{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 6.25, "eval_steps": 20, "global_step": 100, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1328125, "completions/max_length": 3072.0, "completions/max_terminated_length": 3067.0, "completions/mean_length": 1819.9921875, "completions/mean_terminated_length": 1628.2432861328125, "completions/min_length": 179.0, "completions/min_terminated_length": 179.0, "entropy": 0.41793088614940643, "epoch": 0.0625, "frac_reward_zero_std": 0.125, "grad_norm": 0.06122519448399544, "kl": 0.0, "learning_rate": 1e-06, "loss": 0.005, "num_tokens": 252799.0, "reward": -0.13906250894069672, "reward_std": 0.1776094138622284, "rewards/accuracy_reward/mean": 0.0390625, "rewards/accuracy_reward/std": 0.194504976272583, "rewards/format_reward_func/mean": -0.890625, "rewards/format_reward_func/std": 0.3375309705734253, "step": 1 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.126953125, "completions/max_length": 3072.0, "completions/max_terminated_length": 2972.5, "completions/mean_length": 1790.494140625, "completions/mean_terminated_length": 1603.4855651855469, "completions/min_length": 52.5, "completions/min_terminated_length": 52.5, "entropy": 0.4873259887099266, "epoch": 0.3125, "frac_reward_zero_std": 0.09375, "grad_norm": 0.06680382788181305, "kl": 0.00013216794150139322, "learning_rate": 9.6e-07, "loss": 0.0075, "num_tokens": 1248892.0, "reward": -0.11953125149011612, "reward_std": 0.21148933842778206, "rewards/accuracy_reward/mean": 0.05859375, "rewards/accuracy_reward/std": 0.23225979879498482, "rewards/format_reward_func/mean": -0.890625, "rewards/format_reward_func/std": 0.31732235848903656, "step": 5 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1546875, "completions/max_length": 3072.0, "completions/max_terminated_length": 2953.4, "completions/mean_length": 1825.1890625, "completions/mean_terminated_length": 1597.2147216796875, "completions/min_length": 42.2, "completions/min_terminated_length": 42.2, "entropy": 0.45153772197663783, "epoch": 0.625, "frac_reward_zero_std": 0.125, "grad_norm": 0.05490436404943466, "kl": 0.0001304413024627138, "learning_rate": 9.1e-07, "loss": 0.0088, "num_tokens": 2516213.0, "reward": -0.12281250208616257, "reward_std": 0.1975747048854828, "rewards/accuracy_reward/mean": 0.0515625, "rewards/accuracy_reward/std": 0.21566526144742965, "rewards/format_reward_func/mean": -0.871875, "rewards/format_reward_func/std": 0.3422063410282135, "step": 10 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.15, "completions/max_length": 3072.0, "completions/max_terminated_length": 2996.4, "completions/mean_length": 1820.44375, "completions/mean_terminated_length": 1599.5802001953125, "completions/min_length": 36.4, "completions/min_terminated_length": 36.4, "entropy": 0.5154426731169224, "epoch": 0.9375, "frac_reward_zero_std": 0.125, "grad_norm": 0.061234962195158005, "kl": 0.00014891433747834525, "learning_rate": 8.599999999999999e-07, "loss": 0.0044, "num_tokens": 3780497.0, "reward": -0.12312500029802323, "reward_std": 0.2081974685192108, "rewards/accuracy_reward/mean": 0.0546875, "rewards/accuracy_reward/std": 0.2231001377105713, "rewards/format_reward_func/mean": -0.8890625, "rewards/format_reward_func/std": 0.3250035583972931, "step": 15 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1296875, "completions/max_length": 3072.0, "completions/max_terminated_length": 2985.8, "completions/mean_length": 1772.940625, "completions/mean_terminated_length": 1580.4586181640625, "completions/min_length": 58.2, "completions/min_terminated_length": 58.2, "entropy": 0.4530036810785532, "epoch": 1.25, "frac_reward_zero_std": 0.075, "grad_norm": 0.05929873138666153, "kl": 0.00016962703539320502, "learning_rate": 8.1e-07, "loss": 0.014, "num_tokens": 5014379.0, "reward": -0.10468750149011612, "reward_std": 0.2389657288789749, "rewards/accuracy_reward/mean": 0.0671875, "rewards/accuracy_reward/std": 0.24869972169399263, "rewards/format_reward_func/mean": -0.859375, "rewards/format_reward_func/std": 0.3756451666355133, "step": 20 }, { "epoch": 1.25, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.096875, "eval_completions/max_length": 3050.4, "eval_completions/max_terminated_length": 2570.75, "eval_completions/mean_length": 1034.7015625, "eval_completions/mean_terminated_length": 817.477993774414, "eval_completions/min_length": 41.1, "eval_completions/min_terminated_length": 41.1, "eval_entropy": 1.047258096933365, "eval_frac_reward_zero_std": 0.05, "eval_kl": 0.00024451220015180296, "eval_loss": -0.0009000980644486845, "eval_num_tokens": 5014379.0, "eval_reward": -0.02437500040978193, "eval_reward_std": 0.3017435781657696, "eval_rewards/accuracy_reward/mean": 0.1390625, "eval_rewards/accuracy_reward/std": 0.32395628839731216, "eval_rewards/format_reward_func/mean": -0.8171875, "eval_rewards/format_reward_func/std": 0.4204400397837162, "eval_runtime": 328.568, "eval_samples_per_second": 0.122, "eval_steps_per_second": 0.006, "step": 20 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.128125, "completions/max_length": 3072.0, "completions/max_terminated_length": 3050.0, "completions/mean_length": 1773.2953125, "completions/mean_terminated_length": 1582.642919921875, "completions/min_length": 52.2, "completions/min_terminated_length": 52.2, "entropy": 0.44564105086028577, "epoch": 1.5625, "frac_reward_zero_std": 0.025, "grad_norm": 0.06874161213636398, "kl": 0.00019518628796504344, "learning_rate": 7.599999999999999e-07, "loss": 0.0009, "num_tokens": 6248488.0, "reward": -0.11281250268220902, "reward_std": 0.231013685464859, "rewards/accuracy_reward/mean": 0.0625, "rewards/accuracy_reward/std": 0.24069232046604155, "rewards/format_reward_func/mean": -0.8765625, "rewards/format_reward_func/std": 0.3404365301132202, "step": 25 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.14375, "completions/max_length": 3072.0, "completions/max_terminated_length": 3040.0, "completions/mean_length": 1823.740625, "completions/mean_terminated_length": 1614.19853515625, "completions/min_length": 55.2, "completions/min_terminated_length": 55.2, "entropy": 0.41670549251139166, "epoch": 1.875, "frac_reward_zero_std": 0.025, "grad_norm": 0.0631536990404129, "kl": 0.00021747380051237997, "learning_rate": 7.1e-07, "loss": 0.0031, "num_tokens": 7514882.0, "reward": -0.08718750402331352, "reward_std": 0.25979888141155244, "rewards/accuracy_reward/mean": 0.0859375, "rewards/accuracy_reward/std": 0.2747330546379089, "rewards/format_reward_func/mean": -0.865625, "rewards/format_reward_func/std": 0.34926688075065615, "step": 30 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1484375, "completions/max_length": 3072.0, "completions/max_terminated_length": 2927.2, "completions/mean_length": 1842.6109375, "completions/mean_terminated_length": 1627.85, "completions/min_length": 63.4, "completions/min_terminated_length": 63.4, "entropy": 0.40708271935582163, "epoch": 2.1875, "frac_reward_zero_std": 0.05, "grad_norm": 0.06106453761458397, "kl": 0.00025953962140192746, "learning_rate": 6.6e-07, "loss": 0.0059, "num_tokens": 8793353.0, "reward": -0.09406250417232513, "reward_std": 0.2604650527238846, "rewards/accuracy_reward/mean": 0.075, "rewards/accuracy_reward/std": 0.2643246352672577, "rewards/format_reward_func/mean": -0.8453125, "rewards/format_reward_func/std": 0.37586394548416135, "step": 35 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1578125, "completions/max_length": 3072.0, "completions/max_terminated_length": 3014.2, "completions/mean_length": 1870.421875, "completions/mean_terminated_length": 1645.2918701171875, "completions/min_length": 95.0, "completions/min_terminated_length": 95.0, "entropy": 0.4321233291178942, "epoch": 2.5, "frac_reward_zero_std": 0.025, "grad_norm": 0.0598464198410511, "kl": 0.0003444181191298412, "learning_rate": 6.1e-07, "loss": 0.0075, "num_tokens": 10089623.0, "reward": -0.0953125, "reward_std": 0.24980489909648895, "rewards/accuracy_reward/mean": 0.0765625, "rewards/accuracy_reward/std": 0.26329803466796875, "rewards/format_reward_func/mean": -0.859375, "rewards/format_reward_func/std": 0.3709796071052551, "step": 40 }, { "epoch": 2.5, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.096875, "eval_completions/max_length": 3052.45, "eval_completions/max_terminated_length": 2589.7, "eval_completions/mean_length": 1035.0125, "eval_completions/mean_terminated_length": 818.5466033935547, "eval_completions/min_length": 44.3, "eval_completions/min_terminated_length": 44.3, "eval_entropy": 0.9328817576169968, "eval_frac_reward_zero_std": 0.0, "eval_kl": 0.0003855928880511783, "eval_loss": 0.02006949670612812, "eval_num_tokens": 10089623.0, "eval_reward": -0.014062502048909664, "eval_reward_std": 0.310027738660574, "eval_rewards/accuracy_reward/mean": 0.146875, "eval_rewards/accuracy_reward/std": 0.32917849868536, "eval_rewards/format_reward_func/mean": -0.8046875, "eval_rewards/format_reward_func/std": 0.43316270783543587, "eval_runtime": 323.5959, "eval_samples_per_second": 0.124, "eval_steps_per_second": 0.006, "step": 40 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.15625, "completions/max_length": 3072.0, "completions/max_terminated_length": 2965.8, "completions/mean_length": 1803.6140625, "completions/mean_terminated_length": 1568.3510009765625, "completions/min_length": 21.0, "completions/min_terminated_length": 21.0, "entropy": 0.3385248489677906, "epoch": 2.8125, "frac_reward_zero_std": 0.05, "grad_norm": 0.048809297382831573, "kl": 0.00041367213452758733, "learning_rate": 5.6e-07, "loss": 0.019, "num_tokens": 11343136.0, "reward": -0.08250000327825546, "reward_std": 0.2559084087610245, "rewards/accuracy_reward/mean": 0.0859375, "rewards/accuracy_reward/std": 0.27599324882030485, "rewards/format_reward_func/mean": -0.8421875, "rewards/format_reward_func/std": 0.37265667915344236, "step": 45 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1171875, "completions/max_length": 3072.0, "completions/max_terminated_length": 2962.0, "completions/mean_length": 1772.4265625, "completions/mean_terminated_length": 1599.689599609375, "completions/min_length": 44.6, "completions/min_terminated_length": 44.6, "entropy": 0.3753761973232031, "epoch": 3.125, "frac_reward_zero_std": 0.025, "grad_norm": 0.05543806403875351, "kl": 0.00046447456552414226, "learning_rate": 5.1e-07, "loss": -0.0036, "num_tokens": 12576689.0, "reward": -0.06843750141561031, "reward_std": 0.28640744686126707, "rewards/accuracy_reward/mean": 0.0984375, "rewards/accuracy_reward/std": 0.2967461168766022, "rewards/format_reward_func/mean": -0.834375, "rewards/format_reward_func/std": 0.39162964224815366, "step": 50 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.125, "completions/max_length": 3072.0, "completions/max_terminated_length": 3031.4, "completions/mean_length": 1806.475, "completions/mean_terminated_length": 1626.0361328125, "completions/min_length": 88.6, "completions/min_terminated_length": 88.6, "entropy": 0.38518749102950095, "epoch": 3.4375, "frac_reward_zero_std": 0.0, "grad_norm": 0.05889149382710457, "kl": 0.0005611305045022163, "learning_rate": 4.6e-07, "loss": 0.0111, "num_tokens": 13832033.0, "reward": -0.07750000357627869, "reward_std": 0.2828702092170715, "rewards/accuracy_reward/mean": 0.0921875, "rewards/accuracy_reward/std": 0.2897244989871979, "rewards/format_reward_func/mean": -0.8484375, "rewards/format_reward_func/std": 0.38314343690872193, "step": 55 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.14375, "completions/max_length": 3072.0, "completions/max_terminated_length": 2984.2, "completions/mean_length": 1825.7015625, "completions/mean_terminated_length": 1615.96591796875, "completions/min_length": 113.8, "completions/min_terminated_length": 113.8, "entropy": 0.3272393707185984, "epoch": 3.75, "frac_reward_zero_std": 0.0, "grad_norm": 0.06472011655569077, "kl": 0.000605658099084394, "learning_rate": 4.0999999999999994e-07, "loss": -0.0019, "num_tokens": 15099682.0, "reward": -0.037812499329447746, "reward_std": 0.3362713873386383, "rewards/accuracy_reward/mean": 0.128125, "rewards/accuracy_reward/std": 0.3307883620262146, "rewards/format_reward_func/mean": -0.8296875, "rewards/format_reward_func/std": 0.4038907289505005, "step": 60 }, { "epoch": 3.75, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.0640625, "eval_completions/max_length": 3069.35, "eval_completions/max_terminated_length": 2524.5, "eval_completions/mean_length": 993.6, "eval_completions/mean_terminated_length": 851.8893005371094, "eval_completions/min_length": 76.55, "eval_completions/min_terminated_length": 76.55, "eval_entropy": 0.9051018625497818, "eval_frac_reward_zero_std": 0.0, "eval_kl": 0.0006124728533905, "eval_loss": 0.02540457993745804, "eval_num_tokens": 15099682.0, "eval_reward": 0.037500002235174176, "eval_reward_std": 0.3434606835246086, "eval_rewards/accuracy_reward/mean": 0.190625, "eval_rewards/accuracy_reward/std": 0.3693762145936489, "eval_rewards/format_reward_func/mean": -0.765625, "eval_rewards/format_reward_func/std": 0.45619865879416466, "eval_runtime": 325.3378, "eval_samples_per_second": 0.123, "eval_steps_per_second": 0.006, "step": 60 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.09375, "completions/max_length": 3072.0, "completions/max_terminated_length": 2981.2, "completions/mean_length": 1777.940625, "completions/mean_terminated_length": 1645.0806640625, "completions/min_length": 76.2, "completions/min_terminated_length": 76.2, "entropy": 0.3321885637938976, "epoch": 4.0625, "frac_reward_zero_std": 0.0, "grad_norm": 0.05014500394463539, "kl": 0.0006261980546696577, "learning_rate": 3.6e-07, "loss": 0.0128, "num_tokens": 16336764.0, "reward": -0.034062501043081284, "reward_std": 0.3477294445037842, "rewards/accuracy_reward/mean": 0.1296875, "rewards/accuracy_reward/std": 0.3338283896446228, "rewards/format_reward_func/mean": -0.81875, "rewards/format_reward_func/std": 0.4009219169616699, "step": 65 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.134375, "completions/max_length": 3072.0, "completions/max_terminated_length": 2972.4, "completions/mean_length": 1835.6953125, "completions/mean_terminated_length": 1644.408154296875, "completions/min_length": 166.6, "completions/min_terminated_length": 166.6, "entropy": 0.3244955573230982, "epoch": 4.375, "frac_reward_zero_std": 0.05, "grad_norm": 0.05708487331867218, "kl": 0.0006808688696764876, "learning_rate": 3.1e-07, "loss": 0.0185, "num_tokens": 17610809.0, "reward": -0.05187500044703484, "reward_std": 0.32314308881759646, "rewards/accuracy_reward/mean": 0.1125, "rewards/accuracy_reward/std": 0.31601338386535643, "rewards/format_reward_func/mean": -0.821875, "rewards/format_reward_func/std": 0.4241487681865692, "step": 70 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.15, "completions/max_length": 3072.0, "completions/max_terminated_length": 2985.2, "completions/mean_length": 1845.7359375, "completions/mean_terminated_length": 1628.3058837890626, "completions/min_length": 66.0, "completions/min_terminated_length": 66.0, "entropy": 0.3411821313202381, "epoch": 4.6875, "frac_reward_zero_std": 0.0, "grad_norm": 0.04883616045117378, "kl": 0.0007200762047432363, "learning_rate": 2.6e-07, "loss": 0.0252, "num_tokens": 18891280.0, "reward": -0.04375000130385161, "reward_std": 0.32528549432754517, "rewards/accuracy_reward/mean": 0.121875, "rewards/accuracy_reward/std": 0.3260016322135925, "rewards/format_reward_func/mean": -0.828125, "rewards/format_reward_func/std": 0.3894888758659363, "step": 75 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1265625, "completions/max_length": 3072.0, "completions/max_terminated_length": 3035.0, "completions/mean_length": 1792.1734375, "completions/mean_terminated_length": 1607.0977783203125, "completions/min_length": 102.2, "completions/min_terminated_length": 102.2, "entropy": 0.26603236347436904, "epoch": 5.0, "frac_reward_zero_std": 0.0, "grad_norm": 0.056599684059619904, "kl": 0.0007625168931554071, "learning_rate": 2.0999999999999997e-07, "loss": 0.0199, "num_tokens": 20137471.0, "reward": -0.019687500037252902, "reward_std": 0.36239105463027954, "rewards/accuracy_reward/mean": 0.1390625, "rewards/accuracy_reward/std": 0.3465570151805878, "rewards/format_reward_func/mean": -0.79375, "rewards/format_reward_func/std": 0.42610687017440796, "step": 80 }, { "epoch": 5.0, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.0875, "eval_completions/max_length": 3063.8, "eval_completions/max_terminated_length": 2673.9, "eval_completions/mean_length": 1059.50625, "eval_completions/mean_terminated_length": 869.2105285644532, "eval_completions/min_length": 77.95, "eval_completions/min_terminated_length": 77.95, "eval_entropy": 0.9029649257659912, "eval_frac_reward_zero_std": 0.0, "eval_kl": 0.000769376612151973, "eval_loss": 0.000913003459572792, "eval_num_tokens": 20137471.0, "eval_reward": 0.021875000186264516, "eval_reward_std": 0.332692901045084, "eval_rewards/accuracy_reward/mean": 0.1765625, "eval_rewards/accuracy_reward/std": 0.366866135597229, "eval_rewards/format_reward_func/mean": -0.7734375, "eval_rewards/format_reward_func/std": 0.4547915384173393, "eval_runtime": 323.7653, "eval_samples_per_second": 0.124, "eval_steps_per_second": 0.006, "step": 80 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.134375, "completions/max_length": 3072.0, "completions/max_terminated_length": 3008.4, "completions/mean_length": 1843.90625, "completions/mean_terminated_length": 1653.5919677734375, "completions/min_length": 92.8, "completions/min_terminated_length": 92.8, "entropy": 0.31505762599408627, "epoch": 5.3125, "frac_reward_zero_std": 0.0, "grad_norm": 0.05114664509892464, "kl": 0.0007654805594938807, "learning_rate": 1.6e-07, "loss": 0.026, "num_tokens": 21416771.0, "reward": -0.048125001788139346, "reward_std": 0.3248369038105011, "rewards/accuracy_reward/mean": 0.1171875, "rewards/accuracy_reward/std": 0.32169962525367735, "rewards/format_reward_func/mean": -0.8265625, "rewards/format_reward_func/std": 0.3790922164916992, "step": 85 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.125, "completions/max_length": 3072.0, "completions/max_terminated_length": 2969.0, "completions/mean_length": 1854.9390625, "completions/mean_terminated_length": 1679.50869140625, "completions/min_length": 292.8, "completions/min_terminated_length": 292.8, "entropy": 0.2736196957528591, "epoch": 5.625, "frac_reward_zero_std": 0.0, "grad_norm": 0.05169934779405594, "kl": 0.0007617359107825906, "learning_rate": 1.0999999999999999e-07, "loss": 0.0116, "num_tokens": 22703132.0, "reward": -0.02031250111758709, "reward_std": 0.34891016483306886, "rewards/accuracy_reward/mean": 0.1421875, "rewards/accuracy_reward/std": 0.34694244861602785, "rewards/format_reward_func/mean": -0.8125, "rewards/format_reward_func/std": 0.4080379784107208, "step": 90 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.13125, "completions/max_length": 3072.0, "completions/max_terminated_length": 2976.8, "completions/mean_length": 1831.88125, "completions/mean_terminated_length": 1644.7082763671874, "completions/min_length": 72.4, "completions/min_terminated_length": 72.4, "entropy": 0.2820565219968557, "epoch": 5.9375, "frac_reward_zero_std": 0.0, "grad_norm": 0.05695090442895889, "kl": 0.0007496081641875207, "learning_rate": 6e-08, "loss": 0.0125, "num_tokens": 23974736.0, "reward": -0.023749999515712262, "reward_std": 0.3324207901954651, "rewards/accuracy_reward/mean": 0.1359375, "rewards/accuracy_reward/std": 0.3433120846748352, "rewards/format_reward_func/mean": -0.7984375, "rewards/format_reward_func/std": 0.43045543432235717, "step": 95 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.1453125, "completions/max_length": 3072.0, "completions/max_terminated_length": 3036.0, "completions/mean_length": 1836.90625, "completions/mean_terminated_length": 1627.1069091796876, "completions/min_length": 100.2, "completions/min_terminated_length": 100.2, "entropy": 0.29408896565437315, "epoch": 6.25, "frac_reward_zero_std": 0.0, "grad_norm": 0.06173178181052208, "kl": 0.0007958828617120161, "learning_rate": 1e-08, "loss": 0.0125, "num_tokens": 25249556.0, "reward": -0.021875, "reward_std": 0.35396260023117065, "rewards/accuracy_reward/mean": 0.1375, "rewards/accuracy_reward/std": 0.34515348076820374, "rewards/format_reward_func/mean": -0.796875, "rewards/format_reward_func/std": 0.41440053582191466, "step": 100 }, { "epoch": 6.25, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.1015625, "eval_completions/max_length": 3072.0, "eval_completions/max_terminated_length": 2494.6, "eval_completions/mean_length": 1018.278125, "eval_completions/mean_terminated_length": 788.2079956054688, "eval_completions/min_length": 42.85, "eval_completions/min_terminated_length": 42.85, "eval_entropy": 0.8883992671966553, "eval_frac_reward_zero_std": 0.0, "eval_kl": 0.0008909575059078634, "eval_loss": -0.0005189818912185729, "eval_num_tokens": 25249556.0, "eval_reward": 0.009375001490116119, "eval_reward_std": 0.32627131789922714, "eval_rewards/accuracy_reward/mean": 0.165625, "eval_rewards/accuracy_reward/std": 0.3576427064836025, "eval_rewards/format_reward_func/mean": -0.78125, "eval_rewards/format_reward_func/std": 0.44526770785450936, "eval_runtime": 326.8794, "eval_samples_per_second": 0.122, "eval_steps_per_second": 0.006, "step": 100 }, { "epoch": 6.25, "step": 100, "total_flos": 0.0, "train_loss": 0.010758781787008048, "train_runtime": 4656.4009, "train_samples_per_second": 2.749, "train_steps_per_second": 0.021 } ], "logging_steps": 5, "max_steps": 100, "num_input_tokens_seen": 25249556, "num_train_epochs": 7, "save_steps": 25, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }