{ "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": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1333.0, "completions/mean_length": 670.328125, "completions/mean_terminated_length": 578.4833374023438, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0025, "frac_reward_zero_std": 0.0, "grad_norm": 3.2621467113494873, "learning_rate": 0.0, "loss": 0.0, "num_tokens": 163837.0, "reward": 0.14531250298023224, "reward_std": 0.2220301628112793, "rewards/format_reward/mean": 0.359375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.109375, "rewards/mcq_exact_match_reward/std": 0.3145764470100403, "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": 1197.0, "completions/mean_length": 550.203125, "completions/mean_terminated_length": 526.4285888671875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.005, "frac_reward_zero_std": 0.125, "grad_norm": 1.0065436363220215, "learning_rate": 2e-07, "loss": -0.0, "num_tokens": 285818.0, "reward": 0.25703123211860657, "reward_std": 0.33974623680114746, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.21347814798355103, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.0, "completions/max_length": 1550.0, "completions/max_terminated_length": 1550.0, "completions/mean_length": 609.8125, "completions/mean_terminated_length": 609.8125, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0075, "frac_reward_zero_std": 0.0, "grad_norm": 16.26173210144043, "learning_rate": 4e-07, "loss": -0.0, "num_tokens": 425830.0, "reward": 0.3218749761581421, "reward_std": 0.38619568943977356, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 1869.0, "completions/mean_length": 608.9375, "completions/mean_terminated_length": 586.0952758789062, "completions/min_length": 150.0, "completions/min_terminated_length": 150.0, "epoch": 0.01, "frac_reward_zero_std": 0.0, "grad_norm": 1.1259515285491943, "learning_rate": 6e-07, "loss": 0.0, "num_tokens": 539634.0, "reward": 0.20624998211860657, "reward_std": 0.2805548310279846, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 1687.0, "completions/max_terminated_length": 1687.0, "completions/mean_length": 474.296875, "completions/mean_terminated_length": 474.296875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0125, "frac_reward_zero_std": 0.0, "grad_norm": 12.274923324584961, "learning_rate": 8e-07, "loss": -0.0, "num_tokens": 650405.0, "reward": 0.23749998211860657, "reward_std": 0.3483607769012451, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2025.0, "completions/mean_length": 759.984375, "completions/mean_terminated_length": 696.6392822265625, "completions/min_length": 7.0, "completions/min_terminated_length": 7.0, "epoch": 0.015, "frac_reward_zero_std": 0.0, "grad_norm": 4.114074230194092, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 813972.0, "reward": 0.21171873807907104, "reward_std": 0.32983919978141785, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.28423789143562317, "rewards/mcq_exact_match_reward/mean": 0.171875, "rewards/mcq_exact_match_reward/std": 0.38025420904159546, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 972.0, "completions/mean_length": 349.5, "completions/mean_terminated_length": 322.5397033691406, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 15.460124015808105, "learning_rate": 9.99726628670463e-07, "loss": 0.0, "num_tokens": 923932.0, "reward": 0.1875, "reward_std": 0.2971411943435669, "rewards/format_reward/mean": 0.3125, "rewards/format_reward/std": 0.2597312331199646, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1659.0, "completions/mean_length": 707.171875, "completions/mean_terminated_length": 685.888916015625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 3.5198187828063965, "learning_rate": 9.989068136093872e-07, "loss": -0.0, "num_tokens": 1082751.0, "reward": 0.2554687261581421, "reward_std": 0.38829028606414795, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.22257846593856812, "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": 0.0, "completions/max_length": 1745.0, "completions/max_terminated_length": 1745.0, "completions/mean_length": 478.65625, "completions/mean_terminated_length": 478.65625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0225, "frac_reward_zero_std": 0.0, "grad_norm": 14.440953254699707, "learning_rate": 9.975414512725056e-07, "loss": 0.0, "num_tokens": 1205465.0, "reward": 0.2867187261581421, "reward_std": 0.37794092297554016, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.2708333432674408, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 1505.0, "completions/mean_length": 593.53125, "completions/mean_terminated_length": 570.4444580078125, "completions/min_length": 121.0, "completions/min_terminated_length": 121.0, "epoch": 0.025, "frac_reward_zero_std": 0.0, "grad_norm": 0.9528844356536865, "learning_rate": 9.956320346634875e-07, "loss": -0.0, "num_tokens": 1338507.0, "reward": 0.3070312440395355, "reward_std": 0.31435659527778625, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.244862899184227, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1932.0, "completions/mean_length": 664.46875, "completions/mean_terminated_length": 642.5079956054688, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0275, "frac_reward_zero_std": 0.0, "grad_norm": 10.59503173828125, "learning_rate": 9.931806517013612e-07, "loss": -0.0, "num_tokens": 1478297.0, "reward": 0.37812498211860657, "reward_std": 0.3966887593269348, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 1732.0, "completions/mean_length": 590.5, "completions/mean_terminated_length": 518.8196411132812, "completions/min_length": 18.0, "completions/min_terminated_length": 18.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 2.6676149368286133, "learning_rate": 9.901899829374047e-07, "loss": -0.0, "num_tokens": 1605553.0, "reward": 0.5046874284744263, "reward_std": 0.4666898548603058, "rewards/format_reward/mean": 0.359375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.46875, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 1759.0, "completions/max_terminated_length": 1759.0, "completions/mean_length": 530.796875, "completions/mean_terminated_length": 530.796875, "completions/min_length": 5.0, "completions/min_terminated_length": 5.0, "epoch": 0.0325, "frac_reward_zero_std": 0.0, "grad_norm": 17.730297088623047, "learning_rate": 9.866632986240029e-07, "loss": 0.0, "num_tokens": 1756204.0, "reward": 0.19374999403953552, "reward_std": 0.2921380400657654, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.30860671401023865, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 1871.0, "completions/max_terminated_length": 1871.0, "completions/mean_length": 671.3125, "completions/mean_terminated_length": 671.3125, "completions/min_length": 112.0, "completions/min_terminated_length": 112.0, "epoch": 0.035, "frac_reward_zero_std": 0.125, "grad_norm": 1.2168720960617065, "learning_rate": 9.826044551386742e-07, "loss": 0.0, "num_tokens": 1905080.0, "reward": 0.3187499940395355, "reward_std": 0.23982028663158417, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.2357022762298584, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 2002.0, "completions/mean_length": 629.546875, "completions/mean_terminated_length": 559.786865234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.0, "grad_norm": 7.674581050872803, "learning_rate": 9.780178907671788e-07, "loss": -0.0, "num_tokens": 2052819.0, "reward": 0.41093748807907104, "reward_std": 0.4283228814601898, "rewards/format_reward/mean": 0.359375, "rewards/format_reward/std": 0.301698237657547, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.0625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1774.0, "completions/mean_length": 649.40625, "completions/mean_terminated_length": 556.1666870117188, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04, "frac_reward_zero_std": 0.0, "grad_norm": 8.375741004943848, "learning_rate": 9.729086208503173e-07, "loss": 0.0, "num_tokens": 2193149.0, "reward": 0.3218750059604645, "reward_std": 0.4034338593482971, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1522.0, "completions/mean_length": 552.21875, "completions/mean_terminated_length": 503.96771240234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0425, "frac_reward_zero_std": 0.125, "grad_norm": 6.322883605957031, "learning_rate": 9.672822322997304e-07, "loss": -0.0, "num_tokens": 2313955.0, "reward": 0.2828124761581421, "reward_std": 0.2805572748184204, "rewards/format_reward/mean": 0.484375, "rewards/format_reward/std": 0.1985812783241272, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 1023.0, "completions/max_terminated_length": 1023.0, "completions/mean_length": 398.1875, "completions/mean_terminated_length": 398.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.045, "frac_reward_zero_std": 0.125, "grad_norm": 15.723276138305664, "learning_rate": 9.611448774886923e-07, "loss": -0.0, "num_tokens": 2432095.0, "reward": 0.32265621423721313, "reward_std": 0.3714010715484619, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.19012710452079773, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1410.0, "completions/mean_length": 321.1875, "completions/mean_terminated_length": 293.7778015136719, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0475, "frac_reward_zero_std": 0.125, "grad_norm": 12.669057846069336, "learning_rate": 9.545032675245813e-07, "loss": 0.0, "num_tokens": 2547315.0, "reward": 0.4078124761581421, "reward_std": 0.32604295015335083, "rewards/format_reward/mean": 0.484375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1520.0, "completions/mean_length": 483.375, "completions/mean_terminated_length": 432.9031982421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.125, "grad_norm": 5.056563854217529, "learning_rate": 9.473646649103817e-07, "loss": 0.0, "num_tokens": 2688251.0, "reward": 0.4179687201976776, "reward_std": 0.28936219215393066, "rewards/format_reward/mean": 0.4296875, "rewards/format_reward/std": 0.19654127955436707, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 1917.0, "completions/max_terminated_length": 1917.0, "completions/mean_length": 395.4375, "completions/mean_terminated_length": 395.4375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0525, "frac_reward_zero_std": 0.0, "grad_norm": 7.2111735343933105, "learning_rate": 9.397368756032444e-07, "loss": 0.0, "num_tokens": 2798335.0, "reward": 0.2992187440395355, "reward_std": 0.40782421827316284, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.3020063638687134, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1532.0, "completions/mean_length": 428.96875, "completions/mean_terminated_length": 349.34423828125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.25, "grad_norm": 6.910099029541016, "learning_rate": 9.316282404787869e-07, "loss": -0.0, "num_tokens": 2936109.0, "reward": 0.38906246423721313, "reward_std": 0.3330124616622925, "rewards/format_reward/mean": 0.453125, "rewards/format_reward/std": 0.14689241349697113, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 1225.0, "completions/max_terminated_length": 1225.0, "completions/mean_length": 410.84375, "completions/mean_terminated_length": 410.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0575, "frac_reward_zero_std": 0.0, "grad_norm": 6.530163288116455, "learning_rate": 9.230476262104676e-07, "loss": 0.0, "num_tokens": 3049211.0, "reward": 0.500781238079071, "reward_std": 0.3821592330932617, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.24282869696617126, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1361.0, "completions/mean_length": 550.046875, "completions/mean_terminated_length": 526.2698974609375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.25, "grad_norm": 6.006289958953857, "learning_rate": 9.1400441557401e-07, "loss": 0.0, "num_tokens": 3182486.0, "reward": 0.2515625059604645, "reward_std": 0.2533089518547058, "rewards/format_reward/mean": 0.484375, "rewards/format_reward/std": 0.125, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1483.0, "completions/mean_length": 562.4375, "completions/mean_terminated_length": 538.857177734375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.0, "grad_norm": 9.911283493041992, "learning_rate": 9.045084971874737e-07, "loss": 0.0, "num_tokens": 3307874.0, "reward": 0.26249998807907104, "reward_std": 0.3797960877418518, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.18898223340511322, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1326.0, "completions/mean_length": 434.34375, "completions/mean_terminated_length": 382.2903137207031, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.065, "frac_reward_zero_std": 0.125, "grad_norm": 15.86373233795166, "learning_rate": 8.945702546981968e-07, "loss": 0.0, "num_tokens": 3443624.0, "reward": 0.4367187023162842, "reward_std": 0.3652403950691223, "rewards/format_reward/mean": 0.4609375, "rewards/format_reward/std": 0.2236899733543396, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1647.0, "completions/mean_length": 336.78125, "completions/mean_terminated_length": 309.61907958984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0675, "frac_reward_zero_std": 0.25, "grad_norm": 10.825465202331543, "learning_rate": 8.842005554284295e-07, "loss": 0.0, "num_tokens": 3555002.0, "reward": 0.33437496423721313, "reward_std": 0.2338321954011917, "rewards/format_reward/mean": 0.53125, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1966.0, "completions/mean_length": 512.546875, "completions/mean_terminated_length": 488.17462158203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07, "frac_reward_zero_std": 0.25, "grad_norm": 7.471256732940674, "learning_rate": 8.734107384920769e-07, "loss": 0.0, "num_tokens": 3685773.0, "reward": 0.3968749940395355, "reward_std": 0.17824815213680267, "rewards/format_reward/mean": 0.53125, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1400.0, "completions/mean_length": 421.359375, "completions/mean_terminated_length": 395.5397033691406, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0725, "frac_reward_zero_std": 0.0, "grad_norm": 8.75351619720459, "learning_rate": 8.622126023955445e-07, "loss": 0.0, "num_tokens": 3815532.0, "reward": 0.29999998211860657, "reward_std": 0.32218557596206665, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.19920477271080017, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1475.0, "completions/mean_length": 393.203125, "completions/mean_terminated_length": 366.9365234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.075, "frac_reward_zero_std": 0.0, "grad_norm": 11.689899444580078, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 3950689.0, "reward": 0.390625, "reward_std": 0.36572664976119995, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.21593283116817474, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1754.0, "completions/mean_length": 249.109375, "completions/mean_terminated_length": 191.0806427001953, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.25, "grad_norm": 14.236430168151855, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 4059408.0, "reward": 0.4140625, "reward_std": 0.2643175721168518, "rewards/format_reward/mean": 0.546875, "rewards/format_reward/std": 0.2630521357059479, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 1746.0, "completions/max_terminated_length": 1746.0, "completions/mean_length": 371.984375, "completions/mean_terminated_length": 371.984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08, "frac_reward_zero_std": 0.125, "grad_norm": 16.345598220825195, "learning_rate": 8.262928807620843e-07, "loss": -0.0, "num_tokens": 4178695.0, "reward": 0.2812499701976776, "reward_std": 0.29161226749420166, "rewards/format_reward/mean": 0.625, "rewards/format_reward/std": 0.26726123690605164, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1352.0, "completions/mean_length": 313.21875, "completions/mean_terminated_length": 285.68255615234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.0, "grad_norm": 11.585888862609863, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 4298965.0, "reward": 0.4039062261581421, "reward_std": 0.3795614242553711, "rewards/format_reward/mean": 0.6015625, "rewards/format_reward/std": 0.26991602778434753, "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": 1348.0, "completions/max_terminated_length": 1348.0, "completions/mean_length": 225.765625, "completions/mean_terminated_length": 225.765625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.085, "frac_reward_zero_std": 0.0, "grad_norm": 24.46437644958496, "learning_rate": 8.005405736415125e-07, "loss": -0.0, "num_tokens": 4395494.0, "reward": 0.639843761920929, "reward_std": 0.3820061683654785, "rewards/format_reward/mean": 0.6171875, "rewards/format_reward/std": 0.21347814798355103, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1381.0, "completions/max_terminated_length": 1381.0, "completions/mean_length": 89.078125, "completions/mean_terminated_length": 89.078125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0875, "frac_reward_zero_std": 0.25, "grad_norm": 19.375133514404297, "learning_rate": 7.871643313414718e-07, "loss": 0.0, "num_tokens": 4504643.0, "reward": 0.6226562261581421, "reward_std": 0.26231664419174194, "rewards/format_reward/mean": 0.6015625, "rewards/format_reward/std": 0.22146137058734894, "rewards/mcq_exact_match_reward/mean": 0.5625, "rewards/mcq_exact_match_reward/std": 0.5, "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": 1769.0, "completions/max_terminated_length": 1769.0, "completions/mean_length": 151.1875, "completions/mean_terminated_length": 151.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09, "frac_reward_zero_std": 0.125, "grad_norm": 21.263975143432617, "learning_rate": 7.734740790612136e-07, "loss": -0.0, "num_tokens": 4609751.0, "reward": 0.5179687738418579, "reward_std": 0.3356044888496399, "rewards/format_reward/mean": 0.6484375, "rewards/format_reward/std": 0.27716949582099915, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 999.0, "completions/max_terminated_length": 999.0, "completions/mean_length": 67.546875, "completions/mean_terminated_length": 67.546875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0925, "frac_reward_zero_std": 0.25, "grad_norm": 39.54612350463867, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 4707386.0, "reward": 0.44218748807907104, "reward_std": 0.22757485508918762, "rewards/format_reward/mean": 0.671875, "rewards/format_reward/std": 0.33592742681503296, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 684.0, "completions/max_terminated_length": 684.0, "completions/mean_length": 60.578125, "completions/mean_terminated_length": 60.578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.375, "grad_norm": 26.625511169433594, "learning_rate": 7.452117519152541e-07, "loss": 0.0, "num_tokens": 4781535.0, "reward": 0.4804687798023224, "reward_std": 0.22189205884933472, "rewards/format_reward/mean": 0.8984375, "rewards/format_reward/std": 0.25479042530059814, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 1075.0, "completions/max_terminated_length": 1075.0, "completions/mean_length": 122.578125, "completions/mean_terminated_length": 122.578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0975, "frac_reward_zero_std": 0.25, "grad_norm": 26.261032104492188, "learning_rate": 7.306705814893439e-07, "loss": -0.0, "num_tokens": 4870412.0, "reward": 0.5367187261581421, "reward_std": 0.2549279034137726, "rewards/format_reward/mean": 0.8359375, "rewards/format_reward/std": 0.2824873626232147, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 835.0, "completions/max_terminated_length": 835.0, "completions/mean_length": 46.765625, "completions/mean_terminated_length": 46.765625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1, "frac_reward_zero_std": 0.0, "grad_norm": 28.509124755859375, "learning_rate": 7.158771761692464e-07, "loss": -0.0, "num_tokens": 4948165.0, "reward": 0.51171875, "reward_std": 0.36350616812705994, "rewards/format_reward/mean": 0.8984375, "rewards/format_reward/std": 0.22146137058734894, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1136.0, "completions/max_terminated_length": 1136.0, "completions/mean_length": 104.03125, "completions/mean_terminated_length": 104.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1025, "frac_reward_zero_std": 0.0, "grad_norm": 67.95924377441406, "learning_rate": 7.008477123264847e-07, "loss": 0.0, "num_tokens": 5034767.0, "reward": 0.7078125476837158, "reward_std": 0.2515289783477783, "rewards/format_reward/mean": 0.828125, "rewards/format_reward/std": 0.25539806485176086, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1586.0, "completions/mean_length": 255.484375, "completions/mean_terminated_length": 197.66128540039062, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.105, "frac_reward_zero_std": 0.25, "grad_norm": 18.142475128173828, "learning_rate": 6.855986244591103e-07, "loss": 0.0, "num_tokens": 5149230.0, "reward": 0.4375, "reward_std": 0.2851329743862152, "rewards/format_reward/mean": 0.78125, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1782.0, "completions/mean_length": 168.296875, "completions/mean_terminated_length": 107.66128540039062, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1075, "frac_reward_zero_std": 0.375, "grad_norm": 14.344269752502441, "learning_rate": 6.701465872208216e-07, "loss": 0.0, "num_tokens": 5262985.0, "reward": 0.4078125059604645, "reward_std": 0.12202177941799164, "rewards/format_reward/mean": 0.796875, "rewards/format_reward/std": 0.2916666865348816, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 1162.0, "completions/max_terminated_length": 1162.0, "completions/mean_length": 149.140625, "completions/mean_terminated_length": 149.140625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.11, "frac_reward_zero_std": 0.375, "grad_norm": 8.833393096923828, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 5357714.0, "reward": 0.633593738079071, "reward_std": 0.25169724225997925, "rewards/format_reward/mean": 0.8671875, "rewards/format_reward/std": 0.23974503576755524, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 125.0, "completions/max_terminated_length": 125.0, "completions/mean_length": 15.0625, "completions/mean_terminated_length": 15.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1125, "frac_reward_zero_std": 0.5, "grad_norm": 15.923088073730469, "learning_rate": 6.387014543809223e-07, "loss": 0.0, "num_tokens": 5422710.0, "reward": 0.5984375476837158, "reward_std": 0.20514346659183502, "rewards/format_reward/mean": 0.984375, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "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": 0.046875, "completions/max_length": 2048.0, "completions/max_terminated_length": 1669.0, "completions/mean_length": 192.6875, "completions/mean_terminated_length": 101.44261932373047, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.115, "frac_reward_zero_std": 0.375, "grad_norm": 8.417112350463867, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 5521690.0, "reward": 0.41796875, "reward_std": 0.21655938029289246, "rewards/format_reward/mean": 0.8984375, "rewards/format_reward/std": 0.25479042530059814, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 465.0, "completions/max_terminated_length": 465.0, "completions/mean_length": 31.28125, "completions/mean_terminated_length": 31.28125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1175, "frac_reward_zero_std": 0.5, "grad_norm": 18.831295013427734, "learning_rate": 6.066498153718734e-07, "loss": 0.0, "num_tokens": 5623596.0, "reward": 0.706250011920929, "reward_std": 0.21585926413536072, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.1510545015335083, "rewards/mcq_exact_match_reward/mean": 0.609375, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 1428.0, "completions/max_terminated_length": 1428.0, "completions/mean_length": 70.96875, "completions/mean_terminated_length": 70.96875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.12, "frac_reward_zero_std": 0.5, "grad_norm": 16.980724334716797, "learning_rate": 5.90440267166055e-07, "loss": 0.0, "num_tokens": 5729122.0, "reward": 0.2734375, "reward_std": 0.1521616280078888, "rewards/format_reward/mean": 0.859375, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 397.0, "completions/max_terminated_length": 397.0, "completions/mean_length": 49.625, "completions/mean_terminated_length": 49.625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1225, "frac_reward_zero_std": 0.5, "grad_norm": 19.450817108154297, "learning_rate": 5.741318238559209e-07, "loss": -0.0, "num_tokens": 5801394.0, "reward": 0.45859378576278687, "reward_std": 0.190185084939003, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 914.0, "completions/max_terminated_length": 914.0, "completions/mean_length": 127.84375, "completions/mean_terminated_length": 127.84375, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.125, "frac_reward_zero_std": 0.25, "grad_norm": 15.480325698852539, "learning_rate": 5.577423184847931e-07, "loss": -0.0, "num_tokens": 5912240.0, "reward": 0.596875011920929, "reward_std": 0.33784300088882446, "rewards/format_reward/mean": 0.96875, "rewards/format_reward/std": 0.1510545015335083, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "step": 50 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 5912240, "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 }