{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.125, "eval_steps": 500, "global_step": 50, "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": 0.0, "completions/max_length": 1572.0, "completions/max_terminated_length": 1572.0, "completions/mean_length": 380.703125, "completions/mean_terminated_length": 380.703125, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0025, "frac_reward_zero_std": 0.125, "grad_norm": 10.104227066040039, "learning_rate": 0.0, "loss": -0.0, "num_tokens": 94445.0, "reward": 0.30937498807907104, "reward_std": 0.22894421219825745, "rewards/format_reward/mean": 0.28125, "rewards/format_reward/std": 0.3069951832294464, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.0, "completions/max_length": 1411.0, "completions/max_terminated_length": 1411.0, "completions/mean_length": 386.203125, "completions/mean_terminated_length": 386.203125, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.005, "frac_reward_zero_std": 0.125, "grad_norm": 23.277118682861328, "learning_rate": 2e-07, "loss": 0.0, "num_tokens": 233178.0, "reward": 0.35390621423721313, "reward_std": 0.3374568819999695, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.19012710452079773, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 2 }, { "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1254.0, "completions/mean_length": 383.484375, "completions/mean_terminated_length": 357.0635070800781, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0075, "frac_reward_zero_std": 0.125, "grad_norm": 5.73276424407959, "learning_rate": 4e-07, "loss": -0.0, "num_tokens": 327625.0, "reward": 0.17109373211860657, "reward_std": 0.28433501720428467, "rewards/format_reward/mean": 0.3046875, "rewards/format_reward/std": 0.3036443889141083, "rewards/mcq_exact_match_reward/mean": 0.140625, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "step": 3 }, { "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": 0.0, "completions/max_length": 1591.0, "completions/max_terminated_length": 1591.0, "completions/mean_length": 422.296875, "completions/mean_terminated_length": 422.296875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.01, "frac_reward_zero_std": 0.0, "grad_norm": 12.001652717590332, "learning_rate": 6e-07, "loss": -0.0, "num_tokens": 437396.0, "reward": 0.4289062023162842, "reward_std": 0.40633565187454224, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.2478829026222229, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "step": 4 }, { "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": 0.0, "completions/max_length": 1573.0, "completions/max_terminated_length": 1573.0, "completions/mean_length": 411.375, "completions/mean_terminated_length": 411.375, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0125, "frac_reward_zero_std": 0.0, "grad_norm": 34.90378952026367, "learning_rate": 8e-07, "loss": 0.0, "num_tokens": 559724.0, "reward": 0.2914062440395355, "reward_std": 0.32816699147224426, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.28942593932151794, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1824.0, "completions/mean_length": 664.4375, "completions/mean_terminated_length": 619.8064575195312, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.015, "frac_reward_zero_std": 0.0, "grad_norm": 5.146732807159424, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 695512.0, "reward": 0.23984374105930328, "reward_std": 0.25555554032325745, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.23974503576755524, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "step": 6 }, { "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": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1313.0, "completions/mean_length": 495.28125, "completions/mean_terminated_length": 391.7666931152344, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 13.87881088256836, "learning_rate": 9.99726628670463e-07, "loss": -0.0, "num_tokens": 839354.0, "reward": 0.3804687261581421, "reward_std": 0.4405275881290436, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.255761981010437, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 7 }, { "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": 0.0, "completions/max_length": 1038.0, "completions/max_terminated_length": 1038.0, "completions/mean_length": 283.828125, "completions/mean_terminated_length": 283.828125, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 16.57680320739746, "learning_rate": 9.989068136093872e-07, "loss": -0.0, "num_tokens": 941223.0, "reward": 0.2601562440395355, "reward_std": 0.3872828483581543, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.27537402510643005, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "step": 8 }, { "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1446.0, "completions/mean_length": 398.59375, "completions/mean_terminated_length": 345.3870849609375, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0225, "frac_reward_zero_std": 0.0, "grad_norm": 27.8460693359375, "learning_rate": 9.975414512725056e-07, "loss": 0.0, "num_tokens": 1046957.0, "reward": 0.3374999761581421, "reward_std": 0.31849682331085205, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.29378482699394226, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "step": 9 }, { "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1777.0, "completions/mean_length": 362.65625, "completions/mean_terminated_length": 335.90478515625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.025, "frac_reward_zero_std": 0.0, "grad_norm": 9.661582946777344, "learning_rate": 9.956320346634875e-07, "loss": -0.0, "num_tokens": 1172647.0, "reward": 0.3343749940395355, "reward_std": 0.3494775593280792, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.26726123690605164, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 834.0, "completions/mean_length": 373.828125, "completions/mean_terminated_length": 319.82257080078125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0275, "frac_reward_zero_std": 0.125, "grad_norm": 11.05769157409668, "learning_rate": 9.931806517013612e-07, "loss": 0.0, "num_tokens": 1282204.0, "reward": 0.15625, "reward_std": 0.2576306164264679, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.279951810836792, "rewards/mcq_exact_match_reward/mean": 0.109375, "rewards/mcq_exact_match_reward/std": 0.3145764470100403, "step": 11 }, { "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": 0.0, "completions/max_length": 1726.0, "completions/max_terminated_length": 1726.0, "completions/mean_length": 401.875, "completions/mean_terminated_length": 401.875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 6.454217910766602, "learning_rate": 9.901899829374047e-07, "loss": -0.0, "num_tokens": 1398452.0, "reward": 0.3242187201976776, "reward_std": 0.4082157611846924, "rewards/format_reward/mean": 0.4296875, "rewards/format_reward/std": 0.23345555365085602, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "step": 12 }, { "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": 0.0, "completions/max_length": 925.0, "completions/max_terminated_length": 925.0, "completions/mean_length": 246.75, "completions/mean_terminated_length": 246.75, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0325, "frac_reward_zero_std": 0.25, "grad_norm": 6.135255813598633, "learning_rate": 9.866632986240029e-07, "loss": -0.0, "num_tokens": 1474604.0, "reward": 0.4921874403953552, "reward_std": 0.30105772614479065, "rewards/format_reward/mean": 0.546875, "rewards/format_reward/std": 0.2777281701564789, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "step": 13 }, { "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1591.0, "completions/mean_length": 239.3125, "completions/mean_terminated_length": 210.60317993164062, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.035, "frac_reward_zero_std": 0.375, "grad_norm": 13.646222114562988, "learning_rate": 9.826044551386742e-07, "loss": 0.0, "num_tokens": 1565424.0, "reward": 0.15312500298023224, "reward_std": 0.20769444108009338, "rewards/format_reward/mean": 0.59375, "rewards/format_reward/std": 0.279951810836792, "rewards/mcq_exact_match_reward/mean": 0.09375, "rewards/mcq_exact_match_reward/std": 0.29378482699394226, "step": 14 }, { "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1182.0, "completions/mean_length": 238.71875, "completions/mean_terminated_length": 210.00001525878906, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.375, "grad_norm": 10.635026931762695, "learning_rate": 9.780178907671788e-07, "loss": -0.0, "num_tokens": 1671630.0, "reward": 0.5648436546325684, "reward_std": 0.23381614685058594, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.16648055613040924, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 0.0, "completions/max_length": 875.0, "completions/max_terminated_length": 875.0, "completions/mean_length": 64.015625, "completions/mean_terminated_length": 64.015625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04, "frac_reward_zero_std": 0.125, "grad_norm": 17.595975875854492, "learning_rate": 9.729086208503173e-07, "loss": -0.0, "num_tokens": 1763119.0, "reward": 0.547656238079071, "reward_std": 0.3905794620513916, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.10652101784944534, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 16 }, { "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": 0.0, "completions/max_length": 475.0, "completions/max_terminated_length": 475.0, "completions/mean_length": 43.203125, "completions/mean_terminated_length": 43.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0425, "frac_reward_zero_std": 0.5, "grad_norm": 13.388673782348633, "learning_rate": 9.672822322997304e-07, "loss": 0.0, "num_tokens": 1849220.0, "reward": 0.5968749523162842, "reward_std": 0.1720314621925354, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.1259881556034088, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 17 }, { "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": 0.0, "completions/max_length": 6.0, "completions/max_terminated_length": 6.0, "completions/mean_length": 6.0, "completions/mean_terminated_length": 6.0, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.045, "frac_reward_zero_std": 0.625, "grad_norm": 16.634624481201172, "learning_rate": 9.611448774886923e-07, "loss": 0.0, "num_tokens": 1954380.0, "reward": 0.4718749523162842, "reward_std": 0.1530819833278656, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 18 }, { "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": 0.0, "completions/max_length": 106.0, "completions/max_terminated_length": 106.0, "completions/mean_length": 7.5625, "completions/mean_terminated_length": 7.5625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0475, "frac_reward_zero_std": 0.75, "grad_norm": 11.946528434753418, "learning_rate": 9.545032675245813e-07, "loss": -0.0, "num_tokens": 2065680.0, "reward": 0.4562499523162842, "reward_std": 0.10888782143592834, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 19 }, { "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 6.609375, "completions/mean_terminated_length": 6.609375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.5, "grad_norm": 30.572359085083008, "learning_rate": 9.473646649103817e-07, "loss": 0.0, "num_tokens": 2144071.0, "reward": 0.4921874701976776, "reward_std": 0.16964474320411682, "rewards/format_reward/mean": 0.546875, "rewards/format_reward/std": 0.14689241349697113, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 0.0, "completions/max_length": 289.0, "completions/max_terminated_length": 289.0, "completions/mean_length": 14.015625, "completions/mean_terminated_length": 14.015625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0525, "frac_reward_zero_std": 0.5, "grad_norm": 20.962142944335938, "learning_rate": 9.397368756032444e-07, "loss": 0.0, "num_tokens": 2218648.0, "reward": 0.34062498807907104, "reward_std": 0.12822021543979645, "rewards/format_reward/mean": 0.59375, "rewards/format_reward/std": 0.19669894874095917, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "step": 21 }, { "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": 0.0, "completions/max_length": 1013.0, "completions/max_terminated_length": 1013.0, "completions/mean_length": 22.03125, "completions/mean_terminated_length": 22.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.625, "grad_norm": 28.532514572143555, "learning_rate": 9.316282404787869e-07, "loss": -0.0, "num_tokens": 2301802.0, "reward": 0.7242187261581421, "reward_std": 0.12579208612442017, "rewards/format_reward/mean": 0.5234375, "rewards/format_reward/std": 0.10652101784944534, "rewards/mcq_exact_match_reward/mean": 0.671875, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "step": 22 }, { "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": 0.0, "completions/max_length": 106.0, "completions/max_terminated_length": 106.0, "completions/mean_length": 7.5625, "completions/mean_terminated_length": 7.5625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0575, "frac_reward_zero_std": 0.625, "grad_norm": 23.159854888916016, "learning_rate": 9.230476262104676e-07, "loss": -0.0, "num_tokens": 2392438.0, "reward": 0.6124999523162842, "reward_std": 0.15520364046096802, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 23 }, { "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": 0.0, "completions/max_length": 779.0, "completions/max_terminated_length": 779.0, "completions/mean_length": 19.4375, "completions/mean_terminated_length": 19.4375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.625, "grad_norm": 14.921491622924805, "learning_rate": 9.1400441557401e-07, "loss": 0.0, "num_tokens": 2473554.0, "reward": 0.5914062261581421, "reward_std": 0.17606157064437866, "rewards/format_reward/mean": 0.6015625, "rewards/format_reward/std": 0.20275264978408813, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "step": 24 }, { "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": 0.0, "completions/max_length": 460.0, "completions/max_terminated_length": 460.0, "completions/mean_length": 14.15625, "completions/mean_terminated_length": 14.15625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.375, "grad_norm": 31.79290771484375, "learning_rate": 9.045084971874737e-07, "loss": -0.0, "num_tokens": 2573596.0, "reward": 0.26953125, "reward_std": 0.20632556080818176, "rewards/format_reward/mean": 0.5078125, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.0, "completions/max_length": 495.0, "completions/max_terminated_length": 495.0, "completions/mean_length": 16.96875, "completions/mean_terminated_length": 16.96875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.065, "frac_reward_zero_std": 0.25, "grad_norm": 23.660274505615234, "learning_rate": 8.945702546981968e-07, "loss": -0.0, "num_tokens": 2642178.0, "reward": 0.5679687261581421, "reward_std": 0.3241649270057678, "rewards/format_reward/mean": 0.6796875, "rewards/format_reward/std": 0.24180518090724945, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 26 }, { "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": 0.0, "completions/max_length": 170.0, "completions/max_terminated_length": 170.0, "completions/mean_length": 10.84375, "completions/mean_terminated_length": 10.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0675, "frac_reward_zero_std": 0.5, "grad_norm": 21.55986976623535, "learning_rate": 8.842005554284295e-07, "loss": -0.0, "num_tokens": 2758224.0, "reward": 0.5343749523162842, "reward_std": 0.2109457403421402, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 27 }, { "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": 0.0, "completions/max_length": 14.0, "completions/max_terminated_length": 14.0, "completions/mean_length": 7.140625, "completions/mean_terminated_length": 7.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07, "frac_reward_zero_std": 0.5, "grad_norm": 23.411895751953125, "learning_rate": 8.734107384920769e-07, "loss": 0.0, "num_tokens": 2833361.0, "reward": 0.63671875, "reward_std": 0.19149912893772125, "rewards/format_reward/mean": 0.5859375, "rewards/format_reward/std": 0.19012710452079773, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 28 }, { "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": 0.0, "completions/max_length": 141.0, "completions/max_terminated_length": 141.0, "completions/mean_length": 8.859375, "completions/mean_terminated_length": 8.859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0725, "frac_reward_zero_std": 0.875, "grad_norm": 4.729227542877197, "learning_rate": 8.622126023955445e-07, "loss": 0.0, "num_tokens": 2922176.0, "reward": 0.5406249761581421, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 29 }, { "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 8.28125, "completions/mean_terminated_length": 8.28125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.075, "frac_reward_zero_std": 0.75, "grad_norm": 11.8333740234375, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 2992354.0, "reward": 0.5687500238418579, "reward_std": 0.1293872892856598, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.24397502839565277, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "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": 0.0, "completions/max_length": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 8.25, "completions/mean_terminated_length": 8.25, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.625, "grad_norm": 14.442317962646484, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 3080066.0, "reward": 0.3500000238418579, "reward_std": 0.1462520956993103, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.24397502839565277, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "step": 31 }, { "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 8.265625, "completions/mean_terminated_length": 8.265625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08, "frac_reward_zero_std": 0.375, "grad_norm": 29.77044105529785, "learning_rate": 8.262928807620843e-07, "loss": -0.0, "num_tokens": 3154547.0, "reward": 0.5218750238418579, "reward_std": 0.25726157426834106, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.24397502839565277, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 32 }, { "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": 0.0, "completions/max_length": 81.0, "completions/max_terminated_length": 81.0, "completions/mean_length": 10.203125, "completions/mean_terminated_length": 10.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.5, "grad_norm": 15.08368968963623, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 3250384.0, "reward": 0.23515623807907104, "reward_std": 0.13479220867156982, "rewards/format_reward/mean": 0.6328125, "rewards/format_reward/std": 0.22257846593856812, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "step": 33 }, { "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": 0.0, "completions/max_length": 493.0, "completions/max_terminated_length": 493.0, "completions/mean_length": 15.1875, "completions/mean_terminated_length": 15.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.085, "frac_reward_zero_std": 0.625, "grad_norm": 18.34843635559082, "learning_rate": 8.005405736415125e-07, "loss": -0.0, "num_tokens": 3332732.0, "reward": 0.624218761920929, "reward_std": 0.1555730253458023, "rewards/format_reward/mean": 0.6171875, "rewards/format_reward/std": 0.2313210517168045, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "step": 34 }, { "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": 0.0, "completions/max_length": 152.0, "completions/max_terminated_length": 152.0, "completions/mean_length": 10.4375, "completions/mean_terminated_length": 10.4375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0875, "frac_reward_zero_std": 0.875, "grad_norm": 7.837115287780762, "learning_rate": 7.871643313414718e-07, "loss": 0.0, "num_tokens": 3417584.0, "reward": 0.7875000238418579, "reward_std": 0.0578637570142746, "rewards/format_reward/mean": 0.6875, "rewards/format_reward/std": 0.24397502839565277, "rewards/mcq_exact_match_reward/mean": 0.71875, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.0, "completions/max_length": 198.0, "completions/max_terminated_length": 198.0, "completions/mean_length": 18.53125, "completions/mean_terminated_length": 18.53125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09, "frac_reward_zero_std": 0.375, "grad_norm": 22.8250732421875, "learning_rate": 7.734740790612136e-07, "loss": -0.0, "num_tokens": 3503610.0, "reward": 0.484375, "reward_std": 0.30563658475875854, "rewards/format_reward/mean": 0.78125, "rewards/format_reward/std": 0.25, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 36 }, { "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 10.515625, "completions/mean_terminated_length": 10.515625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0925, "frac_reward_zero_std": 0.625, "grad_norm": 10.62349796295166, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 3555275.0, "reward": 0.4937500059604645, "reward_std": 0.1462520956993103, "rewards/format_reward/mean": 0.875, "rewards/format_reward/std": 0.2182178944349289, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 37 }, { "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": 0.0, "completions/max_length": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 7.59375, "completions/mean_terminated_length": 7.59375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.625, "grad_norm": 19.927433013916016, "learning_rate": 7.452117519152541e-07, "loss": 0.0, "num_tokens": 3639065.0, "reward": 0.547656238079071, "reward_std": 0.13371865451335907, "rewards/format_reward/mean": 0.6328125, "rewards/format_reward/std": 0.22257846593856812, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 38 }, { "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": 0.0, "completions/max_length": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 9.078125, "completions/mean_terminated_length": 9.078125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0975, "frac_reward_zero_std": 0.5, "grad_norm": 76.84066009521484, "learning_rate": 7.306705814893439e-07, "loss": 0.0, "num_tokens": 3728342.0, "reward": 0.41796875, "reward_std": 0.11669307202100754, "rewards/format_reward/mean": 0.7421875, "rewards/format_reward/std": 0.25185325741767883, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 39 }, { "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": 0.0, "completions/max_length": 76.0, "completions/max_terminated_length": 76.0, "completions/mean_length": 11.578125, "completions/mean_terminated_length": 11.578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1, "frac_reward_zero_std": 0.5, "grad_norm": 60.49788284301758, "learning_rate": 7.158771761692464e-07, "loss": 0.0, "num_tokens": 3825723.0, "reward": 0.5617187023162842, "reward_std": 0.16234643757343292, "rewards/format_reward/mean": 0.7734375, "rewards/format_reward/std": 0.250866562128067, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 0.0, "completions/max_length": 407.0, "completions/max_terminated_length": 407.0, "completions/mean_length": 18.890625, "completions/mean_terminated_length": 18.890625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1025, "frac_reward_zero_std": 0.5, "grad_norm": 26.07494354248047, "learning_rate": 7.008477123264847e-07, "loss": -0.0, "num_tokens": 3907724.0, "reward": 0.35859376192092896, "reward_std": 0.13479222357273102, "rewards/format_reward/mean": 0.9296875, "rewards/format_reward/std": 0.1751912236213684, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "step": 41 }, { "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": 0.0, "completions/max_length": 182.0, "completions/max_terminated_length": 182.0, "completions/mean_length": 13.84375, "completions/mean_terminated_length": 13.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.105, "frac_reward_zero_std": 0.5, "grad_norm": 57.546485900878906, "learning_rate": 6.855986244591103e-07, "loss": -0.0, "num_tokens": 4013906.0, "reward": 0.4281249940395355, "reward_std": 0.06670258939266205, "rewards/format_reward/mean": 0.84375, "rewards/format_reward/std": 0.25, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 42 }, { "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": 0.0, "completions/max_length": 63.0, "completions/max_terminated_length": 63.0, "completions/mean_length": 11.46875, "completions/mean_terminated_length": 11.46875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1075, "frac_reward_zero_std": 0.375, "grad_norm": 23.997573852539062, "learning_rate": 6.701465872208216e-07, "loss": 0.0, "num_tokens": 4101112.0, "reward": 0.9273437261581421, "reward_std": 0.23680157959461212, "rewards/format_reward/mean": 0.8359375, "rewards/format_reward/std": 0.2366211861371994, "rewards/mcq_exact_match_reward/mean": 0.84375, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "step": 43 }, { "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": 0.0, "completions/max_length": 150.0, "completions/max_terminated_length": 150.0, "completions/mean_length": 14.234375, "completions/mean_terminated_length": 14.234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.11, "frac_reward_zero_std": 0.5, "grad_norm": 46.40782165527344, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 4193311.0, "reward": 0.7109375, "reward_std": 0.18755537271499634, "rewards/format_reward/mean": 0.859375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 44 }, { "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": 0.0, "completions/max_length": 262.0, "completions/max_terminated_length": 262.0, "completions/mean_length": 15.765625, "completions/mean_terminated_length": 15.765625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1125, "frac_reward_zero_std": 0.375, "grad_norm": 21.46227264404297, "learning_rate": 6.387014543809223e-07, "loss": -0.0, "num_tokens": 4272336.0, "reward": 0.6179687976837158, "reward_std": 0.28839826583862305, "rewards/format_reward/mean": 0.8671875, "rewards/format_reward/std": 0.23974503576755524, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 0.0, "completions/max_length": 76.0, "completions/max_terminated_length": 76.0, "completions/mean_length": 13.484375, "completions/mean_terminated_length": 13.484375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.115, "frac_reward_zero_std": 0.875, "grad_norm": 21.336334228515625, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 4352943.0, "reward": 0.9437500238418579, "reward_std": 0.0578637570142746, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.84375, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "step": 46 }, { "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": 0.0, "completions/max_length": 49.0, "completions/max_terminated_length": 49.0, "completions/mean_length": 13.296875, "completions/mean_terminated_length": 13.296875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1175, "frac_reward_zero_std": 0.625, "grad_norm": 18.461605072021484, "learning_rate": 6.066498153718734e-07, "loss": 0.0, "num_tokens": 4449594.0, "reward": 0.3656250238418579, "reward_std": 0.17358146607875824, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "step": 47 }, { "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": 0.0, "completions/max_length": 364.0, "completions/max_terminated_length": 364.0, "completions/mean_length": 23.125, "completions/mean_terminated_length": 23.125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.12, "frac_reward_zero_std": 0.625, "grad_norm": 21.556089401245117, "learning_rate": 5.90440267166055e-07, "loss": 0.0, "num_tokens": 4529738.0, "reward": 0.5843750238418579, "reward_std": 0.17358146607875824, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 48 }, { "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": 0.0, "completions/max_length": 71.0, "completions/max_terminated_length": 71.0, "completions/mean_length": 13.3125, "completions/mean_terminated_length": 13.3125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1225, "frac_reward_zero_std": 0.5, "grad_norm": 27.238489151000977, "learning_rate": 5.741318238559209e-07, "loss": 0.0, "num_tokens": 4629662.0, "reward": 0.7250000238418579, "reward_std": 0.2177756428718567, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "step": 49 }, { "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": 0.0, "completions/max_length": 364.0, "completions/max_terminated_length": 364.0, "completions/mean_length": 22.796875, "completions/mean_terminated_length": 22.796875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.125, "frac_reward_zero_std": 0.75, "grad_norm": 23.85200309753418, "learning_rate": 5.577423184847931e-07, "loss": -0.0, "num_tokens": 4705673.0, "reward": 0.6781250238418579, "reward_std": 0.11100947856903076, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 50 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 4705673, "num_train_epochs": 1, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 1, "trial_name": null, "trial_params": null }