{
  "best_global_step": 600,
  "best_metric": 0.10775862068965517,
  "best_model_checkpoint": "/local2/asuvarna31/distractors/Qwen3-0.6B_Mar18-0256_qwen3_0.6b_micro_top2_Mar18-0250_10000/checkpoint-600",
  "epoch": 0.12,
  "eval_steps": 100,
  "global_step": 600,
  "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": 0.0475,
      "completions/max_length": 1907.96,
      "completions/max_terminated_length": 1542.92,
      "completions/mean_length": 794.835,
      "completions/mean_terminated_length": 635.4152294921874,
      "completions/min_length": 231.32,
      "completions/min_terminated_length": 231.32,
      "entropy": 0.39757278442382815,
      "epoch": 0.005,
      "frac_reward_zero_std": 0.26,
      "grad_norm": 0.7908472990123143,
      "learning_rate": 9.952e-07,
      "loss": -0.0219,
      "num_tokens": 370918.0,
      "reward": 0.37,
      "reward_std": 0.42105737447738645,
      "rewards/combined_reward/mean": 0.37,
      "rewards/combined_reward/std": 0.4210573983192444,
      "sampling/importance_sampling_ratio/max": 2.198445258140564,
      "sampling/importance_sampling_ratio/mean": 0.6045227408409118,
      "sampling/importance_sampling_ratio/min": 0.022674707409983057,
      "sampling/sampling_logp_difference/max": 0.698764111995697,
      "sampling/sampling_logp_difference/mean": 0.02563132591545582,
      "step": 25,
      "step_time": 7.23048153122887
    },
    {
      "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": 0.0225,
      "completions/max_length": 1275.04,
      "completions/max_terminated_length": 1168.96,
      "completions/mean_length": 563.5925,
      "completions/mean_terminated_length": 512.9527490234375,
      "completions/min_length": 240.0,
      "completions/min_terminated_length": 240.0,
      "entropy": 0.41257245182991026,
      "epoch": 0.01,
      "frac_reward_zero_std": 0.34,
      "grad_norm": 1.1362625413839582,
      "learning_rate": 9.901999999999999e-07,
      "loss": 0.0233,
      "num_tokens": 649035.0,
      "reward": 0.5125,
      "reward_std": 0.3963914489746094,
      "rewards/combined_reward/mean": 0.5125,
      "rewards/combined_reward/std": 0.39639145612716675,
      "sampling/importance_sampling_ratio/max": 2.1132223176956177,
      "sampling/importance_sampling_ratio/mean": 0.6526571130752563,
      "sampling/importance_sampling_ratio/min": 0.029447911735951494,
      "sampling/sampling_logp_difference/max": 0.6200782227516174,
      "sampling/sampling_logp_difference/mean": 0.026439733169972897,
      "step": 50,
      "step_time": 4.926513614905998
    },
    {
      "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": 0.0575,
      "completions/max_length": 2291.2,
      "completions/max_terminated_length": 1943.84,
      "completions/mean_length": 961.34,
      "completions/mean_terminated_length": 788.9747546386719,
      "completions/min_length": 298.44,
      "completions/min_terminated_length": 298.44,
      "entropy": 0.38926577985286714,
      "epoch": 0.015,
      "frac_reward_zero_std": 0.24,
      "grad_norm": 1.8509531292907888,
      "learning_rate": 9.852e-07,
      "loss": 0.0555,
      "num_tokens": 1089027.0,
      "reward": 0.535,
      "reward_std": 0.44107813596725465,
      "rewards/combined_reward/mean": 0.535,
      "rewards/combined_reward/std": 0.44107815861701966,
      "sampling/importance_sampling_ratio/max": 1.9247741317749023,
      "sampling/importance_sampling_ratio/mean": 0.593499310016632,
      "sampling/importance_sampling_ratio/min": 0.025702737645042363,
      "sampling/sampling_logp_difference/max": 0.7170420527458191,
      "sampling/sampling_logp_difference/mean": 0.02496741436421871,
      "step": 75,
      "step_time": 8.62744049324654
    },
    {
      "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": 0.03,
      "completions/max_length": 1365.76,
      "completions/max_terminated_length": 1239.48,
      "completions/mean_length": 614.215,
      "completions/mean_terminated_length": 525.4255004882813,
      "completions/min_length": 196.08,
      "completions/min_terminated_length": 196.08,
      "entropy": 0.42757334232330324,
      "epoch": 0.02,
      "frac_reward_zero_std": 0.32,
      "grad_norm": 2.7107245292264266,
      "learning_rate": 9.802e-07,
      "loss": -0.016,
      "num_tokens": 1389617.0,
      "reward": 0.625,
      "reward_std": 0.4001571249961853,
      "rewards/combined_reward/mean": 0.625,
      "rewards/combined_reward/std": 0.400157151222229,
      "sampling/importance_sampling_ratio/max": 2.0216542720794677,
      "sampling/importance_sampling_ratio/mean": 0.6695219826698303,
      "sampling/importance_sampling_ratio/min": 0.031132260655103893,
      "sampling/sampling_logp_difference/max": 0.7458893251419068,
      "sampling/sampling_logp_difference/mean": 0.027404132559895517,
      "step": 100,
      "step_time": 5.374540434898808
    },
    {
      "epoch": 0.02,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.3599137931034483,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3399.1724137931033,
      "eval_bbeh_mini_completions/mean_length": 2556.969827586207,
      "eval_bbeh_mini_completions/mean_terminated_length": 1690.4230199353449,
      "eval_bbeh_mini_completions/min_length": 389.13793103448273,
      "eval_bbeh_mini_completions/min_terminated_length": 389.13793103448273,
      "eval_bbeh_mini_entropy": 0.33320264775177527,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 1389617.0,
      "eval_bbeh_mini_reward": 0.0625,
      "eval_bbeh_mini_reward_std": 0.1932970984228726,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.0625,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.19329710458887034,
      "eval_bbeh_mini_runtime": 493.3587,
      "eval_bbeh_mini_samples_per_second": 0.932,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.5544916828130853,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.24806074290696917,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 4.2029642265184464e-05,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 3.1536410635915297,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.022588976112932993,
      "eval_bbeh_mini_steps_per_second": 0.059,
      "step": 100
    },
    {
      "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": 0.0925,
      "completions/max_length": 2125.76,
      "completions/max_terminated_length": 1810.64,
      "completions/mean_length": 1089.9775,
      "completions/mean_terminated_length": 842.0292138671875,
      "completions/min_length": 302.92,
      "completions/min_terminated_length": 302.92,
      "entropy": 0.381637504696846,
      "epoch": 0.025,
      "frac_reward_zero_std": 0.44,
      "grad_norm": 0.8500470780968633,
      "learning_rate": 9.752e-07,
      "loss": 0.0054,
      "num_tokens": 1879680.0,
      "reward": 0.5525,
      "reward_std": 0.3752482628822327,
      "rewards/combined_reward/mean": 0.5525,
      "rewards/combined_reward/std": 0.3752482759952545,
      "sampling/importance_sampling_ratio/max": 2.1725155782699583,
      "sampling/importance_sampling_ratio/mean": 0.5781891107559204,
      "sampling/importance_sampling_ratio/min": 0.014303099608951016,
      "sampling/sampling_logp_difference/max": 0.7780866742134094,
      "sampling/sampling_logp_difference/mean": 0.02495789147913456,
      "step": 125,
      "step_time": 8.32747592881322
    },
    {
      "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": 0.025,
      "completions/max_length": 1443.16,
      "completions/max_terminated_length": 1350.84,
      "completions/mean_length": 602.2575,
      "completions/mean_terminated_length": 526.3002880859375,
      "completions/min_length": 181.92,
      "completions/min_terminated_length": 181.92,
      "entropy": 0.3935022366046905,
      "epoch": 0.03,
      "frac_reward_zero_std": 0.46,
      "grad_norm": 3.9983544495385286,
      "learning_rate": 9.701999999999998e-07,
      "loss": -0.0266,
      "num_tokens": 2173791.0,
      "reward": 0.675,
      "reward_std": 0.4132915163040161,
      "rewards/combined_reward/mean": 0.675,
      "rewards/combined_reward/std": 0.41329152703285216,
      "sampling/importance_sampling_ratio/max": 1.974114351272583,
      "sampling/importance_sampling_ratio/mean": 0.6311233282089234,
      "sampling/importance_sampling_ratio/min": 0.03929259464435745,
      "sampling/sampling_logp_difference/max": 0.6166027712821961,
      "sampling/sampling_logp_difference/mean": 0.02564044661819935,
      "step": 150,
      "step_time": 5.59985625769943
    },
    {
      "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": 0.0475,
      "completions/max_length": 1882.6,
      "completions/max_terminated_length": 1692.64,
      "completions/mean_length": 824.66,
      "completions/mean_terminated_length": 694.9667797851563,
      "completions/min_length": 231.4,
      "completions/min_terminated_length": 231.4,
      "entropy": 0.38833099365234375,
      "epoch": 0.035,
      "frac_reward_zero_std": 0.44,
      "grad_norm": 0.0,
      "learning_rate": 9.651999999999999e-07,
      "loss": -0.0042,
      "num_tokens": 2556495.0,
      "reward": 0.615,
      "reward_std": 0.3986561942100525,
      "rewards/combined_reward/mean": 0.615,
      "rewards/combined_reward/std": 0.39865620732307433,
      "sampling/importance_sampling_ratio/max": 2.0362929487228394,
      "sampling/importance_sampling_ratio/mean": 0.6166802150011063,
      "sampling/importance_sampling_ratio/min": 0.03415079687081743,
      "sampling/sampling_logp_difference/max": 0.9840151739120483,
      "sampling/sampling_logp_difference/mean": 0.025189833119511604,
      "step": 175,
      "step_time": 7.186449560681358
    },
    {
      "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": 0.0475,
      "completions/max_length": 2255.96,
      "completions/max_terminated_length": 1958.16,
      "completions/mean_length": 876.6025,
      "completions/mean_terminated_length": 734.1006811523438,
      "completions/min_length": 209.52,
      "completions/min_terminated_length": 209.52,
      "entropy": 0.3978144943714142,
      "epoch": 0.04,
      "frac_reward_zero_std": 0.34,
      "grad_norm": 2.407379133784635,
      "learning_rate": 9.602e-07,
      "loss": 0.0184,
      "num_tokens": 2959120.0,
      "reward": 0.5975,
      "reward_std": 0.4489552092552185,
      "rewards/combined_reward/mean": 0.5975,
      "rewards/combined_reward/std": 0.4489552354812622,
      "sampling/importance_sampling_ratio/max": 1.9040312719345094,
      "sampling/importance_sampling_ratio/mean": 0.5416174274682999,
      "sampling/importance_sampling_ratio/min": 0.018405352871644087,
      "sampling/sampling_logp_difference/max": 0.8134938716888428,
      "sampling/sampling_logp_difference/mean": 0.025685913935303686,
      "step": 200,
      "step_time": 8.387271651169286
    },
    {
      "epoch": 0.04,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.31896551724137934,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3501.3793103448274,
      "eval_bbeh_mini_completions/mean_length": 2493.7543103448274,
      "eval_bbeh_mini_completions/mean_terminated_length": 1744.4683248585668,
      "eval_bbeh_mini_completions/min_length": 402.0689655172414,
      "eval_bbeh_mini_completions/min_terminated_length": 402.0689655172414,
      "eval_bbeh_mini_entropy": 0.33658467079031057,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 2959120.0,
      "eval_bbeh_mini_reward": 0.07327586206896551,
      "eval_bbeh_mini_reward_std": 0.22547399586644665,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.07327586206896551,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.22547400408777699,
      "eval_bbeh_mini_runtime": 485.7912,
      "eval_bbeh_mini_samples_per_second": 0.947,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.4247941641971982,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.21224350610683704,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 1.9089912678375065e-05,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 3.4790474669686677,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.022729232650378656,
      "eval_bbeh_mini_steps_per_second": 0.06,
      "step": 200
    },
    {
      "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": 0.045,
      "completions/max_length": 2272.24,
      "completions/max_terminated_length": 1942.2,
      "completions/mean_length": 935.3925,
      "completions/mean_terminated_length": 812.9290356445313,
      "completions/min_length": 275.64,
      "completions/min_terminated_length": 275.64,
      "entropy": 0.4116176557540894,
      "epoch": 0.045,
      "frac_reward_zero_std": 0.42,
      "grad_norm": 3.5037476412669593,
      "learning_rate": 9.552e-07,
      "loss": 0.0173,
      "num_tokens": 3383261.0,
      "reward": 0.6625,
      "reward_std": 0.41886321067810056,
      "rewards/combined_reward/mean": 0.6625,
      "rewards/combined_reward/std": 0.41886322021484373,
      "sampling/importance_sampling_ratio/max": 2.0795044040679933,
      "sampling/importance_sampling_ratio/mean": 0.6006859976053238,
      "sampling/importance_sampling_ratio/min": 0.014239588570781052,
      "sampling/sampling_logp_difference/max": 0.7338527536392212,
      "sampling/sampling_logp_difference/mean": 0.02595676988363266,
      "step": 225,
      "step_time": 8.487474374789745
    },
    {
      "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": 0.03,
      "completions/max_length": 1640.28,
      "completions/max_terminated_length": 1389.36,
      "completions/mean_length": 653.0075,
      "completions/mean_terminated_length": 565.9887109375,
      "completions/min_length": 255.52,
      "completions/min_terminated_length": 255.52,
      "entropy": 0.4170915710926056,
      "epoch": 0.05,
      "frac_reward_zero_std": 0.42,
      "grad_norm": 2.1638076114867757,
      "learning_rate": 9.502e-07,
      "loss": -0.0148,
      "num_tokens": 3698864.0,
      "reward": 0.705,
      "reward_std": 0.38157126426696775,
      "rewards/combined_reward/mean": 0.705,
      "rewards/combined_reward/std": 0.38157127737998964,
      "sampling/importance_sampling_ratio/max": 1.9969594597816467,
      "sampling/importance_sampling_ratio/mean": 0.628093301653862,
      "sampling/importance_sampling_ratio/min": 0.018454841256149202,
      "sampling/sampling_logp_difference/max": 0.659376904964447,
      "sampling/sampling_logp_difference/mean": 0.026758400201797487,
      "step": 250,
      "step_time": 6.111578326625749
    },
    {
      "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": 0.055,
      "completions/max_length": 2229.44,
      "completions/max_terminated_length": 1948.52,
      "completions/mean_length": 942.01,
      "completions/mean_terminated_length": 788.1107543945312,
      "completions/min_length": 249.4,
      "completions/min_terminated_length": 249.4,
      "entropy": 0.4110789692401886,
      "epoch": 0.055,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 4.667525978474157,
      "learning_rate": 9.452e-07,
      "loss": -0.0242,
      "num_tokens": 4129764.0,
      "reward": 0.6575,
      "reward_std": 0.4181215476989746,
      "rewards/combined_reward/mean": 0.6575,
      "rewards/combined_reward/std": 0.41812155723571776,
      "sampling/importance_sampling_ratio/max": 1.8673054933547975,
      "sampling/importance_sampling_ratio/mean": 0.5125843727588654,
      "sampling/importance_sampling_ratio/min": 0.01608145761529954,
      "sampling/sampling_logp_difference/max": 0.8164897656440735,
      "sampling/sampling_logp_difference/mean": 0.02648505486547947,
      "step": 275,
      "step_time": 8.40903925454244
    },
    {
      "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": 0.04,
      "completions/max_length": 2057.4,
      "completions/max_terminated_length": 1949.44,
      "completions/mean_length": 919.8125,
      "completions/mean_terminated_length": 816.5338427734375,
      "completions/min_length": 261.0,
      "completions/min_terminated_length": 261.0,
      "entropy": 0.41194292187690734,
      "epoch": 0.06,
      "frac_reward_zero_std": 0.42,
      "grad_norm": 1.1934433538781786,
      "learning_rate": 9.402e-07,
      "loss": 0.0465,
      "num_tokens": 4550697.0,
      "reward": 0.6275,
      "reward_std": 0.417396194934845,
      "rewards/combined_reward/mean": 0.6275,
      "rewards/combined_reward/std": 0.41739620923995974,
      "sampling/importance_sampling_ratio/max": 2.1660247945785525,
      "sampling/importance_sampling_ratio/mean": 0.6122340059280396,
      "sampling/importance_sampling_ratio/min": 0.014834024702245853,
      "sampling/sampling_logp_difference/max": 0.8179536938667298,
      "sampling/sampling_logp_difference/mean": 0.025995058193802833,
      "step": 300,
      "step_time": 7.779672881411389
    },
    {
      "epoch": 0.06,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.36637931034482757,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3524.7241379310344,
      "eval_bbeh_mini_completions/mean_length": 2685.9525862068967,
      "eval_bbeh_mini_completions/mean_terminated_length": 1877.7948924097523,
      "eval_bbeh_mini_completions/min_length": 571.0689655172414,
      "eval_bbeh_mini_completions/min_terminated_length": 571.0689655172414,
      "eval_bbeh_mini_entropy": 0.35236285369971704,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 4550697.0,
      "eval_bbeh_mini_reward": 0.10344827586206896,
      "eval_bbeh_mini_reward_std": 0.2699657020897701,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.10344827586206896,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.26996571339409925,
      "eval_bbeh_mini_runtime": 494.6737,
      "eval_bbeh_mini_samples_per_second": 0.93,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.2562965689034298,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.1846869053511784,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 8.863040915959077e-06,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 3.5924758705599555,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.02370224494872422,
      "eval_bbeh_mini_steps_per_second": 0.059,
      "step": 300
    },
    {
      "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": 0.07,
      "completions/max_length": 2060.08,
      "completions/max_terminated_length": 1839.84,
      "completions/mean_length": 893.005,
      "completions/mean_terminated_length": 669.4148620605469,
      "completions/min_length": 229.84,
      "completions/min_terminated_length": 229.84,
      "entropy": 0.40538407385349273,
      "epoch": 0.065,
      "frac_reward_zero_std": 0.52,
      "grad_norm": 5.036384229761231,
      "learning_rate": 9.352e-07,
      "loss": 0.0219,
      "num_tokens": 4961307.0,
      "reward": 0.61,
      "reward_std": 0.39049545288085935,
      "rewards/combined_reward/mean": 0.61,
      "rewards/combined_reward/std": 0.3904954707622528,
      "sampling/importance_sampling_ratio/max": 2.171120972633362,
      "sampling/importance_sampling_ratio/mean": 0.6055274987220765,
      "sampling/importance_sampling_ratio/min": 0.031129270781748347,
      "sampling/sampling_logp_difference/max": 0.7485472440719605,
      "sampling/sampling_logp_difference/mean": 0.02611954465508461,
      "step": 325,
      "step_time": 7.8838804703298955
    },
    {
      "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": 0.0075,
      "completions/max_length": 1540.16,
      "completions/max_terminated_length": 1477.4,
      "completions/mean_length": 649.275,
      "completions/mean_terminated_length": 629.781923828125,
      "completions/min_length": 224.08,
      "completions/min_terminated_length": 224.08,
      "entropy": 0.44817891597747805,
      "epoch": 0.07,
      "frac_reward_zero_std": 0.4,
      "grad_norm": 1.262847377601464,
      "learning_rate": 9.302e-07,
      "loss": -0.0023,
      "num_tokens": 5272369.0,
      "reward": 0.63,
      "reward_std": 0.40423260688781737,
      "rewards/combined_reward/mean": 0.63,
      "rewards/combined_reward/std": 0.4042326259613037,
      "sampling/importance_sampling_ratio/max": 1.9462181329727173,
      "sampling/importance_sampling_ratio/mean": 0.5410465461015701,
      "sampling/importance_sampling_ratio/min": 0.007918943986296653,
      "sampling/sampling_logp_difference/max": 0.658315749168396,
      "sampling/sampling_logp_difference/mean": 0.028008888587355615,
      "step": 350,
      "step_time": 5.602628886112943
    },
    {
      "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": 0.055,
      "completions/max_length": 1894.84,
      "completions/max_terminated_length": 1623.76,
      "completions/mean_length": 823.715,
      "completions/mean_terminated_length": 638.96466796875,
      "completions/min_length": 199.48,
      "completions/min_terminated_length": 199.48,
      "entropy": 0.4437406814098358,
      "epoch": 0.075,
      "frac_reward_zero_std": 0.42,
      "grad_norm": 1.416585270355939,
      "learning_rate": 9.251999999999999e-07,
      "loss": -0.0063,
      "num_tokens": 5651599.0,
      "reward": 0.6375,
      "reward_std": 0.4141005539894104,
      "rewards/combined_reward/mean": 0.6375,
      "rewards/combined_reward/std": 0.41410056471824647,
      "sampling/importance_sampling_ratio/max": 2.079793190956116,
      "sampling/importance_sampling_ratio/mean": 0.6110782504081727,
      "sampling/importance_sampling_ratio/min": 0.012589475377462805,
      "sampling/sampling_logp_difference/max": 0.7504757452011108,
      "sampling/sampling_logp_difference/mean": 0.027451810911297798,
      "step": 375,
      "step_time": 7.127676211716607
    },
    {
      "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": 0.04,
      "completions/max_length": 1728.44,
      "completions/max_terminated_length": 1628.96,
      "completions/mean_length": 761.1325,
      "completions/mean_terminated_length": 657.4637036132813,
      "completions/min_length": 202.8,
      "completions/min_terminated_length": 202.8,
      "entropy": 0.40143827855587005,
      "epoch": 0.08,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.0,
      "learning_rate": 9.202e-07,
      "loss": 0.0175,
      "num_tokens": 6009092.0,
      "reward": 0.5775,
      "reward_std": 0.38390336275100706,
      "rewards/combined_reward/mean": 0.5775,
      "rewards/combined_reward/std": 0.38390338063240054,
      "sampling/importance_sampling_ratio/max": 2.096400910615921,
      "sampling/importance_sampling_ratio/mean": 0.6684997088462115,
      "sampling/importance_sampling_ratio/min": 0.027380506457557203,
      "sampling/sampling_logp_difference/max": 0.6964709210395813,
      "sampling/sampling_logp_difference/mean": 0.02609425738453865,
      "step": 400,
      "step_time": 6.615593434758484
    },
    {
      "epoch": 0.08,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.33620689655172414,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3580.5172413793102,
      "eval_bbeh_mini_completions/mean_length": 2591.877155172414,
      "eval_bbeh_mini_completions/mean_terminated_length": 1825.3767721241918,
      "eval_bbeh_mini_completions/min_length": 451.62068965517244,
      "eval_bbeh_mini_completions/min_terminated_length": 451.62068965517244,
      "eval_bbeh_mini_entropy": 0.3378912878447565,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 6009092.0,
      "eval_bbeh_mini_reward": 0.10560344827586207,
      "eval_bbeh_mini_reward_std": 0.25904289196277486,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.10560344827586207,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.2590429012117715,
      "eval_bbeh_mini_runtime": 491.0126,
      "eval_bbeh_mini_samples_per_second": 0.937,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.1502898427946815,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.16786166884261985,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 5.115552837594649e-05,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 2.436232163988311,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.023024646565318108,
      "eval_bbeh_mini_steps_per_second": 0.059,
      "step": 400
    },
    {
      "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": 0.0725,
      "completions/max_length": 2179.08,
      "completions/max_terminated_length": 2003.44,
      "completions/mean_length": 953.6125,
      "completions/mean_terminated_length": 736.740048828125,
      "completions/min_length": 218.08,
      "completions/min_terminated_length": 218.08,
      "entropy": 0.4012847465276718,
      "epoch": 0.085,
      "frac_reward_zero_std": 0.42,
      "grad_norm": 0.5398184196341114,
      "learning_rate": 9.151999999999999e-07,
      "loss": -0.0237,
      "num_tokens": 6441177.0,
      "reward": 0.695,
      "reward_std": 0.3829051518440247,
      "rewards/combined_reward/mean": 0.695,
      "rewards/combined_reward/std": 0.382905170917511,
      "sampling/importance_sampling_ratio/max": 1.9372561621665954,
      "sampling/importance_sampling_ratio/mean": 0.5673770725727081,
      "sampling/importance_sampling_ratio/min": 0.03025458916346361,
      "sampling/sampling_logp_difference/max": 0.7728560495376587,
      "sampling/sampling_logp_difference/mean": 0.02587804526090622,
      "step": 425,
      "step_time": 8.324159880718216
    },
    {
      "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": 0.03,
      "completions/max_length": 1513.88,
      "completions/max_terminated_length": 1419.48,
      "completions/mean_length": 615.395,
      "completions/mean_terminated_length": 520.6183935546875,
      "completions/min_length": 177.0,
      "completions/min_terminated_length": 177.0,
      "entropy": 0.41070492506027223,
      "epoch": 0.09,
      "frac_reward_zero_std": 0.52,
      "grad_norm": 1.5270349642075316,
      "learning_rate": 9.102e-07,
      "loss": -0.0022,
      "num_tokens": 6739631.0,
      "reward": 0.735,
      "reward_std": 0.33413492679595946,
      "rewards/combined_reward/mean": 0.735,
      "rewards/combined_reward/std": 0.3341349399089813,
      "sampling/importance_sampling_ratio/max": 2.027468514442444,
      "sampling/importance_sampling_ratio/mean": 0.6084394156932831,
      "sampling/importance_sampling_ratio/min": 0.02255286922645837,
      "sampling/sampling_logp_difference/max": 0.5962899017333985,
      "sampling/sampling_logp_difference/mean": 0.02680632047355175,
      "step": 450,
      "step_time": 5.793452201336622
    },
    {
      "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": 0.025,
      "completions/max_length": 1631.92,
      "completions/max_terminated_length": 1465.8,
      "completions/mean_length": 605.425,
      "completions/mean_terminated_length": 531.262001953125,
      "completions/min_length": 168.24,
      "completions/min_terminated_length": 168.24,
      "entropy": 0.4006887876987457,
      "epoch": 0.095,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.0,
      "learning_rate": 9.051999999999999e-07,
      "loss": 0.051,
      "num_tokens": 7030409.0,
      "reward": 0.685,
      "reward_std": 0.3645173120498657,
      "rewards/combined_reward/mean": 0.685,
      "rewards/combined_reward/std": 0.3645173156261444,
      "sampling/importance_sampling_ratio/max": 1.9704220461845399,
      "sampling/importance_sampling_ratio/mean": 0.62828111410141,
      "sampling/importance_sampling_ratio/min": 0.012673925184790278,
      "sampling/sampling_logp_difference/max": 0.7063542747497559,
      "sampling/sampling_logp_difference/mean": 0.026762503162026406,
      "step": 475,
      "step_time": 6.15127008873038
    },
    {
      "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": 0.0275,
      "completions/max_length": 1725.32,
      "completions/max_terminated_length": 1643.52,
      "completions/mean_length": 700.63,
      "completions/mean_terminated_length": 630.799814453125,
      "completions/min_length": 195.16,
      "completions/min_terminated_length": 195.16,
      "entropy": 0.40278098702430726,
      "epoch": 0.1,
      "frac_reward_zero_std": 0.46,
      "grad_norm": 0.0,
      "learning_rate": 9.002e-07,
      "loss": 0.0015,
      "num_tokens": 7360269.0,
      "reward": 0.7075,
      "reward_std": 0.32035229206085203,
      "rewards/combined_reward/mean": 0.7075,
      "rewards/combined_reward/std": 0.3203523027896881,
      "sampling/importance_sampling_ratio/max": 1.9510860872268676,
      "sampling/importance_sampling_ratio/mean": 0.5961837387084961,
      "sampling/importance_sampling_ratio/min": 0.02190992054884191,
      "sampling/sampling_logp_difference/max": 0.6417275023460388,
      "sampling/sampling_logp_difference/mean": 0.026219148635864258,
      "step": 500,
      "step_time": 6.48362862716429
    },
    {
      "epoch": 0.1,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.3017241379310345,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3449.0689655172414,
      "eval_bbeh_mini_completions/mean_length": 2359.5905172413795,
      "eval_bbeh_mini_completions/mean_terminated_length": 1620.2735132677801,
      "eval_bbeh_mini_completions/min_length": 336.0689655172414,
      "eval_bbeh_mini_completions/min_terminated_length": 336.0689655172414,
      "eval_bbeh_mini_entropy": 0.338118112292783,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 7360269.0,
      "eval_bbeh_mini_reward": 0.10344827586206896,
      "eval_bbeh_mini_reward_std": 0.27482735083020965,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.10344827586206896,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.27482736316220513,
      "eval_bbeh_mini_runtime": 478.438,
      "eval_bbeh_mini_samples_per_second": 0.961,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.6468630525572547,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.2634151639609501,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 2.0942759789977906e-05,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 3.216358838410213,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.023251901348603183,
      "eval_bbeh_mini_steps_per_second": 0.061,
      "step": 500
    },
    {
      "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": 0.03,
      "completions/max_length": 1828.68,
      "completions/max_terminated_length": 1601.08,
      "completions/mean_length": 683.7675,
      "completions/mean_terminated_length": 593.3429418945312,
      "completions/min_length": 211.0,
      "completions/min_terminated_length": 211.0,
      "entropy": 0.4082434940338135,
      "epoch": 0.105,
      "frac_reward_zero_std": 0.44,
      "grad_norm": 0.0,
      "learning_rate": 8.951999999999999e-07,
      "loss": 0.0051,
      "num_tokens": 7685824.0,
      "reward": 0.6175,
      "reward_std": 0.4563824462890625,
      "rewards/combined_reward/mean": 0.6175,
      "rewards/combined_reward/std": 0.4563824665546417,
      "sampling/importance_sampling_ratio/max": 2.16779132604599,
      "sampling/importance_sampling_ratio/mean": 0.6069332575798034,
      "sampling/importance_sampling_ratio/min": 0.019685860924364532,
      "sampling/sampling_logp_difference/max": 0.6934374618530273,
      "sampling/sampling_logp_difference/mean": 0.026688539236783982,
      "step": 525,
      "step_time": 6.72239572064951
    },
    {
      "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": 0.03,
      "completions/max_length": 1910.32,
      "completions/max_terminated_length": 1841.88,
      "completions/mean_length": 776.53,
      "completions/mean_terminated_length": 694.096181640625,
      "completions/min_length": 208.48,
      "completions/min_terminated_length": 208.48,
      "entropy": 0.4136668372154236,
      "epoch": 0.11,
      "frac_reward_zero_std": 0.52,
      "grad_norm": 1.4963386849443154,
      "learning_rate": 8.902e-07,
      "loss": 0.0071,
      "num_tokens": 8047172.0,
      "reward": 0.705,
      "reward_std": 0.37073400259017947,
      "rewards/combined_reward/mean": 0.705,
      "rewards/combined_reward/std": 0.3707340121269226,
      "sampling/importance_sampling_ratio/max": 2.074637842178345,
      "sampling/importance_sampling_ratio/mean": 0.5935343915224075,
      "sampling/importance_sampling_ratio/min": 0.033500756603752964,
      "sampling/sampling_logp_difference/max": 0.6976750469207764,
      "sampling/sampling_logp_difference/mean": 0.02682505249977112,
      "step": 550,
      "step_time": 7.161583522455767
    },
    {
      "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": 0.04,
      "completions/max_length": 2021.56,
      "completions/max_terminated_length": 1834.32,
      "completions/mean_length": 893.8875,
      "completions/mean_terminated_length": 778.9953515625,
      "completions/min_length": 246.56,
      "completions/min_terminated_length": 246.56,
      "entropy": 0.38500270128250125,
      "epoch": 0.115,
      "frac_reward_zero_std": 0.3,
      "grad_norm": 0.0,
      "learning_rate": 8.851999999999999e-07,
      "loss": 0.0164,
      "num_tokens": 8457615.0,
      "reward": 0.555,
      "reward_std": 0.4233776617050171,
      "rewards/combined_reward/mean": 0.555,
      "rewards/combined_reward/std": 0.42337767004966737,
      "sampling/importance_sampling_ratio/max": 2.0310398817062376,
      "sampling/importance_sampling_ratio/mean": 0.5933060163259506,
      "sampling/importance_sampling_ratio/min": 0.04317198476808926,
      "sampling/sampling_logp_difference/max": 0.821718487739563,
      "sampling/sampling_logp_difference/mean": 0.02523998260498047,
      "step": 575,
      "step_time": 7.77931922157295
    },
    {
      "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": 0.0125,
      "completions/max_length": 1184.28,
      "completions/max_terminated_length": 1105.92,
      "completions/mean_length": 524.98,
      "completions/mean_terminated_length": 489.793486328125,
      "completions/min_length": 213.0,
      "completions/min_terminated_length": 213.0,
      "entropy": 0.3971246266365051,
      "epoch": 0.12,
      "frac_reward_zero_std": 0.46,
      "grad_norm": 0.0,
      "learning_rate": 8.802e-07,
      "loss": -0.0002,
      "num_tokens": 8715415.0,
      "reward": 0.705,
      "reward_std": 0.376721715927124,
      "rewards/combined_reward/mean": 0.705,
      "rewards/combined_reward/std": 0.3767217338085175,
      "sampling/importance_sampling_ratio/max": 2.0127450704574583,
      "sampling/importance_sampling_ratio/mean": 0.6475339376926422,
      "sampling/importance_sampling_ratio/min": 0.04534088842570782,
      "sampling/sampling_logp_difference/max": 0.7423757553100586,
      "sampling/sampling_logp_difference/mean": 0.025974995717406274,
      "step": 600,
      "step_time": 4.796778986351565
    },
    {
      "epoch": 0.12,
      "eval_bbeh_mini_clip_ratio/high_max": 0.0,
      "eval_bbeh_mini_clip_ratio/high_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_mean": 0.0,
      "eval_bbeh_mini_clip_ratio/low_min": 0.0,
      "eval_bbeh_mini_clip_ratio/region_mean": 0.0,
      "eval_bbeh_mini_completions/clipped_ratio": 0.2650862068965517,
      "eval_bbeh_mini_completions/max_length": 4096.0,
      "eval_bbeh_mini_completions/max_terminated_length": 3541.310344827586,
      "eval_bbeh_mini_completions/mean_length": 2310.0086206896553,
      "eval_bbeh_mini_completions/mean_terminated_length": 1660.6524237271012,
      "eval_bbeh_mini_completions/min_length": 390.48275862068965,
      "eval_bbeh_mini_completions/min_terminated_length": 390.48275862068965,
      "eval_bbeh_mini_entropy": 0.33623800195496656,
      "eval_bbeh_mini_frac_reward_zero_std": 1.0,
      "eval_bbeh_mini_loss": 0.0,
      "eval_bbeh_mini_num_tokens": 8715415.0,
      "eval_bbeh_mini_reward": 0.10775862068965517,
      "eval_bbeh_mini_reward_std": 0.27093698238504343,
      "eval_bbeh_mini_rewards/combined_reward/mean": 0.10775862068965517,
      "eval_bbeh_mini_rewards/combined_reward/std": 0.27093699163404006,
      "eval_bbeh_mini_runtime": 475.4716,
      "eval_bbeh_mini_samples_per_second": 0.967,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/max": 1.5325407144324532,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/mean": 0.24085120471387073,
      "eval_bbeh_mini_sampling/importance_sampling_ratio/min": 5.8291943258501656e-05,
      "eval_bbeh_mini_sampling/sampling_logp_difference/max": 3.5541970894254487,
      "eval_bbeh_mini_sampling/sampling_logp_difference/mean": 0.023053794135821277,
      "eval_bbeh_mini_steps_per_second": 0.061,
      "step": 600
    }
  ],
  "logging_steps": 25,
  "max_steps": 5000,
  "num_input_tokens_seen": 8715415,
  "num_train_epochs": 1,
  "save_steps": 100,
  "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": 8,
  "trial_name": null,
  "trial_params": null
}