |
| }, | |
| { | |
| "completion_length": 370.35894775390625, | |
| "epoch": 0.3979907264296754, | |
| "grad_norm": 1.2042393684387207, | |
| "kl": 0.3263671875, | |
| "learning_rate": 3.767489111288757e-06, | |
| "loss": 0.0131, | |
| "reward": 1.8642858028411866, | |
| "reward_std": 0.5265406250953675, | |
| "rewards/accuracy_reward": 0.9267857611179352, | |
| "rewards/format_reward": 0.9375000476837159, | |
| "step": 515 | |
| }, | |
| { | |
| "completion_length": 396.01519775390625, | |
| "epoch": 0.401854714064915, | |
| "grad_norm": 2.4166510105133057, | |
| "kl": 0.45439453125, | |
| "learning_rate": 3.7382949942249695e-06, | |
| "loss": 0.0182, | |
| "reward": 1.8446429371833801, | |
| "reward_std": 0.5753746211528779, | |
| "rewards/accuracy_reward": 0.9410714566707611, | |
| "rewards/format_reward": 0.90357146859169, | |
| "step": 520 | |
| }, | |
| { | |
| "completion_length": 387.351806640625, | |
| "epoch": 0.40571870170015456, | |
| "grad_norm": 1.1032941341400146, | |
| "kl": 0.33154296875, | |
| "learning_rate": 3.7088753746824896e-06, | |
| "loss": 0.0133, | |
| "reward": 1.9517857789993287, | |
| "reward_std": 0.62562136054039, | |
| "rewards/accuracy_reward": 1.0428571939468383, | |
| "rewards/format_reward": 0.9089286208152771, | |
| "step": 525 | |
| }, | |
| { | |
| "completion_length": 399.0643035888672, | |
| "epoch": 0.4095826893353941, | |
| "grad_norm": 0.6850565075874329, | |
| "kl": 0.26015625, | |
| "learning_rate": 3.6792356101869157e-06, | |
| "loss": 0.0104, | |
| "reward": 1.8660715222358704, | |
| "reward_std": 0.6411642551422119, | |
| "rewards/accuracy_reward": 0.9500000476837158, | |
| "rewards/format_reward": 0.9160714626312256, | |
| "step": 530 | |
| }, | |
| { | |
| "completion_length": 409.71162109375, | |
| "epoch": 0.4134466769706337, | |
| "grad_norm": 0.8124867677688599, | |
| "kl": 0.6361328125, | |
| "learning_rate": 3.649381098353834e-06, | |
| "loss": 0.0255, | |
| "reward": 1.9285714864730834, | |
| "reward_std": 0.6074248850345612, | |
| "rewards/accuracy_reward": 1.0392857670783997, | |
| "rewards/format_reward": 0.8892857611179352, | |
| "step": 535 | |
| }, | |
| { | |
| "completion_length": 388.5687622070312, | |
| "epoch": 0.41731066460587324, | |
| "grad_norm": 0.6513718366622925, | |
| "kl": 0.31435546875, | |
| "learning_rate": 3.619317275905874e-06, | |
| "loss": 0.0126, | |
| "reward": 1.9750000953674316, | |
| "reward_std": 0.6598183512687683, | |
| "rewards/accuracy_reward": 1.0875000476837158, | |
| "rewards/format_reward": 0.8875000417232514, | |
| "step": 540 | |
| }, | |
| { | |
| "completion_length": 375.90537109375, | |
| "epoch": 0.42117465224111283, | |
| "grad_norm": 0.9380148649215698, | |
| "kl": 0.3015625, | |
| "learning_rate": 3.589049617682646e-06, | |
| "loss": 0.0121, | |
| "reward": 1.9455357909202575, | |
| "reward_std": 0.6462159514427185, | |
| "rewards/accuracy_reward": 1.0446428954601288, | |
| "rewards/format_reward": 0.9008929073810578, | |
| "step": 545 | |
| }, | |
| { | |
| "completion_length": 415.32323303222654, | |
| "epoch": 0.4250386398763524, | |
| "grad_norm": 3.0105764865875244, | |
| "kl": 0.60146484375, | |
| "learning_rate": 3.5585836356437266e-06, | |
| "loss": 0.0241, | |
| "reward": 1.81071435213089, | |
| "reward_std": 0.6503873735666275, | |
| "rewards/accuracy_reward": 0.9339286088943481, | |
| "rewards/format_reward": 0.8767857670783996, | |
| "step": 550 | |
| }, | |
| { | |
| "completion_length": 440.2902008056641, | |
| "epoch": 0.42890262751159197, | |
| "grad_norm": 3.932904005050659, | |
| "kl": 0.581640625, | |
| "learning_rate": 3.5279248778648944e-06, | |
| "loss": 0.0233, | |
| "reward": 1.725892925262451, | |
| "reward_std": 0.7691966652870178, | |
| "rewards/accuracy_reward": 0.9178571760654449, | |
| "rewards/format_reward": 0.8080357551574707, | |
| "step": 555 | |
| }, | |
| { | |
| "completion_length": 392.79644165039065, | |
| "epoch": 0.4327666151468315, | |
| "grad_norm": 1.320512294769287, | |
| "kl": 0.74921875, | |
| "learning_rate": 3.4970789275277878e-06, | |
| "loss": 0.03, | |
| "reward": 1.7607143640518188, | |
| "reward_std": 0.6493774831295014, | |
| "rewards/accuracy_reward": 0.9071428954601288, | |
| "rewards/format_reward": 0.853571480512619, | |
| "step": 560 | |
| }, | |
| { | |
| "completion_length": 335.8830505371094, | |
| "epoch": 0.4366306027820711, | |
| "grad_norm": 0.6699833273887634, | |
| "kl": 1.0361328125, | |
| "learning_rate": 3.466051401903162e-06, | |
| "loss": 0.0416, | |
| "reward": 1.929464375972748, | |
| "reward_std": 0.5458748698234558, | |
| "rewards/accuracy_reward": 0.9892857730388641, | |
| "rewards/format_reward": 0.9401786088943481, | |
| "step": 565 | |
| }, | |
| { | |
| "completion_length": 366.95447998046876, | |
| "epoch": 0.44049459041731065, | |
| "grad_norm": 1.531845211982727, | |
| "kl": 1.183984375, | |
| "learning_rate": 3.434847951327949e-06, | |
| "loss": 0.0474, | |
| "reward": 1.9107143759727478, | |
| "reward_std": 0.57608083486557, | |
| "rewards/accuracy_reward": 0.9750000536441803, | |
| "rewards/format_reward": 0.9357143223285675, | |
| "step": 570 | |
| }, | |
| { | |
| "completion_length": 406.04376525878905, | |
| "epoch": 0.44435857805255025, | |
| "grad_norm": 2.7006077766418457, | |
| "kl": 1.67265625, | |
| "learning_rate": 3.403474258176283e-06, | |
| "loss": 0.067, | |
| "reward": 1.8223215341567993, | |
| "reward_std": 0.6289676070213318, | |
| "rewards/accuracy_reward": 0.9214286029338836, | |
| "rewards/format_reward": 0.9008929073810578, | |
| "step": 575 | |
| }, | |
| { | |
| "completion_length": 434.67323608398436, | |
| "epoch": 0.4482225656877898, | |
| "grad_norm": 2.290830135345459, | |
| "kl": 2.16640625, | |
| "learning_rate": 3.3719360358247054e-06, | |
| "loss": 0.0866, | |
| "reward": 1.746428644657135, | |
| "reward_std": 0.7944486320018769, | |
| "rewards/accuracy_reward": 0.9267857491970062, | |
| "rewards/format_reward": 0.8196428894996644, | |
| "step": 580 | |
| }, | |
| { | |
| "completion_length": 459.25270385742186, | |
| "epoch": 0.4520865533230294, | |
| "grad_norm": 1.7977360486984253, | |
| "kl": 1.92578125, | |
| "learning_rate": 3.3402390276117175e-06, | |
| "loss": 0.077, | |
| "reward": 1.6910715222358703, | |
| "reward_std": 0.8230761885643005, | |
| "rewards/accuracy_reward": 0.8892857611179352, | |
| "rewards/format_reward": 0.8017857432365417, | |
| "step": 585 | |
| }, | |
| { | |
| "completion_length": 437.5250213623047, | |
| "epoch": 0.4559505409582689, | |
| "grad_norm": 1.8455265760421753, | |
| "kl": 1.071484375, | |
| "learning_rate": 3.308389005791872e-06, | |
| "loss": 0.0429, | |
| "reward": 1.7035714983940125, | |
| "reward_std": 0.7656769216060638, | |
| "rewards/accuracy_reward": 0.878571480512619, | |
| "rewards/format_reward": 0.8250000417232514, | |
| "step": 590 | |
| }, | |
| { | |
| "completion_length": 435.51698303222656, | |
| "epoch": 0.4598145285935085, | |
| "grad_norm": 4.333515167236328, | |
| "kl": 0.9978515625, | |
| "learning_rate": 3.276391770484606e-06, | |
| "loss": 0.0399, | |
| "reward": 1.6803572058677674, | |
| "reward_std": 0.7570467174053193, | |
| "rewards/accuracy_reward": 0.8267857551574707, | |
| "rewards/format_reward": 0.8535714685916901, | |
| "step": 595 | |
| }, | |
| { | |
| "completion_length": 410.84645080566406, | |
| "epoch": 0.46367851622874806, | |
| "grad_norm": 2.310718536376953, | |
| "kl": 0.76328125, | |
| "learning_rate": 3.244253148618002e-06, | |
| "loss": 0.0305, | |
| "reward": 1.771428644657135, | |
| "reward_std": 0.7016718983650208, | |
| "rewards/accuracy_reward": 0.9071428954601288, | |
| "rewards/format_reward": 0.8642857551574707, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.46367851622874806, | |
| "eval_completion_length": 385.1636678059896, | |
| "eval_kl": 0.6589192708333333, | |
| "eval_loss": 0.025798479095101357, | |
| "eval_reward": 1.7809524695078531, | |
| "eval_reward_std": 0.6921192049980164, | |
| "eval_rewards/accuracy_reward": 0.9285714546839396, | |
| "eval_rewards/format_reward": 0.8523809949556986, | |
| "eval_runtime": 92.176, | |
| "eval_samples_per_second": 1.074, | |
| "eval_steps_per_second": 0.043, | |
| "step": 600 | |
| }, | |
| { | |
| "completion_length": 408.04198303222654, | |
| "epoch": 0.46754250386398766, | |
| "grad_norm": 1.6843591928482056, | |
| "kl": 0.76171875, | |
| "learning_rate": 3.211978992867653e-06, | |
| "loss": 0.0305, | |
| "reward": 1.771428656578064, | |
| "reward_std": 0.6202342182397842, | |
| "rewards/accuracy_reward": 0.8875000417232514, | |
| "rewards/format_reward": 0.8839286148548127, | |
| "step": 605 | |
| }, | |
| { | |
| "completion_length": 395.34287414550784, | |
| "epoch": 0.4714064914992272, | |
| "grad_norm": 1.7479908466339111, | |
| "kl": 0.844140625, | |
| "learning_rate": 3.1795751805908578e-06, | |
| "loss": 0.0338, | |
| "reward": 1.8946429252624513, | |
| "reward_std": 0.6609594583511352, | |
| "rewards/accuracy_reward": 1.0071428954601287, | |
| "rewards/format_reward": 0.8875000476837158, | |
| "step": 610 | |
| }, | |
| { | |
| "completion_length": 426.66698303222654, | |
| "epoch": 0.4752704791344668, | |
| "grad_norm": 4.6757588386535645, | |
| "kl": 0.909375, | |
| "learning_rate": 3.147047612756302e-06, | |
| "loss": 0.0364, | |
| "reward": 1.8125000596046448, | |
| "reward_std": 0.7211765825748444, | |
| "rewards/accuracy_reward": 0.9125000417232514, | |
| "rewards/format_reward": 0.9000000476837158, | |
| "step": 615 | |
| }, | |
| { | |
| "completion_length": 409.9562713623047, | |
| "epoch": 0.47913446676970634, | |
| "grad_norm": 0.9073500633239746, | |
| "kl": 0.8787109375, | |
| "learning_rate": 3.1144022128694583e-06, | |
| "loss": 0.0352, | |
| "reward": 1.821428644657135, | |
| "reward_std": 0.6061809420585632, | |
| "rewards/accuracy_reward": 0.9285714626312256, | |
| "rewards/format_reward": 0.8928571879863739, | |
| "step": 620 | |
| }, | |
| { | |
| "completion_length": 369.6464477539063, | |
| "epoch": 0.48299845440494593, | |
| "grad_norm": 0.41812455654144287, | |
| "kl": 0.39560546875, | |
| "learning_rate": 3.081644925893866e-06, | |
| "loss": 0.0158, | |
| "reward": 1.9875000715255737, | |
| "reward_std": 0.535971587896347, | |
| "rewards/accuracy_reward": 1.051785761117935, | |
| "rewards/format_reward": 0.935714328289032, | |
| "step": 625 | |
| }, | |
| { | |
| "completion_length": 377.90537109375, | |
| "epoch": 0.4868624420401855, | |
| "grad_norm": 0.5811069011688232, | |
| "kl": 0.44228515625, | |
| "learning_rate": 3.048781717168513e-06, | |
| "loss": 0.0177, | |
| "reward": 1.9285714983940125, | |
| "reward_std": 0.5182920306921005, | |
| "rewards/accuracy_reward": 0.9875000536441803, | |
| "rewards/format_reward": 0.9410714745521546, | |
| "step": 630 | |
| }, | |
| { | |
| "completion_length": 372.5982299804688, | |
| "epoch": 0.490726429675425, | |
| "grad_norm": 0.9005226492881775, | |
| "kl": 0.37333984375, | |
| "learning_rate": 3.015818571321504e-06, | |
| "loss": 0.0149, | |
| "reward": 2.0464287042617797, | |
| "reward_std": 0.5508933126926422, | |
| "rewards/accuracy_reward": 1.0946429014205932, | |
| "rewards/format_reward": 0.9517857551574707, | |
| "step": 635 | |
| }, | |
| { | |
| "completion_length": 361.0571594238281, | |
| "epoch": 0.4945904173106646, | |
| "grad_norm": 1.6126573085784912, | |
| "kl": 0.625390625, | |
| "learning_rate": 2.9827614911802205e-06, | |
| "loss": 0.025, | |
| "reward": 1.9017857789993287, | |
| "reward_std": 0.594823956489563, | |
| "rewards/accuracy_reward": 0.9803571939468384, | |
| "rewards/format_reward": 0.9214286148548126, | |
| "step": 640 | |
| }, | |
| { | |
| "completion_length": 379.9589447021484, | |
| "epoch": 0.49845440494590415, | |
| "grad_norm": 1.6361123323440552, | |
| "kl": 0.5390625, | |
| "learning_rate": 2.949616496678153e-06, | |
| "loss": 0.0215, | |
| "reward": 1.8375000596046447, | |
| "reward_std": 0.5699772477149964, | |
| "rewards/accuracy_reward": 0.9125000298023224, | |
| "rewards/format_reward": 0.9250000357627869, | |
| "step": 645 | |
| }, | |
| { | |
| "completion_length": 368.4125183105469, | |
| "epoch": 0.5023183925811437, | |
| "grad_norm": 2.0848145484924316, | |
| "kl": 0.698046875, | |
| "learning_rate": 2.9163896237586365e-06, | |
| "loss": 0.0279, | |
| "reward": 1.8410715103149413, | |
| "reward_std": 0.5587106943130493, | |
| "rewards/accuracy_reward": 0.9125000417232514, | |
| "rewards/format_reward": 0.92857146859169, | |
| "step": 650 | |
| }, | |
| { | |
| "completion_length": 376.9428741455078, | |
| "epoch": 0.5061823802163833, | |
| "grad_norm": 1.3533092737197876, | |
| "kl": 0.916015625, | |
| "learning_rate": 2.883086923275658e-06, | |
| "loss": 0.0366, | |
| "reward": 1.8705358147621154, | |
| "reward_std": 0.7300362765789032, | |
| "rewards/accuracy_reward": 0.9839286029338836, | |
| "rewards/format_reward": 0.886607187986374, | |
| "step": 655 | |
| }, | |
| { | |
| "completion_length": 385.7473419189453, | |
| "epoch": 0.5100463678516228, | |
| "grad_norm": 2.705904960632324, | |
| "kl": 0.831640625, | |
| "learning_rate": 2.849714459891953e-06, | |
| "loss": 0.0333, | |
| "reward": 1.8303572177886962, | |
| "reward_std": 0.7160472482442856, | |
| "rewards/accuracy_reward": 0.9553571879863739, | |
| "rewards/format_reward": 0.8750000476837159, | |
| "step": 660 | |
| }, | |
| { | |
| "completion_length": 380.43662109375, | |
| "epoch": 0.5139103554868625, | |
| "grad_norm": 0.9756877422332764, | |
| "kl": 0.5552734375, | |
| "learning_rate": 2.8162783109745833e-06, | |
| "loss": 0.0222, | |
| "reward": 1.9214286804199219, | |
| "reward_std": 0.6294688701629638, | |
| "rewards/accuracy_reward": 1.0196429073810578, | |
| "rewards/format_reward": 0.9017857551574707, | |
| "step": 665 | |
| }, | |
| { | |
| "completion_length": 384.23841247558596, | |
| "epoch": 0.517774343122102, | |
| "grad_norm": 0.8622917532920837, | |
| "kl": 0.34326171875, | |
| "learning_rate": 2.7827845654882112e-06, | |
| "loss": 0.0137, | |
| "reward": 1.896428680419922, | |
| "reward_std": 0.5160377115011215, | |
| "rewards/accuracy_reward": 0.9660714745521546, | |
| "rewards/format_reward": 0.9303571879863739, | |
| "step": 670 | |
| }, | |
| { | |
| "completion_length": 362.0500152587891, | |
| "epoch": 0.5216383307573416, | |
| "grad_norm": 0.9457660913467407, | |
| "kl": 0.4822265625, | |
| "learning_rate": 2.749239322886248e-06, | |
| "loss": 0.0193, | |
| "reward": 2.019642949104309, | |
| "reward_std": 0.5436660468578338, | |
| "rewards/accuracy_reward": 1.0803572058677673, | |
| "rewards/format_reward": 0.9392857611179352, | |
| "step": 675 | |
| }, | |
| { | |
| "completion_length": 367.0884094238281, | |
| "epoch": 0.5255023183925811, | |
| "grad_norm": 0.9137616157531738, | |
| "kl": 0.378515625, | |
| "learning_rate": 2.7156486920001024e-06, | |
| "loss": 0.0151, | |
| "reward": 1.9035715103149413, | |
| "reward_std": 0.5723553836345673, | |
| "rewards/accuracy_reward": 0.9732143402099609, | |
| "rewards/format_reward": 0.9303571820259094, | |
| "step": 680 | |
| }, | |
| { | |
| "completion_length": 360.2259063720703, | |
| "epoch": 0.5293663060278208, | |
| "grad_norm": 1.603175401687622, | |
| "kl": 0.57578125, | |
| "learning_rate": 2.6820187899267203e-06, | |
| "loss": 0.023, | |
| "reward": 1.9187500834465028, | |
| "reward_std": 0.6464061141014099, | |
| "rewards/accuracy_reward": 1.0089286267757416, | |
| "rewards/format_reward": 0.9098214745521546, | |
| "step": 685 | |
| }, | |
| { | |
| "completion_length": 362.6777008056641, | |
| "epoch": 0.5332302936630603, | |
| "grad_norm": 1.3457704782485962, | |
| "kl": 0.8306640625, | |
| "learning_rate": 2.6483557409146133e-06, | |
| "loss": 0.0332, | |
| "reward": 1.8000000834465026, | |
| "reward_std": 0.6633238136768341, | |
| "rewards/accuracy_reward": 0.9178571820259094, | |
| "rewards/format_reward": 0.8821429014205933, | |
| "step": 690 | |
| }, | |
| { | |
| "completion_length": 379.5053741455078, | |
| "epoch": 0.5370942812982998, | |
| "grad_norm": 2.6664907932281494, | |
| "kl": 0.89375, | |
| "learning_rate": 2.6146656752485904e-06, | |
| "loss": 0.0358, | |
| "reward": 1.7928572058677674, | |
| "reward_std": 0.7002339541912079, | |
| "rewards/accuracy_reward": 0.910714328289032, | |
| "rewards/format_reward": 0.8821428954601288, | |
| "step": 695 | |
| }, | |
| { | |
| "completion_length": 382.989306640625, | |
| "epoch": 0.5409582689335394, | |
| "grad_norm": 2.186499834060669, | |
| "kl": 0.8466796875, | |
| "learning_rate": 2.5809547281333904e-06, | |
| "loss": 0.0339, | |
| "reward": 1.7866072177886962, | |
| "reward_std": 0.6277125418186188, | |
| "rewards/accuracy_reward": 0.8892857551574707, | |
| "rewards/format_reward": 0.8973214805126191, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.5409582689335394, | |
| "eval_completion_length": 327.28858642578126, | |
| "eval_kl": 0.5944010416666666, | |
| "eval_loss": 0.024204090237617493, | |
| "eval_reward": 1.959523892402649, | |
| "eval_reward_std": 0.5741641213496526, | |
| "eval_rewards/accuracy_reward": 1.0190476616223654, | |
| "eval_rewards/format_reward": 0.9404762188593546, | |
| "eval_runtime": 83.2638, | |
| "eval_samples_per_second": 1.189, | |
| "eval_steps_per_second": 0.048, | |
| "step": 700 | |
| }, | |
| { | |
| "completion_length": 327.90180053710935, | |
| "epoch": 0.544822256568779, | |
| "grad_norm": 2.957988739013672, | |
| "kl": 0.41298828125, | |
| "learning_rate": 2.5472290385764115e-06, | |
| "loss": 0.0166, | |
| "reward": 1.9491072416305542, | |
| "reward_std": 0.5571832716464996, | |
| "rewards/accuracy_reward": 0.9982143402099609, | |
| "rewards/format_reward": 0.9508928894996643, | |
| "step": 705 | |
| }, | |
| { | |
| "completion_length": 335.6509094238281, | |
| "epoch": 0.5486862442040186, | |
| "grad_norm": 2.2409234046936035, | |
| "kl": 0.6271484375, | |
| "learning_rate": 2.5134947482697615e-06, | |
| "loss": 0.0251, | |
| "reward": 1.8776786565780639, | |
| "reward_std": 0.5538718163967132, | |
| "rewards/accuracy_reward": 0.9535714745521545, | |
| "rewards/format_reward": 0.9241071820259095, | |
| "step": 710 | |
| }, | |
| { | |
| "completion_length": 316.4598358154297, | |
| "epoch": 0.5525502318392581, | |
| "grad_norm": 4.647254943847656, | |
| "kl": 1.2830078125, | |
| "learning_rate": 2.4797580004718038e-06, | |
| "loss": 0.0513, | |
| "reward": 1.9089286565780639, | |
| "reward_std": 0.6047213137149811, | |
| "rewards/accuracy_reward": 0.9821428954601288, | |
| "rewards/format_reward": 0.9267857551574707, | |
| "step": 715 | |
| }, | |
| { | |
| "completion_length": 331.44822692871094, | |
| "epoch": 0.5564142194744977, | |
| "grad_norm": 1.3831017017364502, | |
| "kl": 0.9451171875, | |
| "learning_rate": 2.446024938888431e-06, | |
| "loss": 0.0378, | |
| "reward": 1.891964364051819, | |
| "reward_std": 0.60347221493721, | |
| "rewards/accuracy_reward": 0.9500000476837158, | |
| "rewards/format_reward": 0.941964328289032, | |
| "step": 720 | |
| }, | |
| { | |
| "completion_length": 348.0705505371094, | |
| "epoch": 0.5602782071097373, | |
| "grad_norm": 2.458526849746704, | |
| "kl": 0.82890625, | |
| "learning_rate": 2.412301706554247e-06, | |
| "loss": 0.0332, | |
| "reward": 1.8625000715255737, | |
| "reward_std": 0.5647337824106217, | |
| "rewards/accuracy_reward": 0.9232143223285675, | |
| "rewards/format_reward": 0.9392857611179352, | |
| "step": 725 | |
| }, | |
| { | |
| "completion_length": 351.4446594238281, | |
| "epoch": 0.5641421947449768, | |
| "grad_norm": 0.9637498259544373, | |
| "kl": 1.144140625, | |
| "learning_rate": 2.3785944447138804e-06, | |
| "loss": 0.0458, | |
| "reward": 1.82589293718338, | |
| "reward_std": 0.6317932158708572, | |
| "rewards/accuracy_reward": 0.92857146859169, | |
| "rewards/format_reward": 0.8973214805126191, | |
| "step": 730 | |
| }, | |
| { | |
| "completion_length": 368.37501525878906, | |
| "epoch": 0.5680061823802164, | |
| "grad_norm": 0.7366132736206055, | |
| "kl": 0.6658203125, | |
| "learning_rate": 2.344909291703615e-06, | |
| "loss": 0.0266, | |
| "reward": 1.9098215222358703, | |
| "reward_std": 0.6351912558078766, | |
| "rewards/accuracy_reward": 1.0089286148548127, | |
| "rewards/format_reward": 0.9008929014205933, | |
| "step": 735 | |
| }, | |
| { | |
| "completion_length": 381.0259094238281, | |
| "epoch": 0.5718701700154559, | |
| "grad_norm": 1.8531627655029297, | |
| "kl": 0.59921875, | |
| "learning_rate": 2.3112523818335606e-06, | |
| "loss": 0.024, | |
| "reward": 1.8910715222358703, | |
| "reward_std": 0.5899456530809403, | |
| "rewards/accuracy_reward": 0.9803571939468384, | |
| "rewards/format_reward": 0.910714328289032, | |
| "step": 740 | |
| }, | |
| { | |
| "completion_length": 362.7080535888672, | |
| "epoch": 0.5757341576506955, | |
| "grad_norm": 1.4462131261825562, | |
| "kl": 0.6189453125, | |
| "learning_rate": 2.2776298442705434e-06, | |
| "loss": 0.0247, | |
| "reward": 1.933928668498993, | |
| "reward_std": 0.5655193716287613, | |
| "rewards/accuracy_reward": 0.9946428894996643, | |
| "rewards/format_reward": 0.9392857491970062, | |
| "step": 745 | |
| }, | |
| { | |
| "completion_length": 339.10358276367185, | |
| "epoch": 0.5795981452859351, | |
| "grad_norm": 0.6540559530258179, | |
| "kl": 0.383984375, | |
| "learning_rate": 2.244047801921944e-06, | |
| "loss": 0.0154, | |
| "reward": 2.0205358028411866, | |
| "reward_std": 0.542399314045906, | |
| "rewards/accuracy_reward": 1.0642857611179353, | |
| "rewards/format_reward": 0.9562500357627869, | |
| "step": 750 | |
| }, | |
| { | |
| "completion_length": 370.5160888671875, | |
| "epoch": 0.5834621329211747, | |
| "grad_norm": 0.6533460021018982, | |
| "kl": 0.4796875, | |
| "learning_rate": 2.2105123703206727e-06, | |
| "loss": 0.0192, | |
| "reward": 1.9312501072883606, | |
| "reward_std": 0.6289328157901763, | |
| "rewards/accuracy_reward": 0.998214328289032, | |
| "rewards/format_reward": 0.9330357611179352, | |
| "step": 755 | |
| }, | |
| { | |
| "completion_length": 372.5491180419922, | |
| "epoch": 0.5873261205564142, | |
| "grad_norm": 3.128967761993408, | |
| "kl": 0.53544921875, | |
| "learning_rate": 2.1770296565114847e-06, | |
| "loss": 0.0214, | |
| "reward": 1.9785715460777282, | |
| "reward_std": 0.5253061711788177, | |
| "rewards/accuracy_reward": 1.0357143342494965, | |
| "rewards/format_reward": 0.942857176065445, | |
| "step": 760 | |
| }, | |
| { | |
| "completion_length": 366.4009063720703, | |
| "epoch": 0.5911901081916537, | |
| "grad_norm": 0.8277491927146912, | |
| "kl": 0.46806640625, | |
| "learning_rate": 2.1436057579388443e-06, | |
| "loss": 0.0187, | |
| "reward": 1.865178656578064, | |
| "reward_std": 0.5838503152132034, | |
| "rewards/accuracy_reward": 0.9196429252624512, | |
| "rewards/format_reward": 0.9455357491970062, | |
| "step": 765 | |
| }, | |
| { | |
| "completion_length": 362.5089447021484, | |
| "epoch": 0.5950540958268934, | |
| "grad_norm": 0.9837579131126404, | |
| "kl": 0.43828125, | |
| "learning_rate": 2.1102467613365336e-06, | |
| "loss": 0.0175, | |
| "reward": 2.007142972946167, | |
| "reward_std": 0.5876386165618896, | |
| "rewards/accuracy_reward": 1.0625000476837159, | |
| "rewards/format_reward": 0.9446429014205933, | |
| "step": 770 | |
| }, | |
| { | |
| "completion_length": 373.8339447021484, | |
| "epoch": 0.5989180834621329, | |
| "grad_norm": 1.1622740030288696, | |
| "kl": 0.44111328125, | |
| "learning_rate": 2.0769587416192212e-06, | |
| "loss": 0.0177, | |
| "reward": 1.949107253551483, | |
| "reward_std": 0.5259672313928604, | |
| "rewards/accuracy_reward": 0.99107146859169, | |
| "rewards/format_reward": 0.9580357670783997, | |
| "step": 775 | |
| }, | |
| { | |
| "completion_length": 388.4268035888672, | |
| "epoch": 0.6027820710973725, | |
| "grad_norm": 2.432398557662964, | |
| "kl": 0.5046875, | |
| "learning_rate": 2.0437477607761656e-06, | |
| "loss": 0.0202, | |
| "reward": 1.8973215103149415, | |
| "reward_std": 0.6027686059474945, | |
| "rewards/accuracy_reward": 0.960714328289032, | |
| "rewards/format_reward": 0.936607176065445, | |
| "step": 780 | |
| }, | |
| { | |
| "completion_length": 376.93572692871095, | |
| "epoch": 0.606646058732612, | |
| "grad_norm": 0.8899438977241516, | |
| "kl": 0.475390625, | |
| "learning_rate": 2.0106198667672926e-06, | |
| "loss": 0.019, | |
| "reward": 1.929464340209961, | |
| "reward_std": 0.5717435568571091, | |
| "rewards/accuracy_reward": 0.9946429014205933, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 785 | |
| }, | |
| { | |
| "completion_length": 375.6580505371094, | |
| "epoch": 0.6105100463678517, | |
| "grad_norm": 1.3664709329605103, | |
| "kl": 0.3677734375, | |
| "learning_rate": 1.9775810924218126e-06, | |
| "loss": 0.0147, | |
| "reward": 1.969642949104309, | |
| "reward_std": 0.577250525355339, | |
| "rewards/accuracy_reward": 1.0267857670783997, | |
| "rewards/format_reward": 0.9428571879863739, | |
| "step": 790 | |
| }, | |
| { | |
| "completion_length": 372.2384094238281, | |
| "epoch": 0.6143740340030912, | |
| "grad_norm": 0.5266821384429932, | |
| "kl": 0.34619140625, | |
| "learning_rate": 1.944637454339601e-06, | |
| "loss": 0.0139, | |
| "reward": 1.9500000715255736, | |
| "reward_std": 0.547850227355957, | |
| "rewards/accuracy_reward": 0.9982143342494965, | |
| "rewards/format_reward": 0.9517857670783997, | |
| "step": 795 | |
| }, | |
| { | |
| "completion_length": 363.0500122070313, | |
| "epoch": 0.6182380216383307, | |
| "grad_norm": 0.4656059145927429, | |
| "kl": 0.34765625, | |
| "learning_rate": 1.9117949517955313e-06, | |
| "loss": 0.0139, | |
| "reward": 1.9125000834465027, | |
| "reward_std": 0.5355327159166337, | |
| "rewards/accuracy_reward": 0.9553571999073028, | |
| "rewards/format_reward": 0.9571429073810578, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.6182380216383307, | |
| "eval_completion_length": 339.02525024414064, | |
| "eval_kl": 0.5237955729166667, | |
| "eval_loss": 0.020239148288965225, | |
| "eval_reward": 1.8928572177886962, | |
| "eval_reward_std": 0.520437486966451, | |
| "eval_rewards/accuracy_reward": 0.947619092464447, | |
| "eval_rewards/format_reward": 0.9452381173769633, | |
| "eval_runtime": 82.526, | |
| "eval_samples_per_second": 1.2, | |
| "eval_steps_per_second": 0.048, | |
| "step": 800 | |
| }, | |
| { | |
| "completion_length": 355.1919799804688, | |
| "epoch": 0.6221020092735703, | |
| "grad_norm": 1.5192807912826538, | |
| "kl": 0.4408203125, | |
| "learning_rate": 1.8790595656469628e-06, | |
| "loss": 0.0176, | |
| "reward": 1.9366072177886964, | |
| "reward_std": 0.5646369904279709, | |
| "rewards/accuracy_reward": 0.9892857491970062, | |
| "rewards/format_reward": 0.9473214685916901, | |
| "step": 805 | |
| }, | |
| { | |
| "completion_length": 354.4946624755859, | |
| "epoch": 0.6259659969088099, | |
| "grad_norm": 1.2222896814346313, | |
| "kl": 0.398046875, | |
| "learning_rate": 1.8464372572445867e-06, | |
| "loss": 0.0159, | |
| "reward": 1.9169643640518188, | |
| "reward_std": 0.5843247085809707, | |
| "rewards/accuracy_reward": 0.9785714685916901, | |
| "rewards/format_reward": 0.9383928954601288, | |
| "step": 810 | |
| }, | |
| { | |
| "completion_length": 342.8660827636719, | |
| "epoch": 0.6298299845440495, | |
| "grad_norm": 1.4474635124206543, | |
| "kl": 0.32158203125, | |
| "learning_rate": 1.8139339673468142e-06, | |
| "loss": 0.0129, | |
| "reward": 1.9285715222358704, | |
| "reward_std": 0.5523967891931534, | |
| "rewards/accuracy_reward": 0.9678571939468383, | |
| "rewards/format_reward": 0.9607143342494965, | |
| "step": 815 | |
| }, | |
| { | |
| "completion_length": 368.4000183105469, | |
| "epoch": 0.633693972179289, | |
| "grad_norm": 2.314242362976074, | |
| "kl": 0.63984375, | |
| "learning_rate": 1.7815556150379298e-06, | |
| "loss": 0.0256, | |
| "reward": 1.8758929371833801, | |
| "reward_std": 0.5279008090496063, | |
| "rewards/accuracy_reward": 0.9392857670783996, | |
| "rewards/format_reward": 0.9366071879863739, | |
| "step": 820 | |
| }, | |
| { | |
| "completion_length": 379.89019470214845, | |
| "epoch": 0.6375579598145286, | |
| "grad_norm": 0.9705050587654114, | |
| "kl": 0.5564453125, | |
| "learning_rate": 1.7493080966501764e-06, | |
| "loss": 0.0222, | |
| "reward": 1.858928644657135, | |
| "reward_std": 0.5612190932035446, | |
| "rewards/accuracy_reward": 0.92857146859169, | |
| "rewards/format_reward": 0.930357176065445, | |
| "step": 825 | |
| }, | |
| { | |
| "completion_length": 363.3250213623047, | |
| "epoch": 0.6414219474497682, | |
| "grad_norm": 0.9212985038757324, | |
| "kl": 0.2513671875, | |
| "learning_rate": 1.7171972846899942e-06, | |
| "loss": 0.01, | |
| "reward": 1.8767857909202577, | |
| "reward_std": 0.5048141717910767, | |
| "rewards/accuracy_reward": 0.9267857611179352, | |
| "rewards/format_reward": 0.9500000417232514, | |
| "step": 830 | |
| }, | |
| { | |
| "completion_length": 369.3312683105469, | |
| "epoch": 0.6452859350850078, | |
| "grad_norm": 0.9865968227386475, | |
| "kl": 0.358203125, | |
| "learning_rate": 1.685229026768593e-06, | |
| "loss": 0.0143, | |
| "reward": 1.9616072297096252, | |
| "reward_std": 0.5067572951316833, | |
| "rewards/accuracy_reward": 1.0107143223285675, | |
| "rewards/format_reward": 0.9508929073810577, | |
| "step": 835 | |
| }, | |
| { | |
| "completion_length": 373.60805053710936, | |
| "epoch": 0.6491499227202473, | |
| "grad_norm": 0.8022987842559814, | |
| "kl": 0.5169921875, | |
| "learning_rate": 1.6534091445370604e-06, | |
| "loss": 0.0207, | |
| "reward": 1.9375001072883606, | |
| "reward_std": 0.6279206275939941, | |
| "rewards/accuracy_reward": 1.0125000476837158, | |
| "rewards/format_reward": 0.9250000476837158, | |
| "step": 840 | |
| }, | |
| { | |
| "completion_length": 372.23037109375, | |
| "epoch": 0.6530139103554868, | |
| "grad_norm": 1.627290964126587, | |
| "kl": 0.51416015625, | |
| "learning_rate": 1.6217434326261999e-06, | |
| "loss": 0.0206, | |
| "reward": 1.8633929371833802, | |
| "reward_std": 0.5344478011131286, | |
| "rewards/accuracy_reward": 0.9267857611179352, | |
| "rewards/format_reward": 0.936607176065445, | |
| "step": 845 | |
| }, | |
| { | |
| "completion_length": 368.5562683105469, | |
| "epoch": 0.6568778979907264, | |
| "grad_norm": 0.7936336994171143, | |
| "kl": 0.2919921875, | |
| "learning_rate": 1.5902376575912815e-06, | |
| "loss": 0.0117, | |
| "reward": 1.9357143878936767, | |
| "reward_std": 0.5026874512434005, | |
| "rewards/accuracy_reward": 0.9821428954601288, | |
| "rewards/format_reward": 0.9535714745521545, | |
| "step": 850 | |
| }, | |
| { | |
| "completion_length": 366.2107360839844, | |
| "epoch": 0.660741885625966, | |
| "grad_norm": 0.8993034958839417, | |
| "kl": 0.40234375, | |
| "learning_rate": 1.5588975568619124e-06, | |
| "loss": 0.0161, | |
| "reward": 1.9267858266830444, | |
| "reward_std": 0.5611187249422074, | |
| "rewards/accuracy_reward": 0.983928632736206, | |
| "rewards/format_reward": 0.9428571820259094, | |
| "step": 855 | |
| }, | |
| { | |
| "completion_length": 358.8482299804688, | |
| "epoch": 0.6646058732612056, | |
| "grad_norm": 1.1370081901550293, | |
| "kl": 0.6099609375, | |
| "learning_rate": 1.5277288376972116e-06, | |
| "loss": 0.0244, | |
| "reward": 1.8241072416305542, | |
| "reward_std": 0.5759775102138519, | |
| "rewards/accuracy_reward": 0.8892857551574707, | |
| "rewards/format_reward": 0.9348214745521546, | |
| "step": 860 | |
| }, | |
| { | |
| "completion_length": 369.9125183105469, | |
| "epoch": 0.6684698608964451, | |
| "grad_norm": 0.681843101978302, | |
| "kl": 0.424609375, | |
| "learning_rate": 1.4967371761464738e-06, | |
| "loss": 0.017, | |
| "reward": 1.9053572297096253, | |
| "reward_std": 0.541083812713623, | |
| "rewards/accuracy_reward": 0.9750000417232514, | |
| "rewards/format_reward": 0.9303571879863739, | |
| "step": 865 | |
| }, | |
| { | |
| "completion_length": 362.09466247558595, | |
| "epoch": 0.6723338485316847, | |
| "grad_norm": 1.1364926099777222, | |
| "kl": 0.491015625, | |
| "learning_rate": 1.4659282160155222e-06, | |
| "loss": 0.0196, | |
| "reward": 1.8607143878936767, | |
| "reward_std": 0.5730058521032333, | |
| "rewards/accuracy_reward": 0.9267857611179352, | |
| "rewards/format_reward": 0.9339286148548126, | |
| "step": 870 | |
| }, | |
| { | |
| "completion_length": 360.81340942382815, | |
| "epoch": 0.6761978361669243, | |
| "grad_norm": 1.0204797983169556, | |
| "kl": 0.5099609375, | |
| "learning_rate": 1.4353075678389284e-06, | |
| "loss": 0.0204, | |
| "reward": 1.9785715460777282, | |
| "reward_std": 0.5866712421178818, | |
| "rewards/accuracy_reward": 1.0392857611179351, | |
| "rewards/format_reward": 0.9392857491970062, | |
| "step": 875 | |
| }, | |
| { | |
| "completion_length": 364.41787109375, | |
| "epoch": 0.6800618238021638, | |
| "grad_norm": 1.010473608970642, | |
| "kl": 0.63125, | |
| "learning_rate": 1.4048808078582943e-06, | |
| "loss": 0.0253, | |
| "reward": 1.908928668498993, | |
| "reward_std": 0.611833056807518, | |
| "rewards/accuracy_reward": 0.9928571820259094, | |
| "rewards/format_reward": 0.9160714685916901, | |
| "step": 880 | |
| }, | |
| { | |
| "completion_length": 354.1143005371094, | |
| "epoch": 0.6839258114374034, | |
| "grad_norm": 1.0668288469314575, | |
| "kl": 0.5849609375, | |
| "learning_rate": 1.3746534770067803e-06, | |
| "loss": 0.0234, | |
| "reward": 2.0035715460777284, | |
| "reward_std": 0.5578104436397553, | |
| "rewards/accuracy_reward": 1.066071480512619, | |
| "rewards/format_reward": 0.9375000417232513, | |
| "step": 885 | |
| }, | |
| { | |
| "completion_length": 364.14198303222656, | |
| "epoch": 0.6877897990726429, | |
| "grad_norm": 1.7410629987716675, | |
| "kl": 0.37431640625, | |
| "learning_rate": 1.3446310799000578e-06, | |
| "loss": 0.015, | |
| "reward": 1.9598215341567993, | |
| "reward_std": 0.5321772754192352, | |
| "rewards/accuracy_reward": 1.0303571939468383, | |
| "rewards/format_reward": 0.929464328289032, | |
| "step": 890 | |
| }, | |
| { | |
| "completion_length": 345.7285888671875, | |
| "epoch": 0.6916537867078826, | |
| "grad_norm": 0.8086636662483215, | |
| "kl": 0.453515625, | |
| "learning_rate": 1.3148190838338804e-06, | |
| "loss": 0.0181, | |
| "reward": 1.9848214864730835, | |
| "reward_std": 0.5839688003063201, | |
| "rewards/accuracy_reward": 1.0517857670783997, | |
| "rewards/format_reward": 0.9330357491970063, | |
| "step": 895 | |
| }, | |
| { | |
| "completion_length": 391.9473358154297, | |
| "epoch": 0.6955177743431221, | |
| "grad_norm": 1.5891263484954834, | |
| "kl": 0.8416015625, | |
| "learning_rate": 1.2852229177884492e-06, | |
| "loss": 0.0337, | |
| "reward": 1.813392949104309, | |
| "reward_std": 0.6605880260467529, | |
| "rewards/accuracy_reward": 0.9160714685916901, | |
| "rewards/format_reward": 0.8973214745521545, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.6955177743431221, | |
| "eval_completion_length": 342.9001749674479, | |
| "eval_kl": 0.8555989583333333, | |
| "eval_loss": 0.03408632054924965, | |
| "eval_reward": 2.0619048436482745, | |
| "eval_reward_std": 0.5802303751309713, | |
| "eval_rewards/accuracy_reward": 1.1428571979204813, | |
| "eval_rewards/format_reward": 0.9190476576487223, | |
| "eval_runtime": 90.4473, | |
| "eval_samples_per_second": 1.095, | |
| "eval_steps_per_second": 0.044, | |
| "step": 900 | |
| }, | |
| { | |
| "completion_length": 354.5241180419922, | |
| "epoch": 0.6993817619783617, | |
| "grad_norm": 2.794023036956787, | |
| "kl": 0.760546875, | |
| "learning_rate": 1.2558479714397585e-06, | |
| "loss": 0.0304, | |
| "reward": 1.9258929371833802, | |
| "reward_std": 0.6209361255168915, | |
| "rewards/accuracy_reward": 0.9946429073810578, | |
| "rewards/format_reward": 0.9312500417232513, | |
| "step": 905 | |
| }, | |
| { | |
| "completion_length": 375.51787719726565, | |
| "epoch": 0.7032457496136012, | |
| "grad_norm": 1.4545397758483887, | |
| "kl": 0.623828125, | |
| "learning_rate": 1.2266995941780934e-06, | |
| "loss": 0.025, | |
| "reward": 1.858928644657135, | |
| "reward_std": 0.6513124674558639, | |
| "rewards/accuracy_reward": 0.9464286029338836, | |
| "rewards/format_reward": 0.9125000476837158, | |
| "step": 910 | |
| }, | |
| { | |
| "completion_length": 352.10894775390625, | |
| "epoch": 0.7071097372488409, | |
| "grad_norm": 1.136331558227539, | |
| "kl": 0.5474609375, | |
| "learning_rate": 1.197783094133869e-06, | |
| "loss": 0.0219, | |
| "reward": 1.902678644657135, | |
| "reward_std": 0.5192236006259918, | |
| "rewards/accuracy_reward": 0.973214328289032, | |
| "rewards/format_reward": 0.929464328289032, | |
| "step": 915 | |
| }, | |
| { | |
| "completion_length": 343.6509094238281, | |
| "epoch": 0.7109737248840804, | |
| "grad_norm": 1.513318419456482, | |
| "kl": 0.474609375, | |
| "learning_rate": 1.1691037372109835e-06, | |
| "loss": 0.019, | |
| "reward": 2.0866072535514832, | |
| "reward_std": 0.4978013187646866, | |
| "rewards/accuracy_reward": 1.1410714983940125, | |
| "rewards/format_reward": 0.9455357551574707, | |
| "step": 920 | |
| }, | |
| { | |
| "completion_length": 365.5428741455078, | |
| "epoch": 0.7148377125193199, | |
| "grad_norm": 2.5464484691619873, | |
| "kl": 0.6884765625, | |
| "learning_rate": 1.140666746127854e-06, | |
| "loss": 0.0276, | |
| "reward": 1.988392949104309, | |
| "reward_std": 0.5555747985839844, | |
| "rewards/accuracy_reward": 1.048214328289032, | |
| "rewards/format_reward": 0.9401786088943481, | |
| "step": 925 | |
| }, | |
| { | |
| "completion_length": 359.7901947021484, | |
| "epoch": 0.7187017001545595, | |
| "grad_norm": 0.9951956272125244, | |
| "kl": 0.61796875, | |
| "learning_rate": 1.1124772994663258e-06, | |
| "loss": 0.0247, | |
| "reward": 2.0348215341567992, | |
| "reward_std": 0.5215620249509811, | |
| "rewards/accuracy_reward": 1.0964286148548126, | |
| "rewards/format_reward": 0.9383929014205933, | |
| "step": 930 | |
| }, | |
| { | |
| "completion_length": 344.65537719726564, | |
| "epoch": 0.7225656877897991, | |
| "grad_norm": 0.8325075507164001, | |
| "kl": 0.4759765625, | |
| "learning_rate": 1.084540530728613e-06, | |
| "loss": 0.0191, | |
| "reward": 2.0116072416305544, | |
| "reward_std": 0.48731706738471986, | |
| "rewards/accuracy_reward": 1.0625000417232513, | |
| "rewards/format_reward": 0.949107187986374, | |
| "step": 935 | |
| }, | |
| { | |
| "completion_length": 374.4696563720703, | |
| "epoch": 0.7264296754250387, | |
| "grad_norm": 0.7985464930534363, | |
| "kl": 0.4962890625, | |
| "learning_rate": 1.0568615274024521e-06, | |
| "loss": 0.0199, | |
| "reward": 1.9285715460777282, | |
| "reward_std": 0.5753836840391159, | |
| "rewards/accuracy_reward": 0.99107146859169, | |
| "rewards/format_reward": 0.9375000476837159, | |
| "step": 940 | |
| }, | |
| { | |
| "completion_length": 360.10626831054685, | |
| "epoch": 0.7302936630602782, | |
| "grad_norm": 1.1231186389923096, | |
| "kl": 0.700390625, | |
| "learning_rate": 1.029445330034633e-06, | |
| "loss": 0.028, | |
| "reward": 1.946428644657135, | |
| "reward_std": 0.618572261929512, | |
| "rewards/accuracy_reward": 1.0303571939468383, | |
| "rewards/format_reward": 0.9160714685916901, | |
| "step": 945 | |
| }, | |
| { | |
| "completion_length": 376.62322998046875, | |
| "epoch": 0.7341576506955177, | |
| "grad_norm": 1.4602620601654053, | |
| "kl": 0.5408203125, | |
| "learning_rate": 1.0022969313130773e-06, | |
| "loss": 0.0216, | |
| "reward": 1.9642858266830445, | |
| "reward_std": 0.5609104305505752, | |
| "rewards/accuracy_reward": 1.028571480512619, | |
| "rewards/format_reward": 0.935714328289032, | |
| "step": 950 | |
| }, | |
| { | |
| "completion_length": 381.98126525878905, | |
| "epoch": 0.7380216383307573, | |
| "grad_norm": 2.153388023376465, | |
| "kl": 0.7783203125, | |
| "learning_rate": 9.754212751576386e-07, | |
| "loss": 0.0311, | |
| "reward": 1.8812500834465027, | |
| "reward_std": 0.6705505669116973, | |
| "rewards/accuracy_reward": 0.9785714745521545, | |
| "rewards/format_reward": 0.9026786148548126, | |
| "step": 955 | |
| }, | |
| { | |
| "completion_length": 364.32322387695314, | |
| "epoch": 0.7418856259659969, | |
| "grad_norm": 1.639176845550537, | |
| "kl": 0.65078125, | |
| "learning_rate": 9.488232558197732e-07, | |
| "loss": 0.0261, | |
| "reward": 1.8741072297096253, | |
| "reward_std": 0.6317374408245087, | |
| "rewards/accuracy_reward": 0.948214328289032, | |
| "rewards/format_reward": 0.9258929073810578, | |
| "step": 960 | |
| }, | |
| { | |
| "completion_length": 371.18751525878906, | |
| "epoch": 0.7457496136012365, | |
| "grad_norm": 1.5717941522598267, | |
| "kl": 0.4546875, | |
| "learning_rate": 9.225077169912644e-07, | |
| "loss": 0.0182, | |
| "reward": 1.9750001072883605, | |
| "reward_std": 0.5823389858007431, | |
| "rewards/accuracy_reward": 1.048214328289032, | |
| "rewards/format_reward": 0.9267857551574707, | |
| "step": 965 | |
| }, | |
| { | |
| "completion_length": 374.3330505371094, | |
| "epoch": 0.749613601236476, | |
| "grad_norm": 1.1975027322769165, | |
| "kl": 0.566796875, | |
| "learning_rate": 8.964794509221508e-07, | |
| "loss": 0.0227, | |
| "reward": 1.9089286565780639, | |
| "reward_std": 0.5614093959331512, | |
| "rewards/accuracy_reward": 0.9875000357627869, | |
| "rewards/format_reward": 0.9214286208152771, | |
| "step": 970 | |
| }, | |
| { | |
| "completion_length": 367.66876525878905, | |
| "epoch": 0.7534775888717156, | |
| "grad_norm": 1.064929723739624, | |
| "kl": 0.7947265625, | |
| "learning_rate": 8.707431975480221e-07, | |
| "loss": 0.0318, | |
| "reward": 1.9767858147621156, | |
| "reward_std": 0.611976683139801, | |
| "rewards/accuracy_reward": 1.0571429133415222, | |
| "rewards/format_reward": 0.9196428894996643, | |
| "step": 975 | |
| }, | |
| { | |
| "completion_length": 393.2553741455078, | |
| "epoch": 0.7573415765069552, | |
| "grad_norm": 1.5490120649337769, | |
| "kl": 0.845703125, | |
| "learning_rate": 8.453036436268458e-07, | |
| "loss": 0.0338, | |
| "reward": 1.8348215222358704, | |
| "reward_std": 0.6252253264188766, | |
| "rewards/accuracy_reward": 0.9321429014205933, | |
| "rewards/format_reward": 0.9026786148548126, | |
| "step": 980 | |
| }, | |
| { | |
| "completion_length": 385.7125183105469, | |
| "epoch": 0.7612055641421948, | |
| "grad_norm": 1.1330265998840332, | |
| "kl": 0.6615234375, | |
| "learning_rate": 8.20165421885469e-07, | |
| "loss": 0.0265, | |
| "reward": 1.8267858147621154, | |
| "reward_std": 0.641253513097763, | |
| "rewards/accuracy_reward": 0.935714328289032, | |
| "rewards/format_reward": 0.8910714685916901, | |
| "step": 985 | |
| }, | |
| { | |
| "completion_length": 347.54822998046876, | |
| "epoch": 0.7650695517774343, | |
| "grad_norm": 1.4882538318634033, | |
| "kl": 0.6056640625, | |
| "learning_rate": 7.953331101759706e-07, | |
| "loss": 0.0242, | |
| "reward": 1.9312501072883606, | |
| "reward_std": 0.6073622226715087, | |
| "rewards/accuracy_reward": 1.00357146859169, | |
| "rewards/format_reward": 0.9276786208152771, | |
| "step": 990 | |
| }, | |
| { | |
| "completion_length": 372.0375183105469, | |
| "epoch": 0.7689335394126738, | |
| "grad_norm": 1.4863989353179932, | |
| "kl": 0.7060546875, | |
| "learning_rate": 7.708112306419968e-07, | |
| "loss": 0.0282, | |
| "reward": 1.8142857909202577, | |
| "reward_std": 0.582288071513176, | |
| "rewards/accuracy_reward": 0.898214328289032, | |
| "rewards/format_reward": 0.9160714626312256, | |
| "step": 995 | |
| }, | |
| { | |
| "completion_length": 348.9973388671875, | |
| "epoch": 0.7727975270479135, | |
| "grad_norm": 1.5836974382400513, | |
| "kl": 0.795703125, | |
| "learning_rate": 7.466042488952521e-07, | |
| "loss": 0.0318, | |
| "reward": 1.958928656578064, | |
| "reward_std": 0.6530672192573548, | |
| "rewards/accuracy_reward": 1.0357143342494965, | |
| "rewards/format_reward": 0.9232143223285675, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.7727975270479135, | |
| "eval_completion_length": 376.2670003255208, | |
| "eval_kl": 0.5617838541666667, | |
| "eval_loss": 0.022667212411761284, | |
| "eval_reward": 2.011904811859131, | |
| "eval_reward_std": 0.5900813996791839, | |
| "eval_rewards/accuracy_reward": 1.0809524337450662, | |
| "eval_rewards/format_reward": 0.9309524138768513, | |
| "eval_runtime": 86.4556, | |
| "eval_samples_per_second": 1.145, | |
| "eval_steps_per_second": 0.046, | |
| "step": 1000 | |
| }, | |
| { | |
| "completion_length": 351.85447998046874, | |
| "epoch": 0.776661514683153, | |
| "grad_norm": 1.1455680131912231, | |
| "kl": 0.5607421875, | |
| "learning_rate": 7.227165732022717e-07, | |
| "loss": 0.0224, | |
| "reward": 2.1116072654724123, | |
| "reward_std": 0.5147848486900329, | |
| "rewards/accuracy_reward": 1.1607143521308898, | |
| "rewards/format_reward": 0.9508928954601288, | |
| "step": 1005 | |
| }, | |
| { | |
| "completion_length": 361.89020080566405, | |
| "epoch": 0.7805255023183926, | |
| "grad_norm": 1.188633680343628, | |
| "kl": 0.5509765625, | |
| "learning_rate": 6.991525536816498e-07, | |
| "loss": 0.022, | |
| "reward": 1.9598215103149415, | |
| "reward_std": 0.48385874927043915, | |
| "rewards/accuracy_reward": 1.0196429014205932, | |
| "rewards/format_reward": 0.9401786148548126, | |
| "step": 1010 | |
| }, | |
| { | |
| "completion_length": 373.9660919189453, | |
| "epoch": 0.7843894899536321, | |
| "grad_norm": 0.8180363774299622, | |
| "kl": 0.598828125, | |
| "learning_rate": 6.759164815118493e-07, | |
| "loss": 0.024, | |
| "reward": 1.9000000953674316, | |
| "reward_std": 0.5949198305606842, | |
| "rewards/accuracy_reward": 0.9625000357627869, | |
| "rewards/format_reward": 0.9375000417232513, | |
| "step": 1015 | |
| }, | |
| { | |
| "completion_length": 377.0375152587891, | |
| "epoch": 0.7882534775888718, | |
| "grad_norm": 0.7868214845657349, | |
| "kl": 0.5365234375, | |
| "learning_rate": 6.530125881497473e-07, | |
| "loss": 0.0215, | |
| "reward": 1.8812501072883605, | |
| "reward_std": 0.541077944636345, | |
| "rewards/accuracy_reward": 0.9357143342494965, | |
| "rewards/format_reward": 0.9455357491970062, | |
| "step": 1020 | |
| }, | |
| { | |
| "completion_length": 365.1553680419922, | |
| "epoch": 0.7921174652241113, | |
| "grad_norm": 2.3109092712402344, | |
| "kl": 0.604296875, | |
| "learning_rate": 6.30445044560056e-07, | |
| "loss": 0.0242, | |
| "reward": 1.9892858147621155, | |
| "reward_std": 0.5382134824991226, | |
| "rewards/accuracy_reward": 1.0464286267757417, | |
| "rewards/format_reward": 0.9428571820259094, | |
| "step": 1025 | |
| }, | |
| { | |
| "completion_length": 362.01787719726565, | |
| "epoch": 0.7959814528593508, | |
| "grad_norm": 1.4343385696411133, | |
| "kl": 0.7544921875, | |
| "learning_rate": 6.082179604557617e-07, | |
| "loss": 0.0302, | |
| "reward": 1.960714375972748, | |
| "reward_std": 0.5546221494674682, | |
| "rewards/accuracy_reward": 1.0303571939468383, | |
| "rewards/format_reward": 0.9303571820259094, | |
| "step": 1030 | |
| }, | |
| { | |
| "completion_length": 365.1625122070312, | |
| "epoch": 0.7998454404945904, | |
| "grad_norm": 1.025671362876892, | |
| "kl": 0.5177734375, | |
| "learning_rate": 5.863353835497137e-07, | |
| "loss": 0.0207, | |
| "reward": 1.940178644657135, | |
| "reward_std": 0.6024579167366028, | |
| "rewards/accuracy_reward": 1.0071429014205933, | |
| "rewards/format_reward": 0.9330357551574707, | |
| "step": 1035 | |
| }, | |
| { | |
| "completion_length": 372.9285888671875, | |
| "epoch": 0.80370942812983, | |
| "grad_norm": 0.9761889576911926, | |
| "kl": 0.61640625, | |
| "learning_rate": 5.648012988175075e-07, | |
| "loss": 0.0247, | |
| "reward": 1.8482143759727478, | |
| "reward_std": 0.6087313055992126, | |
| "rewards/accuracy_reward": 0.9375000476837159, | |
| "rewards/format_reward": 0.9107143342494964, | |
| "step": 1040 | |
| }, | |
| { | |
| "completion_length": 376.95359191894534, | |
| "epoch": 0.8075734157650696, | |
| "grad_norm": 1.214440941810608, | |
| "kl": 0.875390625, | |
| "learning_rate": 5.436196277717928e-07, | |
| "loss": 0.035, | |
| "reward": 1.90089293718338, | |
| "reward_std": 0.6438783019781112, | |
| "rewards/accuracy_reward": 0.9910714864730835, | |
| "rewards/format_reward": 0.9098214626312255, | |
| "step": 1045 | |
| }, | |
| { | |
| "completion_length": 365.9160919189453, | |
| "epoch": 0.8114374034003091, | |
| "grad_norm": 1.3267405033111572, | |
| "kl": 0.6033203125, | |
| "learning_rate": 5.227942277481363e-07, | |
| "loss": 0.0241, | |
| "reward": 1.9250001072883607, | |
| "reward_std": 0.5924921661615372, | |
| "rewards/accuracy_reward": 1.000000035762787, | |
| "rewards/format_reward": 0.9250000298023224, | |
| "step": 1050 | |
| }, | |
| { | |
| "completion_length": 380.3330535888672, | |
| "epoch": 0.8153013910355487, | |
| "grad_norm": 1.3830839395523071, | |
| "kl": 0.4916015625, | |
| "learning_rate": 5.023288912025742e-07, | |
| "loss": 0.0197, | |
| "reward": 1.9303572654724122, | |
| "reward_std": 0.5795833975076675, | |
| "rewards/accuracy_reward": 1.0000000476837159, | |
| "rewards/format_reward": 0.9303571820259094, | |
| "step": 1055 | |
| }, | |
| { | |
| "completion_length": 343.18394470214844, | |
| "epoch": 0.8191653786707882, | |
| "grad_norm": 1.7928149700164795, | |
| "kl": 0.4869140625, | |
| "learning_rate": 4.822273450209767e-07, | |
| "loss": 0.0195, | |
| "reward": 2.000000071525574, | |
| "reward_std": 0.5124461591243744, | |
| "rewards/accuracy_reward": 1.0571428954601287, | |
| "rewards/format_reward": 0.9428571879863739, | |
| "step": 1060 | |
| }, | |
| { | |
| "completion_length": 353.4276977539063, | |
| "epoch": 0.8230293663060279, | |
| "grad_norm": 0.9495754837989807, | |
| "kl": 0.83935546875, | |
| "learning_rate": 4.6249324984035863e-07, | |
| "loss": 0.0335, | |
| "reward": 1.9812500834465028, | |
| "reward_std": 0.5092057317495347, | |
| "rewards/accuracy_reward": 1.0214286267757415, | |
| "rewards/format_reward": 0.95982146859169, | |
| "step": 1065 | |
| }, | |
| { | |
| "completion_length": 375.46876831054686, | |
| "epoch": 0.8268933539412674, | |
| "grad_norm": 0.7771401405334473, | |
| "kl": 0.4986328125, | |
| "learning_rate": 4.431301993822471e-07, | |
| "loss": 0.02, | |
| "reward": 2.0125001072883606, | |
| "reward_std": 0.5730317920446396, | |
| "rewards/accuracy_reward": 1.0839286148548126, | |
| "rewards/format_reward": 0.9285714745521545, | |
| "step": 1070 | |
| }, | |
| { | |
| "completion_length": 371.25269470214846, | |
| "epoch": 0.8307573415765069, | |
| "grad_norm": 1.0410772562026978, | |
| "kl": 0.4998046875, | |
| "learning_rate": 4.2414171979824e-07, | |
| "loss": 0.02, | |
| "reward": 1.9660715341567994, | |
| "reward_std": 0.5737438589334488, | |
| "rewards/accuracy_reward": 1.0250000476837158, | |
| "rewards/format_reward": 0.9410714745521546, | |
| "step": 1075 | |
| }, | |
| { | |
| "completion_length": 367.7830474853516, | |
| "epoch": 0.8346213292117465, | |
| "grad_norm": 1.4777237176895142, | |
| "kl": 0.54140625, | |
| "learning_rate": 4.055312690278701e-07, | |
| "loss": 0.0217, | |
| "reward": 1.9062500953674317, | |
| "reward_std": 0.5742262482643128, | |
| "rewards/accuracy_reward": 0.9714286029338837, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 1080 | |
| }, | |
| { | |
| "completion_length": 371.16876831054685, | |
| "epoch": 0.8384853168469861, | |
| "grad_norm": 0.8883810043334961, | |
| "kl": 0.453125, | |
| "learning_rate": 3.8730223616888634e-07, | |
| "loss": 0.0181, | |
| "reward": 1.927678680419922, | |
| "reward_std": 0.47931237816810607, | |
| "rewards/accuracy_reward": 0.9821428954601288, | |
| "rewards/format_reward": 0.9455357551574707, | |
| "step": 1085 | |
| }, | |
| { | |
| "completion_length": 372.65984497070315, | |
| "epoch": 0.8423493044822257, | |
| "grad_norm": 1.4345813989639282, | |
| "kl": 0.4869140625, | |
| "learning_rate": 3.6945794086007706e-07, | |
| "loss": 0.0195, | |
| "reward": 2.0116072297096252, | |
| "reward_std": 0.565495365858078, | |
| "rewards/accuracy_reward": 1.080357199907303, | |
| "rewards/format_reward": 0.9312500357627869, | |
| "step": 1090 | |
| }, | |
| { | |
| "completion_length": 370.21876525878906, | |
| "epoch": 0.8462132921174652, | |
| "grad_norm": 1.1525709629058838, | |
| "kl": 0.47900390625, | |
| "learning_rate": 3.520016326767381e-07, | |
| "loss": 0.0191, | |
| "reward": 1.9973214983940124, | |
| "reward_std": 0.5897004574537277, | |
| "rewards/accuracy_reward": 1.067857176065445, | |
| "rewards/format_reward": 0.9294643342494965, | |
| "step": 1095 | |
| }, | |
| { | |
| "completion_length": 359.56609802246095, | |
| "epoch": 0.8500772797527048, | |
| "grad_norm": 1.2896158695220947, | |
| "kl": 0.5654296875, | |
| "learning_rate": 3.3493649053890325e-07, | |
| "loss": 0.0226, | |
| "reward": 1.9857143998146056, | |
| "reward_std": 0.4953633636236191, | |
| "rewards/accuracy_reward": 1.0446428954601288, | |
| "rewards/format_reward": 0.9410714626312255, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.8500772797527048, | |
| "eval_completion_length": 358.0001708984375, | |
| "eval_kl": 0.4869791666666667, | |
| "eval_loss": 0.019230343401432037, | |
| "eval_reward": 1.9404762903849284, | |
| "eval_reward_std": 0.522600182890892, | |
| "eval_rewards/accuracy_reward": 1.009523856639862, | |
| "eval_rewards/format_reward": 0.9309524138768513, | |
| "eval_runtime": 90.3216, | |
| "eval_samples_per_second": 1.096, | |
| "eval_steps_per_second": 0.044, | |
| "step": 1100 | |
| }, | |
| { | |
| "completion_length": 364.4919799804687, | |
| "epoch": 0.8539412673879444, | |
| "grad_norm": 1.5501704216003418, | |
| "kl": 0.46357421875, | |
| "learning_rate": 3.182656221324384e-07, | |
| "loss": 0.0185, | |
| "reward": 1.9089286923408508, | |
| "reward_std": 0.5076944470405579, | |
| "rewards/accuracy_reward": 0.9750000536441803, | |
| "rewards/format_reward": 0.9339286088943481, | |
| "step": 1105 | |
| }, | |
| { | |
| "completion_length": 349.36341247558596, | |
| "epoch": 0.8578052550231839, | |
| "grad_norm": 0.8084538578987122, | |
| "kl": 0.41845703125, | |
| "learning_rate": 3.019920633431095e-07, | |
| "loss": 0.0167, | |
| "reward": 2.065178656578064, | |
| "reward_std": 0.4948657438158989, | |
| "rewards/accuracy_reward": 1.1071428954601288, | |
| "rewards/format_reward": 0.9580357551574707, | |
| "step": 1110 | |
| }, | |
| { | |
| "completion_length": 371.67323303222656, | |
| "epoch": 0.8616692426584235, | |
| "grad_norm": 1.2597169876098633, | |
| "kl": 0.4419921875, | |
| "learning_rate": 2.861187777037269e-07, | |
| "loss": 0.0177, | |
| "reward": 1.9366072297096253, | |
| "reward_std": 0.5391754776239395, | |
| "rewards/accuracy_reward": 0.9982143342494965, | |
| "rewards/format_reward": 0.9383928954601288, | |
| "step": 1115 | |
| }, | |
| { | |
| "completion_length": 374.87412414550784, | |
| "epoch": 0.865533230293663, | |
| "grad_norm": 0.7964634895324707, | |
| "kl": 0.5240234375, | |
| "learning_rate": 2.706486558544644e-07, | |
| "loss": 0.0209, | |
| "reward": 1.927678644657135, | |
| "reward_std": 0.5668253153562546, | |
| "rewards/accuracy_reward": 0.9928571820259094, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 1120 | |
| }, | |
| { | |
| "completion_length": 357.7071563720703, | |
| "epoch": 0.8693972179289027, | |
| "grad_norm": 0.9206239581108093, | |
| "kl": 0.470703125, | |
| "learning_rate": 2.55584515016451e-07, | |
| "loss": 0.0188, | |
| "reward": 1.9598214864730834, | |
| "reward_std": 0.546750283241272, | |
| "rewards/accuracy_reward": 1.0267857611179352, | |
| "rewards/format_reward": 0.9330357670783996, | |
| "step": 1125 | |
| }, | |
| { | |
| "completion_length": 366.42144775390625, | |
| "epoch": 0.8732612055641422, | |
| "grad_norm": 1.1973419189453125, | |
| "kl": 0.5232421875, | |
| "learning_rate": 2.4092909847873713e-07, | |
| "loss": 0.0209, | |
| "reward": 1.9580357909202575, | |
| "reward_std": 0.5517275601625442, | |
| "rewards/accuracy_reward": 1.023214340209961, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 1130 | |
| }, | |
| { | |
| "completion_length": 365.56787109375, | |
| "epoch": 0.8771251931993818, | |
| "grad_norm": 0.7583184838294983, | |
| "kl": 0.587109375, | |
| "learning_rate": 2.2668507509871957e-07, | |
| "loss": 0.0235, | |
| "reward": 1.894642949104309, | |
| "reward_std": 0.5627562046051026, | |
| "rewards/accuracy_reward": 0.9553571879863739, | |
| "rewards/format_reward": 0.9392857491970062, | |
| "step": 1135 | |
| }, | |
| { | |
| "completion_length": 354.81340942382815, | |
| "epoch": 0.8809891808346213, | |
| "grad_norm": 2.083658218383789, | |
| "kl": 0.714453125, | |
| "learning_rate": 2.128550388161263e-07, | |
| "loss": 0.0286, | |
| "reward": 1.93571435213089, | |
| "reward_std": 0.570975062251091, | |
| "rewards/accuracy_reward": 1.0071428775787354, | |
| "rewards/format_reward": 0.92857146859169, | |
| "step": 1140 | |
| }, | |
| { | |
| "completion_length": 351.0473327636719, | |
| "epoch": 0.884853168469861, | |
| "grad_norm": 1.6536860466003418, | |
| "kl": 0.641796875, | |
| "learning_rate": 1.9944150818063667e-07, | |
| "loss": 0.0257, | |
| "reward": 1.952678632736206, | |
| "reward_std": 0.5870893836021424, | |
| "rewards/accuracy_reward": 1.0178571820259095, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 1145 | |
| }, | |
| { | |
| "completion_length": 371.1955535888672, | |
| "epoch": 0.8887171561051005, | |
| "grad_norm": 0.8971360325813293, | |
| "kl": 0.6400390625, | |
| "learning_rate": 1.864469258932397e-07, | |
| "loss": 0.0256, | |
| "reward": 1.9348215222358705, | |
| "reward_std": 0.5955395519733429, | |
| "rewards/accuracy_reward": 1.008928608894348, | |
| "rewards/format_reward": 0.9258928954601288, | |
| "step": 1150 | |
| }, | |
| { | |
| "completion_length": 350.9419799804688, | |
| "epoch": 0.89258114374034, | |
| "grad_norm": 1.2180782556533813, | |
| "kl": 0.6533203125, | |
| "learning_rate": 1.7387365836139785e-07, | |
| "loss": 0.0261, | |
| "reward": 1.952678668498993, | |
| "reward_std": 0.5585884839296341, | |
| "rewards/accuracy_reward": 1.005357176065445, | |
| "rewards/format_reward": 0.9473214626312256, | |
| "step": 1155 | |
| }, | |
| { | |
| "completion_length": 376.7419769287109, | |
| "epoch": 0.8964451313755796, | |
| "grad_norm": 1.8317776918411255, | |
| "kl": 0.723046875, | |
| "learning_rate": 1.6172399526810822e-07, | |
| "loss": 0.0289, | |
| "reward": 1.846428632736206, | |
| "reward_std": 0.5882213652133942, | |
| "rewards/accuracy_reward": 0.9250000357627869, | |
| "rewards/format_reward": 0.9214286208152771, | |
| "step": 1160 | |
| }, | |
| { | |
| "completion_length": 366.1000244140625, | |
| "epoch": 0.9003091190108191, | |
| "grad_norm": 2.5519237518310547, | |
| "kl": 0.687109375, | |
| "learning_rate": 1.5000014915493467e-07, | |
| "loss": 0.0275, | |
| "reward": 1.9366072297096253, | |
| "reward_std": 0.6008837521076202, | |
| "rewards/accuracy_reward": 1.003571480512619, | |
| "rewards/format_reward": 0.9330357491970063, | |
| "step": 1165 | |
| }, | |
| { | |
| "completion_length": 355.3768035888672, | |
| "epoch": 0.9041731066460588, | |
| "grad_norm": 1.4227640628814697, | |
| "kl": 0.80859375, | |
| "learning_rate": 1.3870425501908674e-07, | |
| "loss": 0.0324, | |
| "reward": 1.908928680419922, | |
| "reward_std": 0.5863888055086136, | |
| "rewards/accuracy_reward": 0.9839286148548126, | |
| "rewards/format_reward": 0.9250000476837158, | |
| "step": 1170 | |
| }, | |
| { | |
| "completion_length": 355.98841247558596, | |
| "epoch": 0.9080370942812983, | |
| "grad_norm": 2.7810275554656982, | |
| "kl": 0.644921875, | |
| "learning_rate": 1.278383699246244e-07, | |
| "loss": 0.0258, | |
| "reward": 1.9991072177886964, | |
| "reward_std": 0.549174913764, | |
| "rewards/accuracy_reward": 1.0678571879863739, | |
| "rewards/format_reward": 0.9312500476837158, | |
| "step": 1175 | |
| }, | |
| { | |
| "completion_length": 354.95090942382814, | |
| "epoch": 0.9119010819165378, | |
| "grad_norm": 1.349503517150879, | |
| "kl": 0.557421875, | |
| "learning_rate": 1.1740447262784782e-07, | |
| "loss": 0.0223, | |
| "reward": 1.9553572297096253, | |
| "reward_std": 0.6008891820907593, | |
| "rewards/accuracy_reward": 1.02857146859169, | |
| "rewards/format_reward": 0.9267857551574707, | |
| "step": 1180 | |
| }, | |
| { | |
| "completion_length": 358.05895080566404, | |
| "epoch": 0.9157650695517774, | |
| "grad_norm": 1.2601906061172485, | |
| "kl": 0.625, | |
| "learning_rate": 1.0740446321695408e-07, | |
| "loss": 0.025, | |
| "reward": 1.9705358266830444, | |
| "reward_std": 0.6071323782205582, | |
| "rewards/accuracy_reward": 1.0517857551574707, | |
| "rewards/format_reward": 0.9187500417232514, | |
| "step": 1185 | |
| }, | |
| { | |
| "completion_length": 356.47412109375, | |
| "epoch": 0.919629057187017, | |
| "grad_norm": 1.0005254745483398, | |
| "kl": 0.508984375, | |
| "learning_rate": 9.78401627660161e-08, | |
| "loss": 0.0204, | |
| "reward": 1.9919643878936768, | |
| "reward_std": 0.5772496670484543, | |
| "rewards/accuracy_reward": 1.0553572058677674, | |
| "rewards/format_reward": 0.9366071939468383, | |
| "step": 1190 | |
| }, | |
| { | |
| "completion_length": 352.22947692871094, | |
| "epoch": 0.9234930448222566, | |
| "grad_norm": 2.757625102996826, | |
| "kl": 0.5763671875, | |
| "learning_rate": 8.871331300335322e-08, | |
| "loss": 0.023, | |
| "reward": 2.016964375972748, | |
| "reward_std": 0.5439181506633759, | |
| "rewards/accuracy_reward": 1.0732143223285675, | |
| "rewards/format_reward": 0.9437500476837158, | |
| "step": 1195 | |
| }, | |
| { | |
| "completion_length": 354.0669860839844, | |
| "epoch": 0.9273570324574961, | |
| "grad_norm": 1.34922456741333, | |
| "kl": 0.555078125, | |
| "learning_rate": 8.002557599434802e-08, | |
| "loss": 0.0222, | |
| "reward": 1.9580357670783997, | |
| "reward_std": 0.5712159514427185, | |
| "rewards/accuracy_reward": 1.023214328289032, | |
| "rewards/format_reward": 0.93482146859169, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.9273570324574961, | |
| "eval_completion_length": 342.553505452474, | |
| "eval_kl": 0.5052734375, | |
| "eval_loss": 0.02041592448949814, | |
| "eval_reward": 2.0404762744903566, | |
| "eval_reward_std": 0.4870568384726842, | |
| "eval_rewards/accuracy_reward": 1.085714336236318, | |
| "eval_rewards/format_reward": 0.9547619263331095, | |
| "eval_runtime": 80.4099, | |
| "eval_samples_per_second": 1.231, | |
| "eval_steps_per_second": 0.05, | |
| "step": 1200 | |
| }, | |
| { | |
| "completion_length": 378.85805358886716, | |
| "epoch": 0.9312210200927357, | |
| "grad_norm": 2.145590305328369, | |
| "kl": 0.67890625, | |
| "learning_rate": 7.177853383877498e-08, | |
| "loss": 0.0272, | |
| "reward": 1.8732143640518188, | |
| "reward_std": 0.5757515370845795, | |
| "rewards/accuracy_reward": 0.9535714745521545, | |
| "rewards/format_reward": 0.9196429073810577, | |
| "step": 1205 | |
| }, | |
| { | |
| "completion_length": 346.4785888671875, | |
| "epoch": 0.9350850077279753, | |
| "grad_norm": 1.1914241313934326, | |
| "kl": 0.5564453125, | |
| "learning_rate": 6.397368838268497e-08, | |
| "loss": 0.0223, | |
| "reward": 1.9446429371833802, | |
| "reward_std": 0.49446034133434297, | |
| "rewards/accuracy_reward": 0.9946429014205933, | |
| "rewards/format_reward": 0.9500000357627869, | |
| "step": 1210 | |
| }, | |
| { | |
| "completion_length": 355.9580535888672, | |
| "epoch": 0.9389489953632149, | |
| "grad_norm": 1.2249726057052612, | |
| "kl": 0.544140625, | |
| "learning_rate": 5.661246094491146e-08, | |
| "loss": 0.0218, | |
| "reward": 1.9660715103149413, | |
| "reward_std": 0.5542481303215027, | |
| "rewards/accuracy_reward": 1.0303571760654449, | |
| "rewards/format_reward": 0.935714328289032, | |
| "step": 1215 | |
| }, | |
| { | |
| "completion_length": 365.9223388671875, | |
| "epoch": 0.9428129829984544, | |
| "grad_norm": 1.0840822458267212, | |
| "kl": 0.63984375, | |
| "learning_rate": 4.969619205823617e-08, | |
| "loss": 0.0256, | |
| "reward": 1.9410715103149414, | |
| "reward_std": 0.5451422989368438, | |
| "rewards/accuracy_reward": 0.9946429073810578, | |
| "rewards/format_reward": 0.9464286208152771, | |
| "step": 1220 | |
| }, | |
| { | |
| "completion_length": 347.46519165039064, | |
| "epoch": 0.9466769706336939, | |
| "grad_norm": 1.4779506921768188, | |
| "kl": 0.5994140625, | |
| "learning_rate": 4.3226141225268804e-08, | |
| "loss": 0.024, | |
| "reward": 1.9232143759727478, | |
| "reward_std": 0.5508476465940475, | |
| "rewards/accuracy_reward": 0.980357187986374, | |
| "rewards/format_reward": 0.9428571879863739, | |
| "step": 1225 | |
| }, | |
| { | |
| "completion_length": 358.8866302490234, | |
| "epoch": 0.9505409582689336, | |
| "grad_norm": 0.8061991930007935, | |
| "kl": 0.5189453125, | |
| "learning_rate": 3.7203486689083857e-08, | |
| "loss": 0.0208, | |
| "reward": 1.983928668498993, | |
| "reward_std": 0.6267418980598449, | |
| "rewards/accuracy_reward": 1.0500000536441803, | |
| "rewards/format_reward": 0.9339286088943481, | |
| "step": 1230 | |
| }, | |
| { | |
| "completion_length": 342.37412414550784, | |
| "epoch": 0.9544049459041731, | |
| "grad_norm": 1.062484622001648, | |
| "kl": 0.5380859375, | |
| "learning_rate": 3.1629325218651695e-08, | |
| "loss": 0.0215, | |
| "reward": 2.0116072297096252, | |
| "reward_std": 0.501123908162117, | |
| "rewards/accuracy_reward": 1.060714316368103, | |
| "rewards/format_reward": 0.9508928954601288, | |
| "step": 1235 | |
| }, | |
| { | |
| "completion_length": 353.96519775390624, | |
| "epoch": 0.9582689335394127, | |
| "grad_norm": 1.2145483493804932, | |
| "kl": 0.5751953125, | |
| "learning_rate": 2.6504671909109993e-08, | |
| "loss": 0.023, | |
| "reward": 1.9794643640518188, | |
| "reward_std": 0.4939621418714523, | |
| "rewards/accuracy_reward": 1.0357143342494965, | |
| "rewards/format_reward": 0.9437500417232514, | |
| "step": 1240 | |
| }, | |
| { | |
| "completion_length": 357.8018035888672, | |
| "epoch": 0.9621329211746522, | |
| "grad_norm": 0.7816463708877563, | |
| "kl": 0.5390625, | |
| "learning_rate": 2.1830459996908527e-08, | |
| "loss": 0.0216, | |
| "reward": 1.9553572177886962, | |
| "reward_std": 0.5724190145730972, | |
| "rewards/accuracy_reward": 1.0125000536441804, | |
| "rewards/format_reward": 0.9428571820259094, | |
| "step": 1245 | |
| }, | |
| { | |
| "completion_length": 347.95180053710936, | |
| "epoch": 0.9659969088098919, | |
| "grad_norm": 1.0017303228378296, | |
| "kl": 0.5392578125, | |
| "learning_rate": 1.7607540689859036e-08, | |
| "loss": 0.0216, | |
| "reward": 1.9767858147621156, | |
| "reward_std": 0.5618195921182633, | |
| "rewards/accuracy_reward": 1.0446429073810577, | |
| "rewards/format_reward": 0.9321429073810578, | |
| "step": 1250 | |
| }, | |
| { | |
| "completion_length": 357.79555053710936, | |
| "epoch": 0.9698608964451314, | |
| "grad_norm": 1.0843381881713867, | |
| "kl": 0.5859375, | |
| "learning_rate": 1.383668301212393e-08, | |
| "loss": 0.0234, | |
| "reward": 1.9732144117355346, | |
| "reward_std": 0.5734216451644898, | |
| "rewards/accuracy_reward": 1.0410714626312256, | |
| "rewards/format_reward": 0.9321428954601287, | |
| "step": 1255 | |
| }, | |
| { | |
| "completion_length": 338.9107269287109, | |
| "epoch": 0.973724884080371, | |
| "grad_norm": 0.9808955192565918, | |
| "kl": 0.611328125, | |
| "learning_rate": 1.0518573664172193e-08, | |
| "loss": 0.0245, | |
| "reward": 1.9794643878936768, | |
| "reward_std": 0.566856500506401, | |
| "rewards/accuracy_reward": 1.035714340209961, | |
| "rewards/format_reward": 0.9437500417232514, | |
| "step": 1260 | |
| }, | |
| { | |
| "completion_length": 368.7884094238281, | |
| "epoch": 0.9775888717156105, | |
| "grad_norm": 1.4401775598526, | |
| "kl": 0.498046875, | |
| "learning_rate": 7.653816897725819e-09, | |
| "loss": 0.0199, | |
| "reward": 1.9937500834465027, | |
| "reward_std": 0.5333459317684174, | |
| "rewards/accuracy_reward": 1.04821435213089, | |
| "rewards/format_reward": 0.9455357551574707, | |
| "step": 1265 | |
| }, | |
| { | |
| "completion_length": 365.1580535888672, | |
| "epoch": 0.98145285935085, | |
| "grad_norm": 1.1229665279388428, | |
| "kl": 0.6189453125, | |
| "learning_rate": 5.242934405720879e-09, | |
| "loss": 0.0248, | |
| "reward": 1.9437500715255738, | |
| "reward_std": 0.5873760730028152, | |
| "rewards/accuracy_reward": 0.9982143342494965, | |
| "rewards/format_reward": 0.9455357670783997, | |
| "step": 1270 | |
| }, | |
| { | |
| "completion_length": 360.70001831054685, | |
| "epoch": 0.9853168469860897, | |
| "grad_norm": 1.2572500705718994, | |
| "kl": 0.56953125, | |
| "learning_rate": 3.286365227304633e-09, | |
| "loss": 0.0228, | |
| "reward": 2.0000000953674317, | |
| "reward_std": 0.5724943190813064, | |
| "rewards/accuracy_reward": 1.0642857611179353, | |
| "rewards/format_reward": 0.935714328289032, | |
| "step": 1275 | |
| }, | |
| { | |
| "completion_length": 357.1821533203125, | |
| "epoch": 0.9891808346213292, | |
| "grad_norm": 0.9178944826126099, | |
| "kl": 0.537109375, | |
| "learning_rate": 1.7844656678817074e-09, | |
| "loss": 0.0215, | |
| "reward": 1.9071429371833801, | |
| "reward_std": 0.5804499536752701, | |
| "rewards/accuracy_reward": 0.9732143223285675, | |
| "rewards/format_reward": 0.9339286088943481, | |
| "step": 1280 | |
| }, | |
| { | |
| "completion_length": 360.2035888671875, | |
| "epoch": 0.9930448222565688, | |
| "grad_norm": 1.0505741834640503, | |
| "kl": 0.52841796875, | |
| "learning_rate": 7.375092342298828e-10, | |
| "loss": 0.0212, | |
| "reward": 1.9723215103149414, | |
| "reward_std": 0.5573354363441467, | |
| "rewards/accuracy_reward": 1.0339286267757415, | |
| "rewards/format_reward": 0.9383928954601288, | |
| "step": 1285 | |
| }, | |
| { | |
| "completion_length": 353.73841247558596, | |
| "epoch": 0.9969088098918083, | |
| "grad_norm": 1.0907429456710815, | |
| "kl": 0.5439453125, | |
| "learning_rate": 1.4568658469132913e-10, | |
| "loss": 0.0218, | |
| "reward": 2.0357143759727476, | |
| "reward_std": 0.5376188308000565, | |
| "rewards/accuracy_reward": 1.0821429014205932, | |
| "rewards/format_reward": 0.9535714745521545, | |
| "step": 1290 | |
| }, | |
| { | |
| "completion_length": 359.22038650512695, | |
| "epoch": 1.0, | |
| "kl": 0.52099609375, | |
| "reward": 1.9821429252624512, | |
| "reward_std": 0.5921385176479816, | |
| "rewards/accuracy_reward": 1.051339328289032, | |
| "rewards/format_reward": 0.9308036118745804, | |
| "step": 1294, | |
| "total_flos": 0.0, | |
| "train_loss": 0.7486699576652505, | |
| "train_runtime": 29158.9846, | |
| "train_samples_per_second": 2.484, | |
| "train_steps_per_second": 0.044 | |
| } | |
| ], | |
| "logging_steps": 5, | |
| "max_steps": 1294, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": false, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |