{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.11160714285714286, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1786.0, "completions/mean_length": 482.234375, "completions/mean_terminated_length": 457.3809814453125, "completions/min_length": 5.0, "completions/min_terminated_length": 5.0, "epoch": 0.002232142857142857, "frac_reward_zero_std": 0.0, "grad_norm": 11.518397331237793, "learning_rate": 0.0, "loss": -0.0, "num_tokens": 110151.0, "reward": 0.2671874761581421, "reward_std": 0.3433631956577301, "rewards/format_reward/mean": 0.328125, "rewards/format_reward/std": 0.25539806485176086, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 1700.0, "completions/mean_length": 460.734375, "completions/mean_terminated_length": 435.5397033691406, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.004464285714285714, "frac_reward_zero_std": 0.0, "grad_norm": 25.52145767211914, "learning_rate": 2e-07, "loss": 0.0, "num_tokens": 222366.0, "reward": 0.1953124850988388, "reward_std": 0.287118136882782, "rewards/format_reward/mean": 0.390625, "rewards/format_reward/std": 0.2592533528804779, "rewards/mcq_exact_match_reward/mean": 0.15625, "rewards/mcq_exact_match_reward/std": 0.36596253514289856, "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": 1822.0, "completions/mean_length": 611.0625, "completions/mean_terminated_length": 588.2540283203125, "completions/min_length": 164.0, "completions/min_terminated_length": 164.0, "epoch": 0.006696428571428571, "frac_reward_zero_std": 0.0, "grad_norm": 1.18820059299469, "learning_rate": 4e-07, "loss": -0.0, "num_tokens": 336234.0, "reward": 0.2554687261581421, "reward_std": 0.3832392692565918, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.255761981010437, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1977.0, "completions/mean_length": 499.90625, "completions/mean_terminated_length": 449.9677429199219, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.008928571428571428, "frac_reward_zero_std": 0.0, "grad_norm": 17.768474578857422, "learning_rate": 6e-07, "loss": 0.0, "num_tokens": 464092.0, "reward": 0.30390623211860657, "reward_std": 0.35698384046554565, "rewards/format_reward/mean": 0.3828125, "rewards/format_reward/std": 0.2780628502368927, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1951.0, "completions/mean_length": 600.375, "completions/mean_terminated_length": 577.3968505859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.011160714285714286, "frac_reward_zero_std": 0.125, "grad_norm": 3.8049168586730957, "learning_rate": 8e-07, "loss": -0.0, "num_tokens": 603132.0, "reward": 0.3671875, "reward_std": 0.25782060623168945, "rewards/format_reward/mean": 0.390625, "rewards/format_reward/std": 0.22658175230026245, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 1384.0, "completions/mean_length": 509.0, "completions/mean_terminated_length": 484.5714416503906, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.013392857142857142, "frac_reward_zero_std": 0.125, "grad_norm": 2.7166552543640137, "learning_rate": 1e-06, "loss": -0.0, "num_tokens": 736044.0, "reward": 0.36562496423721313, "reward_std": 0.3284198045730591, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.2519763112068176, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 1386.0, "completions/max_terminated_length": 1386.0, "completions/mean_length": 572.0625, "completions/mean_terminated_length": 572.0625, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "epoch": 0.015625, "frac_reward_zero_std": 0.125, "grad_norm": 4.848062038421631, "learning_rate": 9.99726628670463e-07, "loss": -0.0, "num_tokens": 869224.0, "reward": 0.43437495827674866, "reward_std": 0.3725493550300598, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 1609.0, "completions/max_terminated_length": 1609.0, "completions/mean_length": 533.875, "completions/mean_terminated_length": 533.875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.017857142857142856, "frac_reward_zero_std": 0.25, "grad_norm": 2.655893325805664, "learning_rate": 9.989068136093872e-07, "loss": -0.0, "num_tokens": 1011408.0, "reward": 0.24062499403953552, "reward_std": 0.26043617725372314, "rewards/format_reward/mean": 0.375, "rewards/format_reward/std": 0.2182178944349289, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 1910.0, "completions/max_terminated_length": 1910.0, "completions/mean_length": 480.796875, "completions/mean_terminated_length": 480.796875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.020089285714285716, "frac_reward_zero_std": 0.0, "grad_norm": 3.9405181407928467, "learning_rate": 9.975414512725056e-07, "loss": 0.0, "num_tokens": 1159619.0, "reward": 0.32265621423721313, "reward_std": 0.3424666225910187, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.2605654299259186, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 1526.0, "completions/mean_length": 532.984375, "completions/mean_terminated_length": 508.9365234375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.022321428571428572, "frac_reward_zero_std": 0.125, "grad_norm": 12.9504976272583, "learning_rate": 9.956320346634875e-07, "loss": 0.0, "num_tokens": 1320586.0, "reward": 0.16562499105930328, "reward_std": 0.22279313206672668, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.19669894874095917, "rewards/mcq_exact_match_reward/mean": 0.125, "rewards/mcq_exact_match_reward/std": 0.3333333432674408, "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": 1780.0, "completions/max_terminated_length": 1780.0, "completions/mean_length": 524.78125, "completions/mean_terminated_length": 524.78125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.024553571428571428, "frac_reward_zero_std": 0.125, "grad_norm": 5.030337810516357, "learning_rate": 9.931806517013612e-07, "loss": 0.0, "num_tokens": 1477204.0, "reward": 0.13515624403953552, "reward_std": 0.12314560264348984, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.19012710452079773, "rewards/mcq_exact_match_reward/mean": 0.09375, "rewards/mcq_exact_match_reward/std": 0.29378482699394226, "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": 1806.0, "completions/mean_length": 546.40625, "completions/mean_terminated_length": 472.5573425292969, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.026785714285714284, "frac_reward_zero_std": 0.125, "grad_norm": 1.6112949848175049, "learning_rate": 9.901899829374047e-07, "loss": 0.0, "num_tokens": 1594782.0, "reward": 0.30937498807907104, "reward_std": 0.24110916256904602, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.22712838649749756, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 1289.0, "completions/mean_length": 390.296875, "completions/mean_terminated_length": 363.9841613769531, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.029017857142857144, "frac_reward_zero_std": 0.0, "grad_norm": 15.314018249511719, "learning_rate": 9.866632986240029e-07, "loss": -0.0, "num_tokens": 1729089.0, "reward": 0.46328121423721313, "reward_std": 0.4383506774902344, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.209963858127594, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1632.0, "completions/max_terminated_length": 1632.0, "completions/mean_length": 380.203125, "completions/mean_terminated_length": 380.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03125, "frac_reward_zero_std": 0.0, "grad_norm": 10.621455192565918, "learning_rate": 9.826044551386742e-07, "loss": -0.0, "num_tokens": 1850926.0, "reward": 0.4257812201976776, "reward_std": 0.33724090456962585, "rewards/format_reward/mean": 0.3515625, "rewards/format_reward/std": 0.24688033759593964, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 0.0, "completions/max_length": 1344.0, "completions/max_terminated_length": 1344.0, "completions/mean_length": 293.84375, "completions/mean_terminated_length": 293.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.033482142857142856, "frac_reward_zero_std": 0.125, "grad_norm": 25.159730911254883, "learning_rate": 9.780178907671788e-07, "loss": -0.0, "num_tokens": 1963156.0, "reward": 0.4210937023162842, "reward_std": 0.336973637342453, "rewards/format_reward/mean": 0.4609375, "rewards/format_reward/std": 0.13524486124515533, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1798.0, "completions/mean_length": 335.84375, "completions/mean_terminated_length": 308.66668701171875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03571428571428571, "frac_reward_zero_std": 0.25, "grad_norm": 13.841256141662598, "learning_rate": 9.729086208503173e-07, "loss": 0.0, "num_tokens": 2090050.0, "reward": 0.34296873211860657, "reward_std": 0.35050320625305176, "rewards/format_reward/mean": 0.4609375, "rewards/format_reward/std": 0.13524486124515533, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 1344.0, "completions/max_terminated_length": 1344.0, "completions/mean_length": 229.28125, "completions/mean_terminated_length": 229.28125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03794642857142857, "frac_reward_zero_std": 0.25, "grad_norm": 12.63813591003418, "learning_rate": 9.672822322997304e-07, "loss": -0.0, "num_tokens": 2208012.0, "reward": 0.484375, "reward_std": 0.3500387668609619, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.12198751419782639, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 776.0, "completions/max_terminated_length": 776.0, "completions/mean_length": 103.421875, "completions/mean_terminated_length": 103.421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04017857142857143, "frac_reward_zero_std": 0.0, "grad_norm": 13.648046493530273, "learning_rate": 9.611448774886923e-07, "loss": 0.0, "num_tokens": 2331679.0, "reward": 0.45390620827674866, "reward_std": 0.38506779074668884, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.13886408507823944, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 810.0, "completions/max_terminated_length": 810.0, "completions/mean_length": 106.296875, "completions/mean_terminated_length": 106.296875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.04241071428571429, "frac_reward_zero_std": 0.25, "grad_norm": 19.259902954101562, "learning_rate": 9.545032675245813e-07, "loss": -0.0, "num_tokens": 2444658.0, "reward": 0.3843749761581421, "reward_std": 0.3423978090286255, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 1564.0, "completions/max_terminated_length": 1564.0, "completions/mean_length": 120.828125, "completions/mean_terminated_length": 120.828125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.044642857142857144, "frac_reward_zero_std": 0.375, "grad_norm": 22.730810165405273, "learning_rate": 9.473646649103817e-07, "loss": -0.0, "num_tokens": 2526119.0, "reward": 0.4859374761581421, "reward_std": 0.2837350070476532, "rewards/format_reward/mean": 0.484375, "rewards/format_reward/std": 0.125, "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": 1069.0, "completions/max_terminated_length": 1069.0, "completions/mean_length": 66.25, "completions/mean_terminated_length": 66.25, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.046875, "frac_reward_zero_std": 0.25, "grad_norm": 18.526090621948242, "learning_rate": 9.397368756032444e-07, "loss": 0.0, "num_tokens": 2628455.0, "reward": 0.26171875, "reward_std": 0.2573496103286743, "rewards/format_reward/mean": 0.4296875, "rewards/format_reward/std": 0.1751912236213684, "rewards/mcq_exact_match_reward/mean": 0.21875, "rewards/mcq_exact_match_reward/std": 0.4166666865348816, "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": 498.0, "completions/max_terminated_length": 498.0, "completions/mean_length": 73.625, "completions/mean_terminated_length": 73.625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.049107142857142856, "frac_reward_zero_std": 0.375, "grad_norm": 19.452939987182617, "learning_rate": 9.316282404787869e-07, "loss": 0.0, "num_tokens": 2716383.0, "reward": 0.44218748807907104, "reward_std": 0.26160189509391785, "rewards/format_reward/mean": 0.515625, "rewards/format_reward/std": 0.08768405020236969, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 973.0, "completions/max_terminated_length": 973.0, "completions/mean_length": 60.78125, "completions/mean_terminated_length": 60.78125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05133928571428571, "frac_reward_zero_std": 0.25, "grad_norm": 29.451814651489258, "learning_rate": 9.230476262104676e-07, "loss": -0.0, "num_tokens": 2824601.0, "reward": 0.4726561903953552, "reward_std": 0.3298916220664978, "rewards/format_reward/mean": 0.5078125, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 811.0, "completions/max_terminated_length": 811.0, "completions/mean_length": 68.5625, "completions/mean_terminated_length": 68.5625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05357142857142857, "frac_reward_zero_std": 0.625, "grad_norm": 17.41834831237793, "learning_rate": 9.1400441557401e-07, "loss": 0.0, "num_tokens": 2921741.0, "reward": 0.40859371423721313, "reward_std": 0.16939526796340942, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 145.0, "completions/max_terminated_length": 145.0, "completions/mean_length": 8.203125, "completions/mean_terminated_length": 8.203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05580357142857143, "frac_reward_zero_std": 0.375, "grad_norm": 20.918785095214844, "learning_rate": 9.045084971874737e-07, "loss": -0.0, "num_tokens": 3015794.0, "reward": 0.6437499523162842, "reward_std": 0.29143062233924866, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.59375, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 424.0, "completions/max_terminated_length": 424.0, "completions/mean_length": 12.53125, "completions/mean_terminated_length": 12.53125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05803571428571429, "frac_reward_zero_std": 0.375, "grad_norm": 25.445863723754883, "learning_rate": 8.945702546981968e-07, "loss": -0.0, "num_tokens": 3097308.0, "reward": 0.4093749523162842, "reward_std": 0.28930896520614624, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 511.0, "completions/max_terminated_length": 511.0, "completions/mean_length": 13.890625, "completions/mean_terminated_length": 13.890625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.060267857142857144, "frac_reward_zero_std": 0.125, "grad_norm": 21.52651023864746, "learning_rate": 8.842005554284295e-07, "loss": -0.0, "num_tokens": 3161941.0, "reward": 0.3937499523162842, "reward_std": 0.3913668990135193, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 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": 0.0625, "frac_reward_zero_std": 0.625, "grad_norm": 19.932497024536133, "learning_rate": 8.734107384920769e-07, "loss": -0.0, "num_tokens": 3269837.0, "reward": 0.39374998211860657, "reward_std": 0.1735912561416626, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1304.0, "completions/mean_length": 159.015625, "completions/mean_terminated_length": 98.08064270019531, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06473214285714286, "frac_reward_zero_std": 0.75, "grad_norm": 10.873550415039062, "learning_rate": 8.622126023955445e-07, "loss": -0.0, "num_tokens": 3399606.0, "reward": 0.5593749284744263, "reward_std": 0.10205792635679245, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.1666666716337204, "rewards/mcq_exact_match_reward/mean": 0.515625, "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": 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": 0.06696428571428571, "frac_reward_zero_std": 0.5, "grad_norm": 16.248550415039062, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 3477030.0, "reward": 0.4718749523162842, "reward_std": 0.2109457403421402, "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": 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": 628.0, "completions/max_terminated_length": 628.0, "completions/mean_length": 15.71875, "completions/mean_terminated_length": 15.71875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06919642857142858, "frac_reward_zero_std": 0.875, "grad_norm": 15.107515335083008, "learning_rate": 8.386407858128706e-07, "loss": -0.0, "num_tokens": 3570292.0, "reward": 0.28437498211860657, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 135.0, "completions/max_terminated_length": 135.0, "completions/mean_length": 8.015625, "completions/mean_terminated_length": 8.015625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07142857142857142, "frac_reward_zero_std": 0.5, "grad_norm": 19.065568923950195, "learning_rate": 8.262928807620843e-07, "loss": 0.0, "num_tokens": 3664733.0, "reward": 0.23749998211860657, "reward_std": 0.19727616012096405, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.1875, "rewards/mcq_exact_match_reward/std": 0.39339789748191833, "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": 458.0, "completions/max_terminated_length": 458.0, "completions/mean_length": 13.0625, "completions/mean_terminated_length": 13.0625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07366071428571429, "frac_reward_zero_std": 0.375, "grad_norm": 18.39124298095703, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 3769953.0, "reward": 0.4562499523162842, "reward_std": 0.2709311544895172, "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": 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": 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": 0.07589285714285714, "frac_reward_zero_std": 0.625, "grad_norm": 17.875316619873047, "learning_rate": 8.005405736415125e-07, "loss": 0.0, "num_tokens": 3870553.0, "reward": 0.4406249523162842, "reward_std": 0.15992167592048645, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 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": 0.078125, "frac_reward_zero_std": 0.5, "grad_norm": 21.13361358642578, "learning_rate": 7.871643313414718e-07, "loss": -0.0, "num_tokens": 3964689.0, "reward": 0.29999998211860657, "reward_std": 0.2540663480758667, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 227.0, "completions/max_terminated_length": 227.0, "completions/mean_length": 9.453125, "completions/mean_terminated_length": 9.453125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08035714285714286, "frac_reward_zero_std": 0.75, "grad_norm": 14.552275657653809, "learning_rate": 7.734740790612136e-07, "loss": 0.0, "num_tokens": 4058494.0, "reward": 0.17499998211860657, "reward_std": 0.0883883386850357, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.125, "rewards/mcq_exact_match_reward/std": 0.3333333432674408, "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": 8.0, "completions/max_terminated_length": 8.0, "completions/mean_length": 6.03125, "completions/mean_terminated_length": 6.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08258928571428571, "frac_reward_zero_std": 0.625, "grad_norm": 35.18708419799805, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 4148304.0, "reward": 0.43984371423721313, "reward_std": 0.12476710975170135, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 8.0, "completions/max_terminated_length": 8.0, "completions/mean_length": 6.03125, "completions/mean_terminated_length": 6.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08482142857142858, "frac_reward_zero_std": 0.25, "grad_norm": 25.0270938873291, "learning_rate": 7.452117519152541e-07, "loss": 0.0, "num_tokens": 4248122.0, "reward": 0.37734371423721313, "reward_std": 0.32051464915275574, "rewards/format_reward/mean": 0.4921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "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": 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": 0.08705357142857142, "frac_reward_zero_std": 0.625, "grad_norm": 29.618751525878906, "learning_rate": 7.306705814893439e-07, "loss": -0.0, "num_tokens": 4352098.0, "reward": 0.4562499523162842, "reward_std": 0.1462520956993103, "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": 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": 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": 0.08928571428571429, "frac_reward_zero_std": 0.875, "grad_norm": 12.265328407287598, "learning_rate": 7.158771761692464e-07, "loss": -0.0, "num_tokens": 4447730.0, "reward": 0.7374999523162842, "reward_std": 0.06681530177593231, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.6875, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 11.0, "completions/max_terminated_length": 11.0, "completions/mean_length": 6.078125, "completions/mean_terminated_length": 6.078125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09151785714285714, "frac_reward_zero_std": 0.75, "grad_norm": 21.324724197387695, "learning_rate": 7.008477123264847e-07, "loss": -0.0, "num_tokens": 4548447.0, "reward": 0.4874999523162842, "reward_std": 0.1157275065779686, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 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": 0.09375, "frac_reward_zero_std": 0.75, "grad_norm": 13.844320297241211, "learning_rate": 6.855986244591103e-07, "loss": 0.0, "num_tokens": 4642335.0, "reward": 0.33124998211860657, "reward_std": 0.0883883386850357, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "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": 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": 0.09598214285714286, "frac_reward_zero_std": 0.5, "grad_norm": 29.367101669311523, "learning_rate": 6.701465872208216e-07, "loss": -0.0, "num_tokens": 4726287.0, "reward": 0.5343749523162842, "reward_std": 0.22673697769641876, "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": 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": 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": 0.09821428571428571, "frac_reward_zero_std": 0.75, "grad_norm": 27.19498062133789, "learning_rate": 6.545084971874736e-07, "loss": -0.0, "num_tokens": 4799447.0, "reward": 0.7062499523162842, "reward_std": 0.1293872892856598, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 2016.0, "completions/max_terminated_length": 2016.0, "completions/mean_length": 37.40625, "completions/mean_terminated_length": 37.40625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.10044642857142858, "frac_reward_zero_std": 0.625, "grad_norm": 19.650163650512695, "learning_rate": 6.387014543809223e-07, "loss": -0.0, "num_tokens": 4914689.0, "reward": 0.7062499523162842, "reward_std": 0.16675156354904175, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 8.0, "completions/max_terminated_length": 8.0, "completions/mean_length": 6.03125, "completions/mean_terminated_length": 6.03125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.10267857142857142, "frac_reward_zero_std": 0.75, "grad_norm": 27.214603424072266, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 5024859.0, "reward": 0.36249998211860657, "reward_std": 0.10888781398534775, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 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": 0.10491071428571429, "frac_reward_zero_std": 0.5, "grad_norm": 29.722211837768555, "learning_rate": 6.066498153718734e-07, "loss": -0.0, "num_tokens": 5118379.0, "reward": 0.5812499523162842, "reward_std": 0.2041158527135849, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 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": 0.10714285714285714, "frac_reward_zero_std": 0.875, "grad_norm": 12.212489128112793, "learning_rate": 5.90440267166055e-07, "loss": -0.0, "num_tokens": 5206979.0, "reward": 0.28437498211860657, "reward_std": 0.04419417306780815, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.234375, "rewards/mcq_exact_match_reward/std": 0.42695629596710205, "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": 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": 0.109375, "frac_reward_zero_std": 0.625, "grad_norm": 24.922893524169922, "learning_rate": 5.741318238559209e-07, "loss": -0.0, "num_tokens": 5303235.0, "reward": 0.5499999523162842, "reward_std": 0.1462520956993103, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "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": 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": 0.11160714285714286, "frac_reward_zero_std": 0.5, "grad_norm": 45.630210876464844, "learning_rate": 5.577423184847931e-07, "loss": -0.0, "num_tokens": 5388987.0, "reward": 0.5499999523162842, "reward_std": 0.19727616012096405, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.0, "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": 5388987, "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 }