{ "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": 685.0, "completions/max_terminated_length": 685.0, "completions/mean_length": 222.625, "completions/mean_terminated_length": 222.625, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0025, "frac_reward_zero_std": 0.0, "grad_norm": 59.533512115478516, "learning_rate": 0.0, "loss": 0.0, "num_tokens": 104344.0, "reward": 0.5445312261581421, "reward_std": 0.47544848918914795, "rewards/format_reward/mean": 0.2890625, "rewards/format_reward/std": 0.2789533734321594, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 1 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1001.0, "completions/max_terminated_length": 1001.0, "completions/mean_length": 286.765625, "completions/mean_terminated_length": 286.765625, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.005, "frac_reward_zero_std": 0.25, "grad_norm": 14.809494972229004, "learning_rate": 2e-07, "loss": -0.0, "num_tokens": 208097.0, "reward": 0.4679687023162842, "reward_std": 0.25451505184173584, "rewards/format_reward/mean": 0.3046875, "rewards/format_reward/std": 0.2615155577659607, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 1127.0, "completions/max_terminated_length": 1127.0, "completions/mean_length": 319.671875, "completions/mean_terminated_length": 319.671875, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0075, "frac_reward_zero_std": 0.0, "grad_norm": 8.160704612731934, "learning_rate": 4e-07, "loss": -0.0, "num_tokens": 305484.0, "reward": 0.5460937023162842, "reward_std": 0.2842584550380707, "rewards/format_reward/mean": 0.3046875, "rewards/format_reward/std": 0.32874447107315063, "rewards/mcq_exact_match_reward/mean": 0.515625, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 3 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1389.0, "completions/max_terminated_length": 1389.0, "completions/mean_length": 244.53125, "completions/mean_terminated_length": 244.53125, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.01, "frac_reward_zero_std": 0.0, "grad_norm": 8.07497787475586, "learning_rate": 6e-07, "loss": -0.0, "num_tokens": 420206.0, "reward": 0.3843749761581421, "reward_std": 0.36569860577583313, "rewards/format_reward/mean": 0.25, "rewards/format_reward/std": 0.28171807527542114, "rewards/mcq_exact_match_reward/mean": 0.359375, "rewards/mcq_exact_match_reward/std": 0.4836103618144989, "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": 1638.0, "completions/max_terminated_length": 1638.0, "completions/mean_length": 349.03125, "completions/mean_terminated_length": 349.03125, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "epoch": 0.0125, "frac_reward_zero_std": 0.0, "grad_norm": 25.123981475830078, "learning_rate": 8e-07, "loss": -0.0, "num_tokens": 551312.0, "reward": 0.2671874761581421, "reward_std": 0.34218358993530273, "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": 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": 0.0, "completions/max_length": 1693.0, "completions/max_terminated_length": 1693.0, "completions/mean_length": 357.921875, "completions/mean_terminated_length": 357.921875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.015, "frac_reward_zero_std": 0.125, "grad_norm": 12.899702072143555, "learning_rate": 1e-06, "loss": -0.0, "num_tokens": 670235.0, "reward": 0.45624998211860657, "reward_std": 0.40883612632751465, "rewards/format_reward/mean": 0.34375, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1128.0, "completions/max_terminated_length": 1128.0, "completions/mean_length": 248.90625, "completions/mean_terminated_length": 248.90625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0175, "frac_reward_zero_std": 0.0, "grad_norm": 9.06348705291748, "learning_rate": 9.99726628670463e-07, "loss": 0.0, "num_tokens": 786077.0, "reward": 0.23671872913837433, "reward_std": 0.3678078055381775, "rewards/format_reward/mean": 0.3359375, "rewards/format_reward/std": 0.25283610820770264, "rewards/mcq_exact_match_reward/mean": 0.203125, "rewards/mcq_exact_match_reward/std": 0.40550529956817627, "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": 1530.0, "completions/max_terminated_length": 1530.0, "completions/mean_length": 493.984375, "completions/mean_terminated_length": 493.984375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.02, "frac_reward_zero_std": 0.0, "grad_norm": 3.943413734436035, "learning_rate": 9.989068136093872e-07, "loss": -0.0, "num_tokens": 897284.0, "reward": 0.296875, "reward_std": 0.45596975088119507, "rewards/format_reward/mean": 0.3125, "rewards/format_reward/std": 0.2745848298072815, "rewards/mcq_exact_match_reward/mean": 0.265625, "rewards/mcq_exact_match_reward/std": 0.44515693187713623, "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": 1452.0, "completions/max_terminated_length": 1452.0, "completions/mean_length": 219.5625, "completions/mean_terminated_length": 219.5625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0225, "frac_reward_zero_std": 0.0, "grad_norm": 15.49644947052002, "learning_rate": 9.975414512725056e-07, "loss": -0.0, "num_tokens": 994448.0, "reward": 0.38515621423721313, "reward_std": 0.35887449979782104, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.209963858127594, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.0, "completions/max_length": 1492.0, "completions/max_terminated_length": 1492.0, "completions/mean_length": 414.5625, "completions/mean_terminated_length": 414.5625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.025, "frac_reward_zero_std": 0.0, "grad_norm": 17.415740966796875, "learning_rate": 9.956320346634875e-07, "loss": 0.0, "num_tokens": 1119476.0, "reward": 0.44140625, "reward_std": 0.4423086643218994, "rewards/format_reward/mean": 0.3515625, "rewards/format_reward/std": 0.3044600784778595, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 939.0, "completions/max_terminated_length": 939.0, "completions/mean_length": 261.59375, "completions/mean_terminated_length": 261.59375, "completions/min_length": 3.0, "completions/min_terminated_length": 3.0, "epoch": 0.0275, "frac_reward_zero_std": 0.0, "grad_norm": 21.028289794921875, "learning_rate": 9.931806517013612e-07, "loss": 0.0, "num_tokens": 1218250.0, "reward": 0.3359375, "reward_std": 0.40134403109550476, "rewards/format_reward/mean": 0.390625, "rewards/format_reward/std": 0.3145764470100403, "rewards/mcq_exact_match_reward/mean": 0.296875, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "step": 11 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1675.0, "completions/max_terminated_length": 1675.0, "completions/mean_length": 503.96875, "completions/mean_terminated_length": 503.96875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.03, "frac_reward_zero_std": 0.0, "grad_norm": 5.3941802978515625, "learning_rate": 9.901899829374047e-07, "loss": 0.0, "num_tokens": 1353544.0, "reward": 0.32343748211860657, "reward_std": 0.39192718267440796, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.3238992393016815, "rewards/mcq_exact_match_reward/mean": 0.28125, "rewards/mcq_exact_match_reward/std": 0.4531635046005249, "step": 12 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 2036.0, "completions/mean_length": 615.359375, "completions/mean_terminated_length": 569.1451416015625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0325, "frac_reward_zero_std": 0.25, "grad_norm": 3.4431583881378174, "learning_rate": 9.866632986240029e-07, "loss": -0.0, "num_tokens": 1481711.0, "reward": 0.29140621423721313, "reward_std": 0.3188745379447937, "rewards/format_reward/mean": 0.4140625, "rewards/format_reward/std": 0.209963858127594, "rewards/mcq_exact_match_reward/mean": 0.25, "rewards/mcq_exact_match_reward/std": 0.4364357888698578, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1428.0, "completions/mean_length": 489.390625, "completions/mean_terminated_length": 439.1128845214844, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.035, "frac_reward_zero_std": 0.0, "grad_norm": 4.944615840911865, "learning_rate": 9.826044551386742e-07, "loss": -0.0, "num_tokens": 1601000.0, "reward": 0.3140624761581421, "reward_std": 0.3419710397720337, "rewards/format_reward/mean": 0.328125, "rewards/format_reward/std": 0.3238992393016815, "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": 1766.0, "completions/mean_length": 571.1875, "completions/mean_terminated_length": 498.5573425292969, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0375, "frac_reward_zero_std": 0.0, "grad_norm": 3.2745373249053955, "learning_rate": 9.780178907671788e-07, "loss": -0.0, "num_tokens": 1731604.0, "reward": 0.2867187261581421, "reward_std": 0.40185993909835815, "rewards/format_reward/mean": 0.3671875, "rewards/format_reward/std": 0.23974503576755524, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1636.0, "completions/mean_length": 601.015625, "completions/mean_terminated_length": 578.0476684570312, "completions/min_length": 17.0, "completions/min_terminated_length": 17.0, "epoch": 0.04, "frac_reward_zero_std": 0.0, "grad_norm": 2.9144675731658936, "learning_rate": 9.729086208503173e-07, "loss": -0.0, "num_tokens": 1859949.0, "reward": 0.38749998807907104, "reward_std": 0.43986088037490845, "rewards/format_reward/mean": 0.4375, "rewards/format_reward/std": 0.2597312331199646, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1056.0, "completions/mean_length": 380.1875, "completions/mean_terminated_length": 353.71429443359375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0425, "frac_reward_zero_std": 0.125, "grad_norm": 1.8824231624603271, "learning_rate": 9.672822322997304e-07, "loss": -0.0, "num_tokens": 1970809.0, "reward": 0.567187488079071, "reward_std": 0.4544362723827362, "rewards/format_reward/mean": 0.359375, "rewards/format_reward/std": 0.28824523091316223, "rewards/mcq_exact_match_reward/mean": 0.53125, "rewards/mcq_exact_match_reward/std": 0.5029674172401428, "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": 1673.0, "completions/max_terminated_length": 1673.0, "completions/mean_length": 516.09375, "completions/mean_terminated_length": 516.09375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.045, "frac_reward_zero_std": 0.0, "grad_norm": 2.5847556591033936, "learning_rate": 9.611448774886923e-07, "loss": -0.0, "num_tokens": 2115567.0, "reward": 0.4624999761581421, "reward_std": 0.33887431025505066, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "step": 18 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1511.0, "completions/max_terminated_length": 1511.0, "completions/mean_length": 406.84375, "completions/mean_terminated_length": 406.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0475, "frac_reward_zero_std": 0.125, "grad_norm": 1.9844298362731934, "learning_rate": 9.545032675245813e-07, "loss": -0.0, "num_tokens": 2224709.0, "reward": 0.4781249761581421, "reward_std": 0.3849598169326782, "rewards/format_reward/mean": 0.40625, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1541.0, "completions/mean_length": 444.796875, "completions/mean_terminated_length": 419.3492431640625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.05, "frac_reward_zero_std": 0.0, "grad_norm": 6.835641384124756, "learning_rate": 9.473646649103817e-07, "loss": -0.0, "num_tokens": 2350360.0, "reward": 0.46406248211860657, "reward_std": 0.3951150178909302, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.25539806485176086, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1713.0, "completions/max_terminated_length": 1713.0, "completions/mean_length": 531.78125, "completions/mean_terminated_length": 531.78125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0525, "frac_reward_zero_std": 0.25, "grad_norm": 2.273691177368164, "learning_rate": 9.397368756032444e-07, "loss": 0.0, "num_tokens": 2476410.0, "reward": 0.4195312261581421, "reward_std": 0.19411513209342957, "rewards/format_reward/mean": 0.4453125, "rewards/format_reward/std": 0.2376670390367508, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 1147.0, "completions/max_terminated_length": 1147.0, "completions/mean_length": 415.796875, "completions/mean_terminated_length": 415.796875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.055, "frac_reward_zero_std": 0.0, "grad_norm": 5.067054748535156, "learning_rate": 9.316282404787869e-07, "loss": -0.0, "num_tokens": 2610701.0, "reward": 0.37109375, "reward_std": 0.4409857988357544, "rewards/format_reward/mean": 0.4296875, "rewards/format_reward/std": 0.26528194546699524, "rewards/mcq_exact_match_reward/mean": 0.328125, "rewards/mcq_exact_match_reward/std": 0.4732423722743988, "step": 22 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 1403.0, "completions/max_terminated_length": 1403.0, "completions/mean_length": 364.1875, "completions/mean_terminated_length": 364.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0575, "frac_reward_zero_std": 0.0, "grad_norm": 13.165079116821289, "learning_rate": 9.230476262104676e-07, "loss": -0.0, "num_tokens": 2738369.0, "reward": 0.6960936784744263, "reward_std": 0.44959956407546997, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.22146137058734894, "rewards/mcq_exact_match_reward/mean": 0.65625, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 1887.0, "completions/max_terminated_length": 1887.0, "completions/mean_length": 396.875, "completions/mean_terminated_length": 396.875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.06, "frac_reward_zero_std": 0.0, "grad_norm": 5.967652797698975, "learning_rate": 9.1400441557401e-07, "loss": -0.0, "num_tokens": 2853985.0, "reward": 0.49296873807907104, "reward_std": 0.4339084029197693, "rewards/format_reward/mean": 0.3984375, "rewards/format_reward/std": 0.20275264978408813, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 1321.0, "completions/mean_length": 527.609375, "completions/mean_terminated_length": 503.4762268066406, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0625, "frac_reward_zero_std": 0.0, "grad_norm": 5.454898357391357, "learning_rate": 9.045084971874737e-07, "loss": -0.0, "num_tokens": 2987496.0, "reward": 0.3882812261581421, "reward_std": 0.4757910370826721, "rewards/format_reward/mean": 0.4453125, "rewards/format_reward/std": 0.26899561285972595, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "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": 1604.0, "completions/max_terminated_length": 1604.0, "completions/mean_length": 433.25, "completions/mean_terminated_length": 433.25, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.065, "frac_reward_zero_std": 0.0, "grad_norm": 7.106506824493408, "learning_rate": 8.945702546981968e-07, "loss": -0.0, "num_tokens": 3120200.0, "reward": 0.47968748211860657, "reward_std": 0.47255939245224, "rewards/format_reward/mean": 0.421875, "rewards/format_reward/std": 0.23935678601264954, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 1906.0, "completions/max_terminated_length": 1906.0, "completions/mean_length": 574.765625, "completions/mean_terminated_length": 574.765625, "completions/min_length": 90.0, "completions/min_terminated_length": 90.0, "epoch": 0.0675, "frac_reward_zero_std": 0.125, "grad_norm": 1.6191740036010742, "learning_rate": 8.842005554284295e-07, "loss": -0.0, "num_tokens": 3249513.0, "reward": 0.4820312261581421, "reward_std": 0.4589237570762634, "rewards/format_reward/mean": 0.4453125, "rewards/format_reward/std": 0.2203386276960373, "rewards/mcq_exact_match_reward/mean": 0.4375, "rewards/mcq_exact_match_reward/std": 0.5, "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": 1529.0, "completions/mean_length": 472.84375, "completions/mean_terminated_length": 447.84130859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.07, "frac_reward_zero_std": 0.25, "grad_norm": 3.762904167175293, "learning_rate": 8.734107384920769e-07, "loss": -0.0, "num_tokens": 3379007.0, "reward": 0.3570312261581421, "reward_std": 0.2997973561286926, "rewards/format_reward/mean": 0.4453125, "rewards/format_reward/std": 0.20152568817138672, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 1093.0, "completions/mean_length": 472.203125, "completions/mean_terminated_length": 447.19049072265625, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.0725, "frac_reward_zero_std": 0.0, "grad_norm": 10.816900253295898, "learning_rate": 8.622126023955445e-07, "loss": -0.0, "num_tokens": 3493564.0, "reward": 0.5625, "reward_std": 0.4124056100845337, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.265398770570755, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1644.0, "completions/mean_length": 378.25, "completions/mean_terminated_length": 351.7460632324219, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.075, "frac_reward_zero_std": 0.25, "grad_norm": 6.790505886077881, "learning_rate": 8.506183921362442e-07, "loss": 0.0, "num_tokens": 3593036.0, "reward": 0.703125, "reward_std": 0.25315535068511963, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.65625, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1901.0, "completions/mean_length": 449.421875, "completions/mean_terminated_length": 424.0476379394531, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0775, "frac_reward_zero_std": 0.125, "grad_norm": 6.935174465179443, "learning_rate": 8.386407858128706e-07, "loss": 0.0, "num_tokens": 3710671.0, "reward": 0.40625, "reward_std": 0.29695966839790344, "rewards/format_reward/mean": 0.46875, "rewards/format_reward/std": 0.19669894874095917, "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": 1866.0, "completions/max_terminated_length": 1866.0, "completions/mean_length": 242.046875, "completions/mean_terminated_length": 242.046875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.08, "frac_reward_zero_std": 0.125, "grad_norm": 11.75405502319336, "learning_rate": 8.262928807620843e-07, "loss": -0.0, "num_tokens": 3812746.0, "reward": 0.6046874523162842, "reward_std": 0.3665615916252136, "rewards/format_reward/mean": 0.578125, "rewards/format_reward/std": 0.2847827076911926, "rewards/mcq_exact_match_reward/mean": 0.546875, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "step": 32 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1017.0, "completions/mean_length": 385.109375, "completions/mean_terminated_length": 331.4677429199219, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0825, "frac_reward_zero_std": 0.125, "grad_norm": 9.297252655029297, "learning_rate": 8.135881792367685e-07, "loss": -0.0, "num_tokens": 3935601.0, "reward": 0.36015623807907104, "reward_std": 0.3222334682941437, "rewards/format_reward/mean": 0.4765625, "rewards/format_reward/std": 0.25865477323532104, "rewards/mcq_exact_match_reward/mean": 0.3125, "rewards/mcq_exact_match_reward/std": 0.467176616191864, "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": 0.03125, "completions/max_length": 2048.0, "completions/max_terminated_length": 1522.0, "completions/mean_length": 492.140625, "completions/mean_terminated_length": 441.95159912109375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.085, "frac_reward_zero_std": 0.125, "grad_norm": 9.278347969055176, "learning_rate": 8.005405736415125e-07, "loss": -0.0, "num_tokens": 4055274.0, "reward": 0.4937499761581421, "reward_std": 0.3744146227836609, "rewards/format_reward/mean": 0.5625, "rewards/format_reward/std": 0.22712838649749756, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1146.0, "completions/mean_length": 302.875, "completions/mean_terminated_length": 275.17462158203125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0875, "frac_reward_zero_std": 0.25, "grad_norm": 5.606284141540527, "learning_rate": 7.871643313414718e-07, "loss": -0.0, "num_tokens": 4176362.0, "reward": 0.47187501192092896, "reward_std": 0.24273529648780823, "rewards/format_reward/mean": 0.5, "rewards/format_reward/std": 0.34503278136253357, "rewards/mcq_exact_match_reward/mean": 0.421875, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1482.0, "completions/max_terminated_length": 1482.0, "completions/mean_length": 410.40625, "completions/mean_terminated_length": 410.40625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.09, "frac_reward_zero_std": 0.125, "grad_norm": 9.850747108459473, "learning_rate": 7.734740790612136e-07, "loss": 0.0, "num_tokens": 4292476.0, "reward": 0.6898437738418579, "reward_std": 0.22759628295898438, "rewards/format_reward/mean": 0.6484375, "rewards/format_reward/std": 0.29113471508026123, "rewards/mcq_exact_match_reward/mean": 0.625, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 1006.0, "completions/max_terminated_length": 1006.0, "completions/mean_length": 126.859375, "completions/mean_terminated_length": 126.859375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0925, "frac_reward_zero_std": 0.125, "grad_norm": 20.594633102416992, "learning_rate": 7.594847868906076e-07, "loss": 0.0, "num_tokens": 4413923.0, "reward": 0.5546875, "reward_std": 0.30506113171577454, "rewards/format_reward/mean": 0.703125, "rewards/format_reward/std": 0.2777281701564789, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "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": 1447.0, "completions/max_terminated_length": 1447.0, "completions/mean_length": 381.84375, "completions/mean_terminated_length": 381.84375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.095, "frac_reward_zero_std": 0.125, "grad_norm": 4.137030124664307, "learning_rate": 7.452117519152541e-07, "loss": -0.0, "num_tokens": 4531905.0, "reward": 0.44062501192092896, "reward_std": 0.25260162353515625, "rewards/format_reward/mean": 0.65625, "rewards/format_reward/std": 0.265398770570755, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1119.0, "completions/mean_length": 270.359375, "completions/mean_terminated_length": 242.1428680419922, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.0975, "frac_reward_zero_std": 0.125, "grad_norm": 20.815427780151367, "learning_rate": 7.306705814893439e-07, "loss": 0.0, "num_tokens": 4643792.0, "reward": 0.41874998807907104, "reward_std": 0.4047183394432068, "rewards/format_reward/mean": 0.75, "rewards/format_reward/std": 0.29546841979026794, "rewards/mcq_exact_match_reward/mean": 0.34375, "rewards/mcq_exact_match_reward/std": 0.4787135720252991, "step": 39 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.015625, "completions/max_length": 2048.0, "completions/max_terminated_length": 1158.0, "completions/mean_length": 153.5625, "completions/mean_terminated_length": 123.49207305908203, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1, "frac_reward_zero_std": 0.125, "grad_norm": 16.95915412902832, "learning_rate": 7.158771761692464e-07, "loss": 0.0, "num_tokens": 4758172.0, "reward": 0.5648437738418579, "reward_std": 0.20101891458034515, "rewards/format_reward/mean": 0.8046875, "rewards/format_reward/std": 0.2615155577659607, "rewards/mcq_exact_match_reward/mean": 0.484375, "rewards/mcq_exact_match_reward/std": 0.5037065148353577, "step": 40 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0, "completions/max_length": 604.0, "completions/max_terminated_length": 604.0, "completions/mean_length": 59.65625, "completions/mean_terminated_length": 59.65625, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1025, "frac_reward_zero_std": 0.25, "grad_norm": 19.127452850341797, "learning_rate": 7.008477123264847e-07, "loss": -0.0, "num_tokens": 4861310.0, "reward": 0.48750001192092896, "reward_std": 0.32531416416168213, "rewards/format_reward/mean": 0.8125, "rewards/format_reward/std": 0.24397502839565277, "rewards/mcq_exact_match_reward/mean": 0.40625, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 126.0, "completions/max_terminated_length": 126.0, "completions/mean_length": 18.4375, "completions/mean_terminated_length": 18.4375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.105, "frac_reward_zero_std": 0.375, "grad_norm": 19.860843658447266, "learning_rate": 6.855986244591103e-07, "loss": 0.0, "num_tokens": 4955818.0, "reward": 0.6742187738418579, "reward_std": 0.18233226239681244, "rewards/format_reward/mean": 0.8046875, "rewards/format_reward/std": 0.2615155577659607, "rewards/mcq_exact_match_reward/mean": 0.59375, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "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": 522.0, "completions/max_terminated_length": 522.0, "completions/mean_length": 38.796875, "completions/mean_terminated_length": 38.796875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1075, "frac_reward_zero_std": 0.5, "grad_norm": 11.83443832397461, "learning_rate": 6.701465872208216e-07, "loss": 0.0, "num_tokens": 5052005.0, "reward": 0.5335937738418579, "reward_std": 0.2508957087993622, "rewards/format_reward/mean": 0.8046875, "rewards/format_reward/std": 0.24587368965148926, "rewards/mcq_exact_match_reward/mean": 0.453125, "rewards/mcq_exact_match_reward/std": 0.501733124256134, "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": 865.0, "completions/max_terminated_length": 865.0, "completions/mean_length": 102.421875, "completions/mean_terminated_length": 102.421875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.11, "frac_reward_zero_std": 0.375, "grad_norm": 7.313492774963379, "learning_rate": 6.545084971874736e-07, "loss": 0.0, "num_tokens": 5165008.0, "reward": 0.6687500476837158, "reward_std": 0.2736644446849823, "rewards/format_reward/mean": 0.90625, "rewards/format_reward/std": 0.233588308095932, "rewards/mcq_exact_match_reward/mean": 0.578125, "rewards/mcq_exact_match_reward/std": 0.49776285886764526, "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": 1833.0, "completions/max_terminated_length": 1833.0, "completions/mean_length": 95.1875, "completions/mean_terminated_length": 95.1875, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1125, "frac_reward_zero_std": 0.5, "grad_norm": 20.023258209228516, "learning_rate": 6.387014543809223e-07, "loss": 0.0, "num_tokens": 5266132.0, "reward": 0.47265625, "reward_std": 0.1391630470752716, "rewards/format_reward/mean": 0.8203125, "rewards/format_reward/std": 0.32571274042129517, "rewards/mcq_exact_match_reward/mean": 0.390625, "rewards/mcq_exact_match_reward/std": 0.4917473793029785, "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": 1079.0, "completions/max_terminated_length": 1079.0, "completions/mean_length": 71.578125, "completions/mean_terminated_length": 71.578125, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.115, "frac_reward_zero_std": 0.25, "grad_norm": 19.259967803955078, "learning_rate": 6.227427435703995e-07, "loss": 0.0, "num_tokens": 5373785.0, "reward": 0.5859375, "reward_std": 0.34790486097335815, "rewards/format_reward/mean": 0.859375, "rewards/format_reward/std": 0.2741328477859497, "rewards/mcq_exact_match_reward/mean": 0.5, "rewards/mcq_exact_match_reward/std": 0.5039526224136353, "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": 947.0, "completions/max_terminated_length": 947.0, "completions/mean_length": 128.34375, "completions/mean_terminated_length": 128.34375, "completions/min_length": 6.0, "completions/min_terminated_length": 6.0, "epoch": 0.1175, "frac_reward_zero_std": 0.375, "grad_norm": 28.760831832885742, "learning_rate": 6.066498153718734e-07, "loss": -0.0, "num_tokens": 5500999.0, "reward": 0.557812511920929, "reward_std": 0.24416987597942352, "rewards/format_reward/mean": 0.890625, "rewards/format_reward/std": 0.2083333432674408, "rewards/mcq_exact_match_reward/mean": 0.46875, "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": 638.0, "completions/max_terminated_length": 638.0, "completions/mean_length": 38.078125, "completions/mean_terminated_length": 38.078125, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.12, "frac_reward_zero_std": 0.375, "grad_norm": 8.37263298034668, "learning_rate": 5.90440267166055e-07, "loss": -0.0, "num_tokens": 5585012.0, "reward": 0.47421878576278687, "reward_std": 0.20632557570934296, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.375, "rewards/mcq_exact_match_reward/std": 0.48795005679130554, "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": 533.0, "completions/max_terminated_length": 533.0, "completions/mean_length": 21.296875, "completions/mean_terminated_length": 21.296875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.1225, "frac_reward_zero_std": 0.625, "grad_norm": 15.718689918518066, "learning_rate": 5.741318238559209e-07, "loss": 0.0, "num_tokens": 5652815.0, "reward": 0.8023437857627869, "reward_std": 0.16939528286457062, "rewards/format_reward/mean": 0.9921875, "rewards/format_reward/std": 0.0625, "rewards/mcq_exact_match_reward/mean": 0.703125, "rewards/mcq_exact_match_reward/std": 0.4604927599430084, "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": 396.0, "completions/max_terminated_length": 396.0, "completions/mean_length": 34.1875, "completions/mean_terminated_length": 34.1875, "completions/min_length": 12.0, "completions/min_terminated_length": 12.0, "epoch": 0.125, "frac_reward_zero_std": 0.625, "grad_norm": 7.805715084075928, "learning_rate": 5.577423184847931e-07, "loss": 0.0, "num_tokens": 5754403.0, "reward": 0.69140625, "reward_std": 0.18080364167690277, "rewards/format_reward/mean": 0.9765625, "rewards/format_reward/std": 0.13886408507823944, "rewards/mcq_exact_match_reward/mean": 0.59375, "rewards/mcq_exact_match_reward/std": 0.49501484632492065, "step": 50 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 5754403, "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 }