{ "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": 1697.0, "completions/max_terminated_length": 1697.0, "completions/mean_length": 556.375, "completions/mean_terminated_length": 556.375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0025, "frac_reward_zero_std": 0.0, "grad_norm": 8.208407402038574, "learning_rate": 0.0, "loss": -0.0, "num_tokens": 149640.0, "reward": 0.34375, "reward_std": 0.37133273482322693, "rewards/format_reward/mean": 0.3125, "rewards/format_reward/std": 0.2597312331199646, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1011.0, "completions/mean_length": 358.65625, "completions/mean_terminated_length": 331.8412780761719, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.005, "frac_reward_zero_std": 0.0, "grad_norm": 11.007621765136719, "learning_rate": 2e-07, "loss": -0.0, "num_tokens": 241938.0, "reward": 0.36562496423721313, "reward_std": 0.43796563148498535, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.26726123690605164, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 1215.0, "completions/mean_length": 468.828125, "completions/mean_terminated_length": 443.7619323730469, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0075, "frac_reward_zero_std": 0.125, "grad_norm": 4.224123477935791, "learning_rate": 4e-07, "loss": -0.0, "num_tokens": 374575.0, "reward": 0.3664062023162842, "reward_std": 0.37426015734672546, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.2634054720401764, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1154.0, "completions/mean_length": 509.296875, "completions/mean_terminated_length": 484.873046875, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.01, "frac_reward_zero_std": 0.0, "grad_norm": 16.33107566833496, "learning_rate": 6e-07, "loss": 0.0, "num_tokens": 508722.0, "reward": 0.29921871423721313, "reward_std": 0.4078366756439209, "rewards/format_reward/mean": 0.3359375, "rewards/format_reward/std": 0.25283610820770264, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1163.0, "completions/mean_length": 455.5, "completions/mean_terminated_length": 404.1290283203125, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0125, "frac_reward_zero_std": 0.0, "grad_norm": 7.4960174560546875, "learning_rate": 8e-07, "loss": 0.0, "num_tokens": 600122.0, "reward": 0.34999996423721313, "reward_std": 0.3820914626121521, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.26726123690605164, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1408.0, "completions/mean_length": 526.625, "completions/mean_terminated_length": 502.4762268066406, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.015, "frac_reward_zero_std": 0.0, "grad_norm": 11.325846672058105, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 710722.0, "reward": 0.22578124701976776, "reward_std": 0.2976260781288147, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.3052735924720764, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 0.0, "completions/max_length": 1434.0, "completions/max_terminated_length": 1434.0, "completions/mean_length": 547.0625, "completions/mean_terminated_length": 547.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 4.365859031677246, "learning_rate": 9.99726628670463e-07, "loss": -0.0, "num_tokens": 855958.0, "reward": 0.36249998211860657, "reward_std": 0.42349135875701904, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1187.0, "completions/mean_length": 491.390625, "completions/mean_terminated_length": 441.1773986816406, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 7.249939441680908, "learning_rate": 9.989068136093872e-07, "loss": 0.0, "num_tokens": 971183.0, "reward": 0.34062498807907104, "reward_std": 0.36264413595199585, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.2745848298072815, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 1752.0, "completions/mean_length": 604.5625, "completions/mean_terminated_length": 558.0, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0225, "frac_reward_zero_std": 0.125, "grad_norm": 5.596194744110107, "learning_rate": 9.975414512725056e-07, "loss": 0.0, "num_tokens": 1108787.0, "reward": 0.19843748211860657, "reward_std": 0.30780309438705444, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.2221602201461792, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1997.0, "completions/mean_length": 548.109375, "completions/mean_terminated_length": 499.7257995605469, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.025, "frac_reward_zero_std": 0.0, "grad_norm": 14.296536445617676, "learning_rate": 9.956320346634875e-07, "loss": 0.0, "num_tokens": 1234522.0, "reward": 0.2562499940395355, "reward_std": 0.33966386318206787, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.29546841979026794, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.0, "completions/max_length": 1391.0, "completions/max_terminated_length": 1391.0, "completions/mean_length": 372.140625, "completions/mean_terminated_length": 372.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0275, "frac_reward_zero_std": 0.125, "grad_norm": 8.358329772949219, "learning_rate": 9.931806517013612e-07, "loss": -0.0, "num_tokens": 1351707.0, "reward": 0.5249999761581421, "reward_std": 0.3847616910934448, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1601.0, "completions/mean_length": 597.78125, "completions/mean_terminated_length": 526.458984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 8.153278350830078, "learning_rate": 9.901899829374047e-07, "loss": -0.0, "num_tokens": 1490685.0, "reward": 0.37031248211860657, "reward_std": 0.39562246203422546, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.20351573824882507, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1131.0, "completions/mean_length": 467.21875, "completions/mean_terminated_length": 442.12701416015625, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0325, "frac_reward_zero_std": 0.0, "grad_norm": 13.680706977844238, "learning_rate": 9.866632986240029e-07, "loss": -0.0, "num_tokens": 1590243.0, "reward": 0.38593748211860657, "reward_std": 0.41698139905929565, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.29839184880256653, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.0, "completions/max_length": 1345.0, "completions/max_terminated_length": 1345.0, "completions/mean_length": 346.8125, "completions/mean_terminated_length": 346.8125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.035, "frac_reward_zero_std": 0.375, "grad_norm": 8.916102409362793, "learning_rate": 9.826044551386742e-07, "loss": 0.0, "num_tokens": 1709063.0, "reward": 0.33906248211860657, "reward_std": 0.20275810360908508, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.23935678601264954, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 2018.0, "completions/mean_length": 333.203125, "completions/mean_terminated_length": 305.984130859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.0, "grad_norm": 14.707695960998535, "learning_rate": 9.780178907671788e-07, "loss": 0.0, "num_tokens": 1820412.0, "reward": 0.2953124940395355, "reward_std": 0.41087859869003296, "rewards/format_reward/mean": 0.453125, "rewards/format_reward/std": 0.3177144229412079, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1974.0, "completions/mean_length": 609.546875, "completions/mean_terminated_length": 538.80322265625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04, "frac_reward_zero_std": 0.125, "grad_norm": 2.6404623985290527, "learning_rate": 9.729086208503173e-07, "loss": 0.0, "num_tokens": 1968927.0, "reward": 0.21171873807907104, "reward_std": 0.23303578794002533, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.26991602778434753, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 876.0, "completions/max_terminated_length": 876.0, "completions/mean_length": 302.203125, "completions/mean_terminated_length": 302.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0425, "frac_reward_zero_std": 0.0, "grad_norm": 23.550676345825195, "learning_rate": 9.672822322997304e-07, "loss": 0.0, "num_tokens": 2067324.0, "reward": 0.328125, "reward_std": 0.2806034982204437, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.29378482699394226, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 1015.0, "completions/max_terminated_length": 1015.0, "completions/mean_length": 194.734375, "completions/mean_terminated_length": 194.734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.045, "frac_reward_zero_std": 0.0, "grad_norm": 15.42598819732666, "learning_rate": 9.611448774886923e-07, "loss": -0.0, "num_tokens": 2174483.0, "reward": 0.31562498211860657, "reward_std": 0.39141321182250977, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.26726123690605164, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 1456.0, "completions/max_terminated_length": 1456.0, "completions/mean_length": 248.046875, "completions/mean_terminated_length": 248.046875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0475, "frac_reward_zero_std": 0.0, "grad_norm": 16.15092658996582, "learning_rate": 9.545032675245813e-07, "loss": 0.0, "num_tokens": 2273846.0, "reward": 0.47578123211860657, "reward_std": 0.351205050945282, "rewards/format_reward/mean": 0.5390625, "rewards/format_reward/std": 0.29953256249427795, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1513.0, "completions/max_terminated_length": 1513.0, "completions/mean_length": 193.1875, "completions/mean_terminated_length": 193.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.25, "grad_norm": 13.046883583068848, "learning_rate": 9.473646649103817e-07, "loss": 0.0, "num_tokens": 2385890.0, "reward": 0.2148437649011612, "reward_std": 0.18485845625400543, "rewards/format_reward/mean": 0.5859375, "rewards/format_reward/std": 0.2605654299259186, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 1252.0, "completions/max_terminated_length": 1252.0, "completions/mean_length": 292.875, "completions/mean_terminated_length": 292.875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0525, "frac_reward_zero_std": 0.375, "grad_norm": 5.589727878570557, "learning_rate": 9.397368756032444e-07, "loss": -0.0, "num_tokens": 2479914.0, "reward": 0.23281249403953552, "reward_std": 0.22750720381736755, "rewards/format_reward/mean": 0.609375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 877.0, "completions/mean_length": 120.984375, "completions/mean_terminated_length": 90.3968276977539, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.25, "grad_norm": 8.323296546936035, "learning_rate": 9.316282404787869e-07, "loss": 0.0, "num_tokens": 2564873.0, "reward": 0.538281261920929, "reward_std": 0.23045194149017334, "rewards/format_reward/mean": 0.6953125, "rewards/format_reward/std": 0.31644338369369507, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 1446.0, "completions/max_terminated_length": 1446.0, "completions/mean_length": 164.953125, "completions/mean_terminated_length": 164.953125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0575, "frac_reward_zero_std": 0.125, "grad_norm": 14.68455982208252, "learning_rate": 9.230476262104676e-07, "loss": 0.0, "num_tokens": 2654126.0, "reward": 0.5132812261581421, "reward_std": 0.35147207975387573, "rewards/format_reward/mean": 0.6015625, "rewards/format_reward/std": 0.4380665421485901, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 1430.0, "completions/max_terminated_length": 1430.0, "completions/mean_length": 150.203125, "completions/mean_terminated_length": 150.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.125, "grad_norm": 12.729601860046387, "learning_rate": 9.1400441557401e-07, "loss": 0.0, "num_tokens": 2754451.0, "reward": 0.21875, "reward_std": 0.24718233942985535, "rewards/format_reward/mean": 0.78125, "rewards/format_reward/std": 0.36596253514289856, "rewards/mcq_exact_match_reward/mean": 0.140625, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "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": 1101.0, "completions/max_terminated_length": 1101.0, "completions/mean_length": 135.21875, "completions/mean_terminated_length": 135.21875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.25, "grad_norm": 24.319026947021484, "learning_rate": 9.045084971874737e-07, "loss": 0.0, "num_tokens": 2853273.0, "reward": 0.41718751192092896, "reward_std": 0.3194974362850189, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.22658175230026245, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 843.0, "completions/max_terminated_length": 843.0, "completions/mean_length": 51.78125, "completions/mean_terminated_length": 51.78125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.065, "frac_reward_zero_std": 0.125, "grad_norm": 21.616012573242188, "learning_rate": 8.945702546981968e-07, "loss": 0.0, "num_tokens": 2944211.0, "reward": 0.5789062976837158, "reward_std": 0.32899072766304016, "rewards/format_reward/mean": 0.7890625, "rewards/format_reward/std": 0.2643453776836395, "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": 643.0, "completions/max_terminated_length": 643.0, "completions/mean_length": 28.3125, "completions/mean_terminated_length": 28.3125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0675, "frac_reward_zero_std": 0.5, "grad_norm": 41.8525390625, "learning_rate": 8.842005554284295e-07, "loss": -0.0, "num_tokens": 3057071.0, "reward": 0.23281250894069672, "reward_std": 0.09383273124694824, "rewards/format_reward/mean": 0.765625, "rewards/format_reward/std": 0.2514837086200714, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 1673.0, "completions/max_terminated_length": 1673.0, "completions/mean_length": 90.21875, "completions/mean_terminated_length": 90.21875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07, "frac_reward_zero_std": 0.25, "grad_norm": 31.983003616333008, "learning_rate": 8.734107384920769e-07, "loss": -0.0, "num_tokens": 3146917.0, "reward": 0.47187498211860657, "reward_std": 0.28054213523864746, "rewards/format_reward/mean": 0.8125, "rewards/format_reward/std": 0.2745848298072815, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 227.0, "completions/max_terminated_length": 227.0, "completions/mean_length": 20.828125, "completions/mean_terminated_length": 20.828125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0725, "frac_reward_zero_std": 0.5, "grad_norm": 39.55906295776367, "learning_rate": 8.622126023955445e-07, "loss": -0.0, "num_tokens": 3241050.0, "reward": 0.7132812738418579, "reward_std": 0.09349124133586884, "rewards/format_reward/mean": 0.8828125, "rewards/format_reward/std": 0.21347814798355103, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 208.0, "completions/max_terminated_length": 208.0, "completions/mean_length": 22.578125, "completions/mean_terminated_length": 22.578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.075, "frac_reward_zero_std": 0.375, "grad_norm": 28.997608184814453, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 3318663.0, "reward": 0.6125000715255737, "reward_std": 0.21486152708530426, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.12198751419782639, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 440.0, "completions/max_terminated_length": 440.0, "completions/mean_length": 30.140625, "completions/mean_terminated_length": 30.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.375, "grad_norm": 17.291109085083008, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 3409360.0, "reward": 0.24921876192092896, "reward_std": 0.20632557570934296, "rewards/format_reward/mean": 0.9296875, "rewards/format_reward/std": 0.1751912236213684, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.08, "frac_reward_zero_std": 0.75, "grad_norm": 4.5453386306762695, "learning_rate": 8.262928807620843e-07, "loss": 0.0, "num_tokens": 3503712.0, "reward": 0.5375000238418579, "reward_std": 0.1157275140285492, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 612.0, "completions/max_terminated_length": 612.0, "completions/mean_length": 30.453125, "completions/mean_terminated_length": 30.453125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.375, "grad_norm": 28.091222763061523, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 3601133.0, "reward": 0.4414062798023224, "reward_std": 0.20033013820648193, "rewards/format_reward/mean": 0.9765625, "rewards/format_reward/std": 0.13886408507823944, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 17.0, "completions/max_terminated_length": 17.0, "completions/mean_length": 12.078125, "completions/mean_terminated_length": 12.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.085, "frac_reward_zero_std": 0.625, "grad_norm": 9.407567977905273, "learning_rate": 8.005405736415125e-07, "loss": 0.0, "num_tokens": 3680266.0, "reward": 0.5375000238418579, "reward_std": 0.1462521106004715, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.4375, "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": 91.0, "completions/max_terminated_length": 91.0, "completions/mean_length": 13.234375, "completions/mean_terminated_length": 13.234375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.0875, "frac_reward_zero_std": 0.75, "grad_norm": 17.338909149169922, "learning_rate": 7.871643313414718e-07, "loss": 0.0, "num_tokens": 3758505.0, "reward": 0.6312500238418579, "reward_std": 0.1246790662407875, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 88.0, "completions/max_terminated_length": 88.0, "completions/mean_length": 13.3125, "completions/mean_terminated_length": 13.3125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.09, "frac_reward_zero_std": 0.25, "grad_norm": 14.293464660644531, "learning_rate": 7.734740790612136e-07, "loss": 0.0, "num_tokens": 3855685.0, "reward": 0.37031251192092896, "reward_std": 0.2753506004810333, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.3145764470100403, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 184.0, "completions/max_terminated_length": 184.0, "completions/mean_length": 16.765625, "completions/mean_terminated_length": 16.765625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.0925, "frac_reward_zero_std": 0.625, "grad_norm": 11.55620002746582, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 3975158.0, "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": 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": 13.0, "completions/max_terminated_length": 13.0, "completions/mean_length": 11.734375, "completions/mean_terminated_length": 11.734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.5, "grad_norm": 37.207035064697266, "learning_rate": 7.452117519152541e-07, "loss": 0.0, "num_tokens": 4072213.0, "reward": 0.3148437738418579, "reward_std": 0.1323331594467163, "rewards/format_reward/mean": 0.9609375, "rewards/format_reward/std": 0.16194961965084076, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 201.0, "completions/max_terminated_length": 201.0, "completions/mean_length": 15.328125, "completions/mean_terminated_length": 15.328125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.0975, "frac_reward_zero_std": 0.75, "grad_norm": 6.73227071762085, "learning_rate": 7.306705814893439e-07, "loss": 0.0, "num_tokens": 4170890.0, "reward": 0.2406250238418579, "reward_std": 0.10205793380737305, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.140625, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "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": 117.0, "completions/max_terminated_length": 117.0, "completions/mean_length": 16.265625, "completions/mean_terminated_length": 16.265625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1, "frac_reward_zero_std": 0.625, "grad_norm": 9.96992015838623, "learning_rate": 7.158771761692464e-07, "loss": 0.0, "num_tokens": 4275931.0, "reward": 0.5375000238418579, "reward_std": 0.16675157845020294, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1025, "frac_reward_zero_std": 0.625, "grad_norm": 9.436807632446289, "learning_rate": 7.008477123264847e-07, "loss": -0.0, "num_tokens": 4370739.0, "reward": 0.7718750238418579, "reward_std": 0.1530819982290268, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.671875, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 18.0, "completions/max_terminated_length": 18.0, "completions/mean_length": 12.90625, "completions/mean_terminated_length": 12.90625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.105, "frac_reward_zero_std": 0.5, "grad_norm": 13.235849380493164, "learning_rate": 6.855986244591103e-07, "loss": 0.0, "num_tokens": 4460485.0, "reward": 0.5218750238418579, "reward_std": 0.2109457552433014, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1075, "frac_reward_zero_std": 0.75, "grad_norm": 10.07264518737793, "learning_rate": 6.701465872208216e-07, "loss": 0.0, "num_tokens": 4569461.0, "reward": 0.4125000238418579, "reward_std": 0.10888782143592834, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 12.0, "completions/max_terminated_length": 12.0, "completions/mean_length": 12.0, "completions/mean_terminated_length": 12.0, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.11, "frac_reward_zero_std": 0.625, "grad_norm": 7.959709167480469, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 4650237.0, "reward": 0.2406250238418579, "reward_std": 0.13258251547813416, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.140625, "rewards/mcq_exact_match_reward/std": 0.3503824472427368, "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": 18.0, "completions/max_terminated_length": 18.0, "completions/mean_length": 12.484375, "completions/mean_terminated_length": 12.484375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1125, "frac_reward_zero_std": 0.875, "grad_norm": 6.270750999450684, "learning_rate": 6.387014543809223e-07, "loss": 0.0, "num_tokens": 4743260.0, "reward": 0.6468750238418579, "reward_std": 0.0646936446428299, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 20.0, "completions/max_terminated_length": 20.0, "completions/mean_length": 12.21875, "completions/mean_terminated_length": 12.21875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.115, "frac_reward_zero_std": 0.5, "grad_norm": 15.185348510742188, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 4852386.0, "reward": 0.35468751192092896, "reward_std": 0.23586463928222656, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.3145764470100403, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 0.109375, "completions/max_length": 2048.0, "completions/max_terminated_length": 150.0, "completions/mean_length": 237.328125, "completions/mean_terminated_length": 14.964912414550781, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1175, "frac_reward_zero_std": 0.75, "grad_norm": 9.276894569396973, "learning_rate": 6.066498153718734e-07, "loss": 0.0, "num_tokens": 4957703.0, "reward": 0.24531251192092896, "reward_std": 0.09280776977539062, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.3145764470100403, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 90.0, "completions/max_terminated_length": 90.0, "completions/mean_length": 15.203125, "completions/mean_terminated_length": 15.203125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.12, "frac_reward_zero_std": 0.625, "grad_norm": 12.503599166870117, "learning_rate": 5.90440267166055e-07, "loss": 0.0, "num_tokens": 5075004.0, "reward": 0.7093750238418579, "reward_std": 0.15992169082164764, "rewards/format_reward/mean": 1.0, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 198.0, "completions/max_terminated_length": 198.0, "completions/mean_length": 16.53125, "completions/mean_terminated_length": 16.53125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1225, "frac_reward_zero_std": 0.875, "grad_norm": 4.414046287536621, "learning_rate": 5.741318238559209e-07, "loss": 0.0, "num_tokens": 5159358.0, "reward": 0.43125003576278687, "reward_std": 0.0578637570142746, "rewards/format_reward/mean": 0.875, "rewards/format_reward/std": 0.3333333432674408, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 19.0, "completions/mean_length": 44.890625, "completions/mean_terminated_length": 13.09523868560791, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.125, "frac_reward_zero_std": 0.625, "grad_norm": 17.49028968811035, "learning_rate": 5.577423184847931e-07, "loss": 0.0, "num_tokens": 5264119.0, "reward": 0.7859375476837158, "reward_std": 0.15604791045188904, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "step": 50 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 5264119, "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 }