{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.4705882352941178,
  "eval_steps": 10,
  "global_step": 100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.013888889302810034,
      "eval_completions/max_length": 1051.047619047619,
      "eval_completions/max_terminated_length": 869.1190476190476,
      "eval_completions/mean_length": 481.75398054577056,
      "eval_completions/mean_terminated_length": 445.6996131170364,
      "eval_completions/min_length": 214.35714285714286,
      "eval_completions/min_terminated_length": 214.35714285714286,
      "eval_entropy": 0.19031873114761852,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 0.0,
      "eval_reward": 0.4583333432674408,
      "eval_reward_std": 0.4926568010733241,
      "eval_rewards/reward_correctness/mean": 0.4583333383003871,
      "eval_rewards/reward_correctness/std": 0.49265679788021816,
      "eval_runtime": 621.4365,
      "eval_samples_per_second": 0.805,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9897958558230173,
      "eval_sampling/importance_sampling_ratio/min": 0.5083668609814984,
      "eval_sampling/sampling_logp_difference/max": 2.397807226294563,
      "eval_sampling/sampling_logp_difference/mean": 0.0472170448019391,
      "eval_steps_per_second": 0.135,
      "step": 0
    },
    {
      "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": 0.006510416977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3039.0,
      "completions/mean_length": 544.44140625,
      "completions/mean_terminated_length": 527.8781127929688,
      "completions/min_length": 97.0,
      "completions/min_terminated_length": 97.0,
      "entropy": 0.23509776848368347,
      "epoch": 0.014705882352941176,
      "frac_reward_zero_std": 0.484375,
      "grad_norm": 0.09276555104929421,
      "learning_rate": 0.0,
      "loss": 0.0089,
      "num_tokens": 1008798.0,
      "reward": 0.3372395932674408,
      "reward_std": 0.4729214012622833,
      "rewards/reward_correctness/mean": 0.3372395932674408,
      "rewards/reward_correctness/std": 0.4729214012622833,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9975379109382629,
      "sampling/importance_sampling_ratio/min": 9.726321137382143e-11,
      "sampling/sampling_logp_difference/max": 23.053600311279297,
      "sampling/sampling_logp_difference/mean": 0.042724158614873886,
      "step": 1,
      "step_time": 338.8089781026356
    },
    {
      "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": 0.005859375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3041.0,
      "completions/mean_length": 531.7975463867188,
      "completions/mean_terminated_length": 516.8258056640625,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "entropy": 0.23647121316753328,
      "epoch": 0.029411764705882353,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.09973641051476721,
      "learning_rate": 6.000000000000001e-07,
      "loss": 0.0144,
      "num_tokens": 1986295.0,
      "reward": 0.3059895932674408,
      "reward_std": 0.46097537875175476,
      "rewards/reward_correctness/mean": 0.3059895932674408,
      "rewards/reward_correctness/std": 0.46097537875175476,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000666379928589,
      "sampling/importance_sampling_ratio/min": 0.1762581765651703,
      "sampling/sampling_logp_difference/max": 1.7358055114746094,
      "sampling/sampling_logp_difference/mean": 0.015235913917422295,
      "step": 2,
      "step_time": 280.3220818070695
    },
    {
      "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": 0.005859375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2961.0,
      "completions/mean_length": 541.4909057617188,
      "completions/mean_terminated_length": 526.5762939453125,
      "completions/min_length": 93.0,
      "completions/min_terminated_length": 93.0,
      "entropy": 0.23211678757797927,
      "epoch": 0.04411764705882353,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.09852825942987486,
      "learning_rate": 1.2000000000000002e-06,
      "loss": 0.0094,
      "num_tokens": 2958029.0,
      "reward": 0.3697916865348816,
      "reward_std": 0.4829053580760956,
      "rewards/reward_correctness/mean": 0.3697916567325592,
      "rewards/reward_correctness/std": 0.48290541768074036,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999954104423523,
      "sampling/importance_sampling_ratio/min": 0.1790223866701126,
      "sampling/sampling_logp_difference/max": 2.3153841495513916,
      "sampling/sampling_logp_difference/mean": 0.014817245304584503,
      "step": 3,
      "step_time": 277.6638293429278
    },
    {
      "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": 0.005859375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2881.0,
      "completions/mean_length": 527.78515625,
      "completions/mean_terminated_length": 512.789794921875,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "entropy": 0.2507125422125682,
      "epoch": 0.058823529411764705,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.09895279498528099,
      "learning_rate": 1.8e-06,
      "loss": 0.0037,
      "num_tokens": 3932327.0,
      "reward": 0.2903645932674408,
      "reward_std": 0.45407843589782715,
      "rewards/reward_correctness/mean": 0.2903645932674408,
      "rewards/reward_correctness/std": 0.45407843589782715,
      "sampling/importance_sampling_ratio/max": 2.93314528465271,
      "sampling/importance_sampling_ratio/mean": 1.0000152587890625,
      "sampling/importance_sampling_ratio/min": 0.24576407670974731,
      "sampling/sampling_logp_difference/max": 1.4033832550048828,
      "sampling/sampling_logp_difference/mean": 0.016350775957107544,
      "step": 4,
      "step_time": 275.9684011223726
    },
    {
      "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": 0.0078125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3004.0,
      "completions/mean_length": 557.0267333984375,
      "completions/mean_terminated_length": 537.2237548828125,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "entropy": 0.23487792268861085,
      "epoch": 0.07352941176470588,
      "frac_reward_zero_std": 0.484375,
      "grad_norm": 0.08832062141924793,
      "learning_rate": 2.4000000000000003e-06,
      "loss": 0.0098,
      "num_tokens": 4957852.0,
      "reward": 0.3430989682674408,
      "reward_std": 0.4748988151550293,
      "rewards/reward_correctness/mean": 0.3430989682674408,
      "rewards/reward_correctness/std": 0.4748988151550293,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999831318855286,
      "sampling/importance_sampling_ratio/min": 0.10895482450723648,
      "sampling/sampling_logp_difference/max": 2.2168219089508057,
      "sampling/sampling_logp_difference/mean": 0.014919000677764416,
      "step": 5,
      "step_time": 286.02837926428765
    },
    {
      "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": 0.004557291977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2260.0,
      "completions/mean_length": 525.5950927734375,
      "completions/mean_terminated_length": 513.9371948242188,
      "completions/min_length": 120.0,
      "completions/min_terminated_length": 120.0,
      "entropy": 0.24871817359235138,
      "epoch": 0.08823529411764706,
      "frac_reward_zero_std": 0.4765625,
      "grad_norm": 0.09720010779023536,
      "learning_rate": 3e-06,
      "loss": 0.0129,
      "num_tokens": 5911986.0,
      "reward": 0.388671875,
      "reward_std": 0.4876072406768799,
      "rewards/reward_correctness/mean": 0.388671875,
      "rewards/reward_correctness/std": 0.48760727047920227,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000280141830444,
      "sampling/importance_sampling_ratio/min": 0.06059809401631355,
      "sampling/sampling_logp_difference/max": 2.8034918308258057,
      "sampling/sampling_logp_difference/mean": 0.015820473432540894,
      "step": 6,
      "step_time": 278.1581368963234
    },
    {
      "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": 0.0078125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2586.0,
      "completions/mean_length": 557.4264526367188,
      "completions/mean_terminated_length": 537.6266479492188,
      "completions/min_length": 83.0,
      "completions/min_terminated_length": 83.0,
      "entropy": 0.24867978470865637,
      "epoch": 0.10294117647058823,
      "frac_reward_zero_std": 0.515625,
      "grad_norm": 0.0920658987432061,
      "learning_rate": 2.9996118137817615e-06,
      "loss": 0.0052,
      "num_tokens": 6918073.0,
      "reward": 0.39453125,
      "reward_std": 0.48890894651412964,
      "rewards/reward_correctness/mean": 0.39453125,
      "rewards/reward_correctness/std": 0.48890894651412964,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000068187713623,
      "sampling/importance_sampling_ratio/min": 0.055469363927841187,
      "sampling/sampling_logp_difference/max": 2.8919243812561035,
      "sampling/sampling_logp_difference/mean": 0.015951288864016533,
      "step": 7,
      "step_time": 283.01500198384747
    },
    {
      "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": 3072.0,
      "completions/max_terminated_length": 3047.0,
      "completions/mean_length": 644.6015625,
      "completions/mean_terminated_length": 606.0714111328125,
      "completions/min_length": 152.0,
      "completions/min_terminated_length": 152.0,
      "entropy": 0.2523061840329319,
      "epoch": 0.11764705882352941,
      "frac_reward_zero_std": 0.4765625,
      "grad_norm": 0.08226176001367254,
      "learning_rate": 2.998447478369329e-06,
      "loss": 0.0112,
      "num_tokens": 8052013.0,
      "reward": 0.3118489682674408,
      "reward_std": 0.46339935064315796,
      "rewards/reward_correctness/mean": 0.3118489682674408,
      "rewards/reward_correctness/std": 0.46339938044548035,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000117301940918,
      "sampling/importance_sampling_ratio/min": 0.11431955546140671,
      "sampling/sampling_logp_difference/max": 2.168757677078247,
      "sampling/sampling_logp_difference/mean": 0.015464004129171371,
      "step": 8,
      "step_time": 282.23121579224244
    },
    {
      "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": 0.01302083395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2863.0,
      "completions/mean_length": 615.9193115234375,
      "completions/mean_terminated_length": 583.5171508789062,
      "completions/min_length": 157.0,
      "completions/min_terminated_length": 157.0,
      "entropy": 0.24067589035257697,
      "epoch": 0.1323529411764706,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.09128268806596039,
      "learning_rate": 2.9965076633611604e-06,
      "loss": 0.0138,
      "num_tokens": 9143985.0,
      "reward": 0.35546875,
      "reward_std": 0.47881096601486206,
      "rewards/reward_correctness/mean": 0.35546875,
      "rewards/reward_correctness/std": 0.47881099581718445,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000066757202148,
      "sampling/importance_sampling_ratio/min": 0.20224328339099884,
      "sampling/sampling_logp_difference/max": 1.5982838869094849,
      "sampling/sampling_logp_difference/mean": 0.014961779117584229,
      "step": 9,
      "step_time": 293.5717437211424
    },
    {
      "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": 0.01692708395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3064.0,
      "completions/mean_length": 622.6705932617188,
      "completions/mean_terminated_length": 580.5986938476562,
      "completions/min_length": 142.0,
      "completions/min_terminated_length": 142.0,
      "entropy": 0.22710344230290502,
      "epoch": 0.14705882352941177,
      "frac_reward_zero_std": 0.4296875,
      "grad_norm": 0.09245324413811332,
      "learning_rate": 2.993793484326816e-06,
      "loss": 0.0046,
      "num_tokens": 10266823.0,
      "reward": 0.3509114682674408,
      "reward_std": 0.4774106740951538,
      "rewards/reward_correctness/mean": 0.3509114682674408,
      "rewards/reward_correctness/std": 0.4774107038974762,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000184774398804,
      "sampling/importance_sampling_ratio/min": 0.18761380016803741,
      "sampling/sampling_logp_difference/max": 1.6733696460723877,
      "sampling/sampling_logp_difference/mean": 0.014758117496967316,
      "step": 10,
      "step_time": 302.41576358349994
    },
    {
      "epoch": 0.14705882352941177,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.011904762259551458,
      "eval_completions/max_length": 1140.107142857143,
      "eval_completions/max_terminated_length": 993.6309523809524,
      "eval_completions/mean_length": 548.2758098783947,
      "eval_completions/mean_terminated_length": 518.2567600068592,
      "eval_completions/min_length": 244.98809523809524,
      "eval_completions/min_terminated_length": 244.98809523809524,
      "eval_entropy": 0.19089051132046042,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 10266823.0,
      "eval_reward": 0.4801587403884956,
      "eval_reward_std": 0.48688212249960217,
      "eval_rewards/reward_correctness/mean": 0.48015873471186277,
      "eval_rewards/reward_correctness/std": 0.48688211966128575,
      "eval_runtime": 674.283,
      "eval_samples_per_second": 0.742,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9907156576712927,
      "eval_sampling/importance_sampling_ratio/min": 0.4924284939964612,
      "eval_sampling/sampling_logp_difference/max": 2.736755425021762,
      "eval_sampling/sampling_logp_difference/mean": 0.048229864931532314,
      "eval_steps_per_second": 0.125,
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3035.0,
      "completions/mean_length": 645.2077026367188,
      "completions/mean_terminated_length": 598.5076293945312,
      "completions/min_length": 98.0,
      "completions/min_terminated_length": 98.0,
      "entropy": 0.24283059418667108,
      "epoch": 0.16176470588235295,
      "frac_reward_zero_std": 0.421875,
      "grad_norm": 0.0974447715454529,
      "learning_rate": 2.990306502165398e-06,
      "loss": 0.0102,
      "num_tokens": 11415890.0,
      "reward": 0.3795573115348816,
      "reward_std": 0.48543480038642883,
      "rewards/reward_correctness/mean": 0.3795572817325592,
      "rewards/reward_correctness/std": 0.4854348301887512,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000299215316772,
      "sampling/importance_sampling_ratio/min": 0.14718766510486603,
      "sampling/sampling_logp_difference/max": 1.9160468578338623,
      "sampling/sampling_logp_difference/mean": 0.015639059245586395,
      "step": 11,
      "step_time": 291.7094019954093
    },
    {
      "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": 0.0201822929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2988.0,
      "completions/mean_length": 683.3971557617188,
      "completions/mean_terminated_length": 634.1966552734375,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "entropy": 0.25403576996177435,
      "epoch": 0.17647058823529413,
      "frac_reward_zero_std": 0.4453125,
      "grad_norm": 0.09531950301623979,
      "learning_rate": 2.986048722207899e-06,
      "loss": 0.012,
      "num_tokens": 12632868.0,
      "reward": 0.380859375,
      "reward_std": 0.4857562184333801,
      "rewards/reward_correctness/mean": 0.380859375,
      "rewards/reward_correctness/std": 0.4857562482357025,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999941885471344,
      "sampling/importance_sampling_ratio/min": 0.05190279707312584,
      "sampling/sampling_logp_difference/max": 2.9583826065063477,
      "sampling/sampling_logp_difference/mean": 0.01636284403502941,
      "step": 12,
      "step_time": 290.1786908484064
    },
    {
      "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": 0.0162760429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3024.0,
      "completions/mean_length": 671.1953125,
      "completions/mean_terminated_length": 633.8160400390625,
      "completions/min_length": 128.0,
      "completions/min_terminated_length": 128.0,
      "entropy": 0.23784961050841957,
      "epoch": 0.19117647058823528,
      "frac_reward_zero_std": 0.40625,
      "grad_norm": 0.09189129853735502,
      "learning_rate": 2.981022593063946e-06,
      "loss": 0.0089,
      "num_tokens": 13847616.0,
      "reward": 0.357421875,
      "reward_std": 0.4793965518474579,
      "rewards/reward_correctness/mean": 0.357421875,
      "rewards/reward_correctness/std": 0.47939661145210266,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9972469210624695,
      "sampling/importance_sampling_ratio/min": 1.9157713474415514e-14,
      "sampling/sampling_logp_difference/max": 31.586071014404297,
      "sampling/sampling_logp_difference/mean": 0.04794478043913841,
      "step": 13,
      "step_time": 341.73873996781185
    },
    {
      "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": 0.02083333395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3065.0,
      "completions/mean_length": 707.0162963867188,
      "completions/mean_terminated_length": 656.6974487304688,
      "completions/min_length": 126.0,
      "completions/min_terminated_length": 126.0,
      "entropy": 0.2236228445544839,
      "epoch": 0.20588235294117646,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.08294063821997087,
      "learning_rate": 2.9752310052136353e-06,
      "loss": 0.0156,
      "num_tokens": 15082525.0,
      "reward": 0.3951823115348816,
      "reward_std": 0.4890490174293518,
      "rewards/reward_correctness/mean": 0.3951822817325592,
      "rewards/reward_correctness/std": 0.4890490770339966,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000492334365845,
      "sampling/importance_sampling_ratio/min": 0.06042155250906944,
      "sampling/sampling_logp_difference/max": 2.8064093589782715,
      "sampling/sampling_logp_difference/mean": 0.015129472129046917,
      "step": 14,
      "step_time": 307.8643926582299
    },
    {
      "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": 3072.0,
      "completions/max_terminated_length": 2959.0,
      "completions/mean_length": 675.498046875,
      "completions/mean_terminated_length": 637.7943115234375,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "entropy": 0.23216103401500732,
      "epoch": 0.22058823529411764,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.10101422201216483,
      "learning_rate": 2.968677289345242e-06,
      "loss": 0.0107,
      "num_tokens": 16288186.0,
      "reward": 0.3997395932674408,
      "reward_std": 0.49000421166419983,
      "rewards/reward_correctness/mean": 0.3997395932674408,
      "rewards/reward_correctness/std": 0.4900042414665222,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9976019859313965,
      "sampling/importance_sampling_ratio/min": 5.457596647330436e-18,
      "sampling/sampling_logp_difference/max": 39.7495231628418,
      "sampling/sampling_logp_difference/mean": 0.04638348147273064,
      "step": 15,
      "step_time": 320.05006989324465
    },
    {
      "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": 0.02734375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3016.0,
      "completions/mean_length": 722.0931396484375,
      "completions/mean_terminated_length": 656.0314331054688,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "entropy": 0.24111876578535885,
      "epoch": 0.23529411764705882,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.08050920134854482,
      "learning_rate": 2.9613652144397706e-06,
      "loss": 0.0131,
      "num_tokens": 17574597.0,
      "reward": 0.361328125,
      "reward_std": 0.48054182529449463,
      "rewards/reward_correctness/mean": 0.361328125,
      "rewards/reward_correctness/std": 0.48054182529449463,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000146627426147,
      "sampling/importance_sampling_ratio/min": 0.08782368153333664,
      "sampling/sampling_logp_difference/max": 2.4324240684509277,
      "sampling/sampling_logp_difference/mean": 0.015971485525369644,
      "step": 16,
      "step_time": 324.0193057116121
    },
    {
      "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": 0.017578125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 711.9896240234375,
      "completions/mean_terminated_length": 669.7627563476562,
      "completions/min_length": 131.0,
      "completions/min_terminated_length": 131.0,
      "entropy": 0.22629119083285332,
      "epoch": 0.25,
      "frac_reward_zero_std": 0.4609375,
      "grad_norm": 0.0900011417432106,
      "learning_rate": 2.9532989856034515e-06,
      "loss": 0.0086,
      "num_tokens": 18819317.0,
      "reward": 0.38671875,
      "reward_std": 0.4871568977832794,
      "rewards/reward_correctness/mean": 0.38671875,
      "rewards/reward_correctness/std": 0.4871569275856018,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.997471809387207,
      "sampling/importance_sampling_ratio/min": 3.498903030914151e-14,
      "sampling/sampling_logp_difference/max": 30.983741760253906,
      "sampling/sampling_logp_difference/mean": 0.045513179153203964,
      "step": 17,
      "step_time": 328.7708681197837
    },
    {
      "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": 0.025390625,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3011.0,
      "completions/mean_length": 725.828125,
      "completions/mean_terminated_length": 664.7054443359375,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "entropy": 0.22558519581798464,
      "epoch": 0.2647058823529412,
      "frac_reward_zero_std": 0.421875,
      "grad_norm": 0.09060650894291081,
      "learning_rate": 2.94448324164942e-06,
      "loss": 0.01,
      "num_tokens": 20101097.0,
      "reward": 0.3626302182674408,
      "reward_std": 0.4809158742427826,
      "rewards/reward_correctness/mean": 0.3626302182674408,
      "rewards/reward_correctness/std": 0.480915904045105,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000027418136597,
      "sampling/importance_sampling_ratio/min": 0.1281590312719345,
      "sampling/sampling_logp_difference/max": 2.4219703674316406,
      "sampling/sampling_logp_difference/mean": 0.015497544780373573,
      "step": 18,
      "step_time": 311.72907185554504
    },
    {
      "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": 0.0221354179084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3016.0,
      "completions/mean_length": 702.263671875,
      "completions/mean_terminated_length": 648.6211547851562,
      "completions/min_length": 148.0,
      "completions/min_terminated_length": 148.0,
      "entropy": 0.2129860440036282,
      "epoch": 0.27941176470588236,
      "frac_reward_zero_std": 0.4296875,
      "grad_norm": 0.0864612338873769,
      "learning_rate": 2.934923052429984e-06,
      "loss": 0.0126,
      "num_tokens": 21336422.0,
      "reward": 0.416015625,
      "reward_std": 0.49305665493011475,
      "rewards/reward_correctness/mean": 0.416015625,
      "rewards/reward_correctness/std": 0.49305668473243713,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000278949737549,
      "sampling/importance_sampling_ratio/min": 0.19384071230888367,
      "sampling/sampling_logp_difference/max": 2.108778953552246,
      "sampling/sampling_logp_difference/mean": 0.01505441963672638,
      "step": 19,
      "step_time": 298.8261763001792
    },
    {
      "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": 0.0221354179084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3009.0,
      "completions/mean_length": 698.8724365234375,
      "completions/mean_terminated_length": 645.153076171875,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "entropy": 0.21506994613446295,
      "epoch": 0.29411764705882354,
      "frac_reward_zero_std": 0.4296875,
      "grad_norm": 0.09299737935420878,
      "learning_rate": 2.924623915920992e-06,
      "loss": 0.0199,
      "num_tokens": 22553890.0,
      "reward": 0.3893229365348816,
      "reward_std": 0.48775550723075867,
      "rewards/reward_correctness/mean": 0.3893229067325592,
      "rewards/reward_correctness/std": 0.48775556683540344,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000158548355103,
      "sampling/importance_sampling_ratio/min": 0.051003288477659225,
      "sampling/sampling_logp_difference/max": 2.975865125656128,
      "sampling/sampling_logp_difference/mean": 0.01541055552661419,
      "step": 20,
      "step_time": 296.1154873804189
    },
    {
      "epoch": 0.29411764705882354,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.00992063521629288,
      "eval_completions/max_length": 1160.6309523809523,
      "eval_completions/max_terminated_length": 1036.6785714285713,
      "eval_completions/mean_length": 563.2639083862305,
      "eval_completions/mean_terminated_length": 538.8448602585565,
      "eval_completions/min_length": 253.11904761904762,
      "eval_completions/min_terminated_length": 253.11904761904762,
      "eval_entropy": 0.16844143558825767,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 22553890.0,
      "eval_reward": 0.5218254091838995,
      "eval_reward_std": 0.47886995829287027,
      "eval_rewards/reward_correctness/mean": 0.5218254020881086,
      "eval_rewards/reward_correctness/std": 0.47886995545455385,
      "eval_runtime": 684.1716,
      "eval_samples_per_second": 0.731,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9925182844911303,
      "eval_sampling/importance_sampling_ratio/min": 0.5034990778991154,
      "eval_sampling/sampling_logp_difference/max": 2.5275759952408925,
      "eval_sampling/sampling_logp_difference/mean": 0.04347030202015525,
      "eval_steps_per_second": 0.123,
      "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": 0.0319010429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3067.0,
      "completions/mean_length": 783.7174682617188,
      "completions/mean_terminated_length": 708.4525756835938,
      "completions/min_length": 169.0,
      "completions/min_terminated_length": 169.0,
      "entropy": 0.21329337812494487,
      "epoch": 0.3088235294117647,
      "frac_reward_zero_std": 0.4453125,
      "grad_norm": 0.08555682284566928,
      "learning_rate": 2.913591755060004e-06,
      "loss": 0.0163,
      "num_tokens": 23924636.0,
      "reward": 0.3626302182674408,
      "reward_std": 0.4809158742427826,
      "rewards/reward_correctness/mean": 0.3626302182674408,
      "rewards/reward_correctness/std": 0.480915904045105,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000237226486206,
      "sampling/importance_sampling_ratio/min": 0.17430639266967773,
      "sampling/sampling_logp_difference/max": 1.7469406127929688,
      "sampling/sampling_logp_difference/mean": 0.015152385458350182,
      "step": 21,
      "step_time": 323.7390418453142
    },
    {
      "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": 0.0345052108168602,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3058.0,
      "completions/mean_length": 776.927734375,
      "completions/mean_terminated_length": 694.9056396484375,
      "completions/min_length": 182.0,
      "completions/min_terminated_length": 182.0,
      "entropy": 0.20884651492815465,
      "epoch": 0.3235294117647059,
      "frac_reward_zero_std": 0.4453125,
      "grad_norm": 0.08504344406668593,
      "learning_rate": 2.901832914340062e-06,
      "loss": 0.0145,
      "num_tokens": 25266329.0,
      "reward": 0.3463541865348816,
      "reward_std": 0.4759626090526581,
      "rewards/reward_correctness/mean": 0.3463541567325592,
      "rewards/reward_correctness/std": 0.47596266865730286,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999927818775177,
      "sampling/importance_sampling_ratio/min": 0.10210206359624863,
      "sampling/sampling_logp_difference/max": 2.281782388687134,
      "sampling/sampling_logp_difference/mean": 0.015026840381324291,
      "step": 22,
      "step_time": 308.0862176595256
    },
    {
      "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": 0.0221354179084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3059.0,
      "completions/mean_length": 682.2506713867188,
      "completions/mean_terminated_length": 628.1550903320312,
      "completions/min_length": 150.0,
      "completions/min_terminated_length": 150.0,
      "entropy": 0.18090229877270758,
      "epoch": 0.3382352941176471,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.08417650985796771,
      "learning_rate": 2.889354156161033e-06,
      "loss": 0.0037,
      "num_tokens": 26476398.0,
      "reward": 0.40234375,
      "reward_std": 0.49053022265434265,
      "rewards/reward_correctness/mean": 0.40234375,
      "rewards/reward_correctness/std": 0.49053022265434265,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9974940419197083,
      "sampling/importance_sampling_ratio/min": 2.3660272959608042e-14,
      "sampling/sampling_logp_difference/max": 31.37497901916504,
      "sampling/sampling_logp_difference/mean": 0.0386495515704155,
      "step": 23,
      "step_time": 333.6117503368296
    },
    {
      "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": 0.02083333395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2987.0,
      "completions/mean_length": 681.4134521484375,
      "completions/mean_terminated_length": 630.5498657226562,
      "completions/min_length": 158.0,
      "completions/min_terminated_length": 158.0,
      "entropy": 0.1827188755851239,
      "epoch": 0.35294117647058826,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.0813981211251326,
      "learning_rate": 2.876162656940614e-06,
      "loss": 0.0063,
      "num_tokens": 27691781.0,
      "reward": 0.4134114682674408,
      "reward_std": 0.49260568618774414,
      "rewards/reward_correctness/mean": 0.4134114682674408,
      "rewards/reward_correctness/std": 0.49260571599006653,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999978244304657,
      "sampling/importance_sampling_ratio/min": 0.033229220658540726,
      "sampling/sampling_logp_difference/max": 3.4043257236480713,
      "sampling/sampling_logp_difference/mean": 0.013785876333713531,
      "step": 24,
      "step_time": 318.87752299709246
    },
    {
      "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": 0.0162760429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2978.0,
      "completions/mean_length": 664.8314208984375,
      "completions/mean_terminated_length": 625.0039672851562,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "entropy": 0.17039690352976322,
      "epoch": 0.36764705882352944,
      "frac_reward_zero_std": 0.4609375,
      "grad_norm": 0.08972646850895903,
      "learning_rate": 2.862266002987244e-06,
      "loss": 0.0086,
      "num_tokens": 28863442.0,
      "reward": 0.4075520932674408,
      "reward_std": 0.4915390610694885,
      "rewards/reward_correctness/mean": 0.4075520932674408,
      "rewards/reward_correctness/std": 0.4915390610694885,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999510645866394,
      "sampling/importance_sampling_ratio/min": 0.050326716154813766,
      "sampling/sampling_logp_difference/max": 2.9892191886901855,
      "sampling/sampling_logp_difference/mean": 0.013295136392116547,
      "step": 25,
      "step_time": 308.65431338362396
    },
    {
      "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": 0.01953125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 694.8568115234375,
      "completions/mean_terminated_length": 647.5033569335938,
      "completions/min_length": 129.0,
      "completions/min_terminated_length": 129.0,
      "entropy": 0.1789090399397537,
      "epoch": 0.38235294117647056,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.09002986293169273,
      "learning_rate": 2.847672186137282e-06,
      "loss": 0.014,
      "num_tokens": 30105822.0,
      "reward": 0.359375,
      "reward_std": 0.47997352480888367,
      "rewards/reward_correctness/mean": 0.359375,
      "rewards/reward_correctness/std": 0.47997352480888367,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999814033508301,
      "sampling/importance_sampling_ratio/min": 0.10588235408067703,
      "sampling/sampling_logp_difference/max": 2.4165873527526855,
      "sampling/sampling_logp_difference/mean": 0.013810048811137676,
      "step": 26,
      "step_time": 317.5119425994344
    },
    {
      "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": 0.02669270895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2983.0,
      "completions/mean_length": 712.9296875,
      "completions/mean_terminated_length": 648.2327880859375,
      "completions/min_length": 114.0,
      "completions/min_terminated_length": 114.0,
      "entropy": 0.17184418614488095,
      "epoch": 0.39705882352941174,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.09041602003319889,
      "learning_rate": 2.8323895991589866e-06,
      "loss": 0.0023,
      "num_tokens": 31355742.0,
      "reward": 0.4388020932674408,
      "reward_std": 0.4964022934436798,
      "rewards/reward_correctness/mean": 0.4388020932674408,
      "rewards/reward_correctness/std": 0.4964022934436798,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000118017196655,
      "sampling/importance_sampling_ratio/min": 0.03874439746141434,
      "sampling/sampling_logp_difference/max": 3.2507691383361816,
      "sampling/sampling_logp_difference/mean": 0.013423329219222069,
      "step": 27,
      "step_time": 312.58686931058764
    },
    {
      "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": 0.01692708395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2944.0,
      "completions/mean_length": 640.52734375,
      "completions/mean_terminated_length": 598.6609497070312,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "entropy": 0.1618180574150756,
      "epoch": 0.4117647058823529,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.09128044036460715,
      "learning_rate": 2.8164270309259034e-06,
      "loss": 0.0073,
      "num_tokens": 32496000.0,
      "reward": 0.44921875,
      "reward_std": 0.4975765645503998,
      "rewards/reward_correctness/mean": 0.44921875,
      "rewards/reward_correctness/std": 0.4975765645503998,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9966862201690674,
      "sampling/importance_sampling_ratio/min": 2.0281311352565723e-12,
      "sampling/sampling_logp_difference/max": 26.923906326293945,
      "sampling/sampling_logp_difference/mean": 0.04643641412258148,
      "step": 28,
      "step_time": 353.90896678948775
    },
    {
      "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": 0.012369791977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2864.0,
      "completions/mean_length": 649.8665771484375,
      "completions/mean_terminated_length": 619.5299682617188,
      "completions/min_length": 142.0,
      "completions/min_terminated_length": 142.0,
      "entropy": 0.17093808809295297,
      "epoch": 0.4264705882352941,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.08931129502738985,
      "learning_rate": 2.79979366136247e-06,
      "loss": 0.0086,
      "num_tokens": 33646175.0,
      "reward": 0.37890625,
      "reward_std": 0.48527270555496216,
      "rewards/reward_correctness/mean": 0.37890625,
      "rewards/reward_correctness/std": 0.48527273535728455,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999492764472961,
      "sampling/importance_sampling_ratio/min": 0.07345923781394958,
      "sampling/sampling_logp_difference/max": 2.6110246181488037,
      "sampling/sampling_logp_difference/mean": 0.013474516570568085,
      "step": 29,
      "step_time": 295.66233261767775
    },
    {
      "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": 0.01953125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3038.0,
      "completions/mean_length": 661.0436401367188,
      "completions/mean_terminated_length": 613.0166015625,
      "completions/min_length": 137.0,
      "completions/min_terminated_length": 137.0,
      "entropy": 0.16064971894957125,
      "epoch": 0.4411764705882353,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.09301717877462062,
      "learning_rate": 2.7824990561647276e-06,
      "loss": 0.0064,
      "num_tokens": 34807266.0,
      "reward": 0.4016927182674408,
      "reward_std": 0.4904000461101532,
      "rewards/reward_correctness/mean": 0.4016927182674408,
      "rewards/reward_correctness/std": 0.4904000759124756,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000295639038086,
      "sampling/importance_sampling_ratio/min": 4.900506974081509e-05,
      "sampling/sampling_logp_difference/max": 9.92358684539795,
      "sampling/sampling_logp_difference/mean": 0.0127759650349617,
      "step": 30,
      "step_time": 331.1905814567581
    },
    {
      "epoch": 0.4411764705882353,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.023809524519102915,
      "eval_completions/max_length": 1353.3809523809523,
      "eval_completions/max_terminated_length": 1045.2619047619048,
      "eval_completions/mean_length": 598.2460470653716,
      "eval_completions/mean_terminated_length": 538.2575521923247,
      "eval_completions/min_length": 258.3095238095238,
      "eval_completions/min_terminated_length": 258.3095238095238,
      "eval_entropy": 0.1339410948788836,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 34807266.0,
      "eval_reward": 0.5178571552747772,
      "eval_reward_std": 0.49191097028198694,
      "eval_rewards/reward_correctness/mean": 0.517857147469407,
      "eval_rewards/reward_correctness/std": 0.4919109692176183,
      "eval_runtime": 806.339,
      "eval_samples_per_second": 0.62,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9940516991274697,
      "eval_sampling/importance_sampling_ratio/min": 0.5017738665143648,
      "eval_sampling/sampling_logp_difference/max": 2.459620109626225,
      "eval_sampling/sampling_logp_difference/mean": 0.0351627429370724,
      "eval_steps_per_second": 0.104,
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2916.0,
      "completions/mean_length": 652.9173583984375,
      "completions/mean_terminated_length": 606.3656616210938,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "entropy": 0.1553694016765803,
      "epoch": 0.45588235294117646,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.0903234273155232,
      "learning_rate": 2.7645531612991763e-06,
      "loss": 0.0091,
      "num_tokens": 35967767.0,
      "reward": 0.435546875,
      "reward_std": 0.4959898591041565,
      "rewards/reward_correctness/mean": 0.435546875,
      "rewards/reward_correctness/std": 0.4959898591041565,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999983012676239,
      "sampling/importance_sampling_ratio/min": 0.036659516394138336,
      "sampling/sampling_logp_difference/max": 3.306082248687744,
      "sampling/sampling_logp_difference/mean": 0.01269611157476902,
      "step": 31,
      "step_time": 314.0296438266523
    },
    {
      "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": 0.008463541977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3054.0,
      "completions/mean_length": 680.3131713867188,
      "completions/mean_terminated_length": 659.8982543945312,
      "completions/min_length": 124.0,
      "completions/min_terminated_length": 124.0,
      "entropy": 0.1594469680567272,
      "epoch": 0.47058823529411764,
      "frac_reward_zero_std": 0.5390625,
      "grad_norm": 0.08553326807932113,
      "learning_rate": 2.745966297282944e-06,
      "loss": 0.0035,
      "num_tokens": 37146348.0,
      "reward": 0.3802083432674408,
      "reward_std": 0.4855959713459015,
      "rewards/reward_correctness/mean": 0.3802083432674408,
      "rewards/reward_correctness/std": 0.4855960011482239,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000026822090149,
      "sampling/importance_sampling_ratio/min": 0.004895252175629139,
      "sampling/sampling_logp_difference/max": 5.319489479064941,
      "sampling/sampling_logp_difference/mean": 0.012724189087748528,
      "step": 32,
      "step_time": 284.7576134908013
    },
    {
      "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": 0.02083333395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3037.0,
      "completions/mean_length": 649.078125,
      "completions/mean_terminated_length": 597.564453125,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "entropy": 0.14795389265054837,
      "epoch": 0.4852941176470588,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.09217934339873941,
      "learning_rate": 2.726749153248549e-06,
      "loss": 0.008,
      "num_tokens": 38312112.0,
      "reward": 0.439453125,
      "reward_std": 0.49648213386535645,
      "rewards/reward_correctness/mean": 0.439453125,
      "rewards/reward_correctness/std": 0.4964821934700012,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999476075172424,
      "sampling/importance_sampling_ratio/min": 0.09122932702302933,
      "sampling/sampling_logp_difference/max": 2.394378900527954,
      "sampling/sampling_logp_difference/mean": 0.012328311800956726,
      "step": 33,
      "step_time": 312.3779431232251
    },
    {
      "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": 0.012369791977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2798.0,
      "completions/mean_length": 613.1237182617188,
      "completions/mean_terminated_length": 582.3269653320312,
      "completions/min_length": 146.0,
      "completions/min_terminated_length": 146.0,
      "entropy": 0.1546164610190317,
      "epoch": 0.5,
      "frac_reward_zero_std": 0.4609375,
      "grad_norm": 0.10097534688518955,
      "learning_rate": 2.706912780796687e-06,
      "loss": 0.0144,
      "num_tokens": 39437566.0,
      "reward": 0.46875,
      "reward_std": 0.4991849958896637,
      "rewards/reward_correctness/mean": 0.46875,
      "rewards/reward_correctness/std": 0.4991849958896637,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000132322311401,
      "sampling/importance_sampling_ratio/min": 0.0012084051268175244,
      "sampling/sampling_logp_difference/max": 6.718453884124756,
      "sampling/sampling_logp_difference/mean": 0.012729411944746971,
      "step": 34,
      "step_time": 299.6743005514145
    },
    {
      "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": 0.009765625,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2971.0,
      "completions/mean_length": 645.4049682617188,
      "completions/mean_terminated_length": 621.4740600585938,
      "completions/min_length": 161.0,
      "completions/min_terminated_length": 161.0,
      "entropy": 0.15254527825163677,
      "epoch": 0.5147058823529411,
      "frac_reward_zero_std": 0.390625,
      "grad_norm": 0.09941734222697605,
      "learning_rate": 2.686468587640551e-06,
      "loss": 0.0153,
      "num_tokens": 40596032.0,
      "reward": 0.4173177182674408,
      "reward_std": 0.49327680468559265,
      "rewards/reward_correctness/mean": 0.4173177182674408,
      "rewards/reward_correctness/std": 0.49327683448791504,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000033378601074,
      "sampling/importance_sampling_ratio/min": 0.06955904513597488,
      "sampling/sampling_logp_difference/max": 2.665579319000244,
      "sampling/sampling_logp_difference/mean": 0.012409794144332409,
      "step": 35,
      "step_time": 309.1410033633001
    },
    {
      "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": 0.03059895895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2755.0,
      "completions/mean_length": 697.9694213867188,
      "completions/mean_terminated_length": 623.3291015625,
      "completions/min_length": 111.0,
      "completions/min_terminated_length": 111.0,
      "entropy": 0.16129665705375373,
      "epoch": 0.5294117647058824,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.08853224454748279,
      "learning_rate": 2.6654283310453644e-06,
      "loss": 0.0129,
      "num_tokens": 41840481.0,
      "reward": 0.41015625,
      "reward_std": 0.4920220673084259,
      "rewards/reward_correctness/mean": 0.41015625,
      "rewards/reward_correctness/std": 0.4920220673084259,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000028133392334,
      "sampling/importance_sampling_ratio/min": 0.054633643478155136,
      "sampling/sampling_logp_difference/max": 2.9071054458618164,
      "sampling/sampling_logp_difference/mean": 0.012762945145368576,
      "step": 36,
      "step_time": 331.5215606596321
    },
    {
      "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": 0.014322916977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2977.0,
      "completions/mean_length": 650.126953125,
      "completions/mean_terminated_length": 615.03369140625,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "entropy": 0.13540291565004736,
      "epoch": 0.5441176470588235,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.09558218428111698,
      "learning_rate": 2.643804111066888e-06,
      "loss": 0.0004,
      "num_tokens": 42998640.0,
      "reward": 0.4524739682674408,
      "reward_std": 0.49789825081825256,
      "rewards/reward_correctness/mean": 0.4524739682674408,
      "rewards/reward_correctness/std": 0.49789825081825256,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9963769316673279,
      "sampling/importance_sampling_ratio/min": 2.2342668740924247e-15,
      "sampling/sampling_logp_difference/max": 33.73486328125,
      "sampling/sampling_logp_difference/mean": 0.04356236383318901,
      "step": 37,
      "step_time": 318.7183957109228
    },
    {
      "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": 0.013671875,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3070.0,
      "completions/mean_length": 639.828125,
      "completions/mean_terminated_length": 606.1597290039062,
      "completions/min_length": 107.0,
      "completions/min_terminated_length": 107.0,
      "entropy": 0.14192722243024036,
      "epoch": 0.5588235294117647,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.09128377744038438,
      "learning_rate": 2.6216083635927896e-06,
      "loss": 0.0145,
      "num_tokens": 44127948.0,
      "reward": 0.4055989682674408,
      "reward_std": 0.49116745591163635,
      "rewards/reward_correctness/mean": 0.4055989682674408,
      "rewards/reward_correctness/std": 0.49116748571395874,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999954104423523,
      "sampling/importance_sampling_ratio/min": 0.12891176342964172,
      "sampling/sampling_logp_difference/max": 2.07913875579834,
      "sampling/sampling_logp_difference/mean": 0.0118638901039958,
      "step": 38,
      "step_time": 291.72405133536085
    },
    {
      "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": 0.01106770895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3021.0,
      "completions/mean_length": 651.6712646484375,
      "completions/mean_terminated_length": 624.5839233398438,
      "completions/min_length": 122.0,
      "completions/min_terminated_length": 122.0,
      "entropy": 0.144097798562143,
      "epoch": 0.5735294117647058,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.09866853100849818,
      "learning_rate": 2.598853853190882e-06,
      "loss": 0.0117,
      "num_tokens": 45284507.0,
      "reward": 0.4303385615348816,
      "reward_std": 0.4952847361564636,
      "rewards/reward_correctness/mean": 0.4303385317325592,
      "rewards/reward_correctness/std": 0.495284765958786,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000274181365967,
      "sampling/importance_sampling_ratio/min": 0.07444321364164352,
      "sampling/sampling_logp_difference/max": 2.5977187156677246,
      "sampling/sampling_logp_difference/mean": 0.012085234746336937,
      "step": 39,
      "step_time": 307.95729614887387
    },
    {
      "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": 0.010416666977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2808.0,
      "completions/mean_length": 647.4193115234375,
      "completions/mean_terminated_length": 622.0487060546875,
      "completions/min_length": 126.0,
      "completions/min_terminated_length": 126.0,
      "entropy": 0.14093391405185685,
      "epoch": 0.5882352941176471,
      "frac_reward_zero_std": 0.484375,
      "grad_norm": 0.09003847468670717,
      "learning_rate": 2.5755536657683354e-06,
      "loss": 0.0038,
      "num_tokens": 46434355.0,
      "reward": 0.4361979365348816,
      "reward_std": 0.49607405066490173,
      "rewards/reward_correctness/mean": 0.4361979067325592,
      "rewards/reward_correctness/std": 0.4960741102695465,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999783635139465,
      "sampling/importance_sampling_ratio/min": 0.0060172248631715775,
      "sampling/sampling_logp_difference/max": 5.113129138946533,
      "sampling/sampling_logp_difference/mean": 0.011764663271605968,
      "step": 40,
      "step_time": 293.8941194838844
    },
    {
      "epoch": 0.5882352941176471,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.007936508173034304,
      "eval_completions/max_length": 1165.6904761904761,
      "eval_completions/max_terminated_length": 1080.047619047619,
      "eval_completions/mean_length": 574.1071610223679,
      "eval_completions/mean_terminated_length": 554.6643026442755,
      "eval_completions/min_length": 256.3452380952381,
      "eval_completions/min_terminated_length": 256.3452380952381,
      "eval_entropy": 0.1198243172395797,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 46434355.0,
      "eval_reward": 0.5138889015430496,
      "eval_reward_std": 0.4920797245133491,
      "eval_rewards/reward_correctness/mean": 0.5138888944472585,
      "eval_rewards/reward_correctness/std": 0.4920797199010849,
      "eval_runtime": 686.0979,
      "eval_samples_per_second": 0.729,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9935774937981651,
      "eval_sampling/importance_sampling_ratio/min": 0.5091292620414779,
      "eval_sampling/sampling_logp_difference/max": 2.236756165822347,
      "eval_sampling/sampling_logp_difference/mean": 0.03119110870396807,
      "eval_steps_per_second": 0.122,
      "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": 0.014322916977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2812.0,
      "completions/mean_length": 638.0228271484375,
      "completions/mean_terminated_length": 602.69287109375,
      "completions/min_length": 133.0,
      "completions/min_terminated_length": 133.0,
      "entropy": 0.14125462470110506,
      "epoch": 0.6029411764705882,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.0900402655314074,
      "learning_rate": 2.551721201046098e-06,
      "loss": 0.0076,
      "num_tokens": 47570106.0,
      "reward": 0.421875,
      "reward_std": 0.49401959776878357,
      "rewards/reward_correctness/mean": 0.421875,
      "rewards/reward_correctness/std": 0.49401959776878357,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999483227729797,
      "sampling/importance_sampling_ratio/min": 0.0491253063082695,
      "sampling/sampling_logp_difference/max": 3.013381004333496,
      "sampling/sampling_logp_difference/mean": 0.011780498549342155,
      "step": 41,
      "step_time": 294.60718926927075
    },
    {
      "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": 0.01692708395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2950.0,
      "completions/mean_length": 682.486328125,
      "completions/mean_terminated_length": 641.3424072265625,
      "completions/min_length": 162.0,
      "completions/min_terminated_length": 162.0,
      "entropy": 0.13389375229598954,
      "epoch": 0.6176470588235294,
      "frac_reward_zero_std": 0.515625,
      "grad_norm": 0.08930218418218254,
      "learning_rate": 2.5273701648528393e-06,
      "loss": 0.0104,
      "num_tokens": 48771609.0,
      "reward": 0.439453125,
      "reward_std": 0.49648216366767883,
      "rewards/reward_correctness/mean": 0.439453125,
      "rewards/reward_correctness/std": 0.4964821934700012,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999879002571106,
      "sampling/importance_sampling_ratio/min": 0.04630628600716591,
      "sampling/sampling_logp_difference/max": 3.0724775791168213,
      "sampling/sampling_logp_difference/mean": 0.011399103328585625,
      "step": 42,
      "step_time": 306.2859174082987
    },
    {
      "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": 0.012369791977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2893.0,
      "completions/mean_length": 634.833984375,
      "completions/mean_terminated_length": 604.3091430664062,
      "completions/min_length": 127.0,
      "completions/min_terminated_length": 127.0,
      "entropy": 0.13599129632348195,
      "epoch": 0.6323529411764706,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.08873768343136257,
      "learning_rate": 2.5025145612428566e-06,
      "loss": 0.0022,
      "num_tokens": 49918734.0,
      "reward": 0.4166666865348816,
      "reward_std": 0.4931671619415283,
      "rewards/reward_correctness/mean": 0.4166666567325592,
      "rewards/reward_correctness/std": 0.4931672215461731,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999739527702332,
      "sampling/importance_sampling_ratio/min": 0.04011012613773346,
      "sampling/sampling_logp_difference/max": 3.2161264419555664,
      "sampling/sampling_logp_difference/mean": 0.011722831055521965,
      "step": 43,
      "step_time": 321.6622120910324
    },
    {
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2979.0,
      "completions/mean_length": 688.9193115234375,
      "completions/mean_terminated_length": 643.0603637695312,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "entropy": 0.13310185814043507,
      "epoch": 0.6470588235294118,
      "frac_reward_zero_std": 0.4296875,
      "grad_norm": 0.09454388235202577,
      "learning_rate": 2.47716868444247e-06,
      "loss": 0.0114,
      "num_tokens": 51156518.0,
      "reward": 0.4361979365348816,
      "reward_std": 0.49607405066490173,
      "rewards/reward_correctness/mean": 0.4361979067325592,
      "rewards/reward_correctness/std": 0.4960741102695465,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999701976776123,
      "sampling/importance_sampling_ratio/min": 0.02935052663087845,
      "sampling/sampling_logp_difference/max": 3.528444766998291,
      "sampling/sampling_logp_difference/mean": 0.011353356763720512,
      "step": 44,
      "step_time": 326.7605273928493
    },
    {
      "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": 0.0201822929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2984.0,
      "completions/mean_length": 672.0306396484375,
      "completions/mean_terminated_length": 622.5960083007812,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "entropy": 0.12670818017795682,
      "epoch": 0.6617647058823529,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.08260192050872966,
      "learning_rate": 2.4513471106295523e-06,
      "loss": 0.0051,
      "num_tokens": 52362397.0,
      "reward": 0.4440104365348816,
      "reward_std": 0.4970170855522156,
      "rewards/reward_correctness/mean": 0.4440104067325592,
      "rewards/reward_correctness/std": 0.49701711535453796,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000724792480469,
      "sampling/importance_sampling_ratio/min": 0.09955490380525589,
      "sampling/sampling_logp_difference/max": 2.3070459365844727,
      "sampling/sampling_logp_difference/mean": 0.011110104620456696,
      "step": 45,
      "step_time": 311.18877177499235
    },
    {
      "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": 0.01953125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2785.0,
      "completions/mean_length": 703.7669677734375,
      "completions/mean_terminated_length": 656.5910034179688,
      "completions/min_length": 120.0,
      "completions/min_terminated_length": 120.0,
      "entropy": 0.12684163701487705,
      "epoch": 0.6764705882352942,
      "frac_reward_zero_std": 0.515625,
      "grad_norm": 0.09331266808858216,
      "learning_rate": 2.4250646895508992e-06,
      "loss": 0.0032,
      "num_tokens": 53619579.0,
      "reward": 0.3912760615348816,
      "reward_std": 0.48819488286972046,
      "rewards/reward_correctness/mean": 0.3912760317325592,
      "rewards/reward_correctness/std": 0.48819491267204285,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0001139640808105,
      "sampling/importance_sampling_ratio/min": 0.06684955954551697,
      "sampling/sampling_logp_difference/max": 2.705310583114624,
      "sampling/sampling_logp_difference/mean": 0.011112010106444359,
      "step": 46,
      "step_time": 313.3187058418989
    },
    {
      "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": 0.0240885429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2982.0,
      "completions/mean_length": 668.234375,
      "completions/mean_terminated_length": 609.0033569335938,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "entropy": 0.11769095557974651,
      "epoch": 0.6911764705882353,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.09763888479973193,
      "learning_rate": 2.3983365359822804e-06,
      "loss": 0.0065,
      "num_tokens": 54811371.0,
      "reward": 0.4407552182674408,
      "reward_std": 0.49663931131362915,
      "rewards/reward_correctness/mean": 0.4407552182674408,
      "rewards/reward_correctness/std": 0.49663934111595154,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9966242909431458,
      "sampling/importance_sampling_ratio/min": 1.1548664751366232e-17,
      "sampling/sampling_logp_difference/max": 38.999961853027344,
      "sampling/sampling_logp_difference/mean": 0.040139876306056976,
      "step": 47,
      "step_time": 321.21418457012624
    },
    {
      "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": 0.021484375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2728.0,
      "completions/mean_length": 687.427734375,
      "completions/mean_terminated_length": 635.0718383789062,
      "completions/min_length": 107.0,
      "completions/min_terminated_length": 107.0,
      "entropy": 0.12559914664598182,
      "epoch": 0.7058823529411765,
      "frac_reward_zero_std": 0.5390625,
      "grad_norm": 0.09487698141413321,
      "learning_rate": 2.3711780210360726e-06,
      "loss": 0.0109,
      "num_tokens": 56026512.0,
      "reward": 0.3912760615348816,
      "reward_std": 0.48819488286972046,
      "rewards/reward_correctness/mean": 0.3912760317325592,
      "rewards/reward_correctness/std": 0.48819491267204285,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000499486923218,
      "sampling/importance_sampling_ratio/min": 1.1616037227213383e-05,
      "sampling/sampling_logp_difference/max": 11.363123893737793,
      "sampling/sampling_logp_difference/mean": 0.010995877906680107,
      "step": 48,
      "step_time": 309.3666119822301
    },
    {
      "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": 0.017578125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2970.0,
      "completions/mean_length": 698.48046875,
      "completions/mean_terminated_length": 656.011962890625,
      "completions/min_length": 157.0,
      "completions/min_terminated_length": 157.0,
      "entropy": 0.12552364665316418,
      "epoch": 0.7205882352941176,
      "frac_reward_zero_std": 0.484375,
      "grad_norm": 0.09602299763561906,
      "learning_rate": 2.343604763321476e-06,
      "loss": 0.0123,
      "num_tokens": 57251862.0,
      "reward": 0.4134114682674408,
      "reward_std": 0.49260571599006653,
      "rewards/reward_correctness/mean": 0.4134114682674408,
      "rewards/reward_correctness/std": 0.49260571599006653,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000723600387573,
      "sampling/importance_sampling_ratio/min": 0.0009446050971746445,
      "sampling/sampling_logp_difference/max": 6.964743614196777,
      "sampling/sampling_logp_difference/mean": 0.010939620435237885,
      "step": 49,
      "step_time": 297.2958505139686
    },
    {
      "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": 0.0162760429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2870.0,
      "completions/mean_length": 650.4967651367188,
      "completions/mean_terminated_length": 610.43212890625,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "entropy": 0.11314014665549621,
      "epoch": 0.7352941176470589,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.0970926202181468,
      "learning_rate": 2.3156326199623965e-06,
      "loss": 0.0121,
      "num_tokens": 58399213.0,
      "reward": 0.4583333432674408,
      "reward_std": 0.49842309951782227,
      "rewards/reward_correctness/mean": 0.4583333432674408,
      "rewards/reward_correctness/std": 0.49842312932014465,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999960720539093,
      "sampling/importance_sampling_ratio/min": 0.06841433048248291,
      "sampling/sampling_logp_difference/max": 2.682173013687134,
      "sampling/sampling_logp_difference/mean": 0.010027723386883736,
      "step": 50,
      "step_time": 305.2847441327758
    },
    {
      "epoch": 0.7352941176470589,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.023809524519102915,
      "eval_completions/max_length": 1246.345238095238,
      "eval_completions/max_terminated_length": 1003.8809523809524,
      "eval_completions/mean_length": 589.6845405215308,
      "eval_completions/mean_terminated_length": 530.4769995553153,
      "eval_completions/min_length": 251.0,
      "eval_completions/min_terminated_length": 251.0,
      "eval_entropy": 0.10575014485844544,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 58399213.0,
      "eval_reward": 0.5000000120628447,
      "eval_reward_std": 0.4889630675315857,
      "eval_rewards/reward_correctness/mean": 0.5000000056766328,
      "eval_rewards/reward_correctness/std": 0.4889630675315857,
      "eval_runtime": 741.191,
      "eval_samples_per_second": 0.675,
      "eval_sampling/importance_sampling_ratio/max": 2.990457282179878,
      "eval_sampling/importance_sampling_ratio/mean": 0.9947371142251151,
      "eval_sampling/importance_sampling_ratio/min": 0.5308067000338009,
      "eval_sampling/sampling_logp_difference/max": 2.112378262338184,
      "eval_sampling/sampling_logp_difference/mean": 0.027818603640688316,
      "eval_steps_per_second": 0.113,
      "step": 50
    },
    {
      "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": 0.01302083395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3064.0,
      "completions/mean_length": 648.2448120117188,
      "completions/mean_terminated_length": 616.2691650390625,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "entropy": 0.1220089640119113,
      "epoch": 0.75,
      "frac_reward_zero_std": 0.3984375,
      "grad_norm": 0.10917632380343659,
      "learning_rate": 2.2872776774781627e-06,
      "loss": -0.0,
      "num_tokens": 59534657.0,
      "reward": 0.4407552182674408,
      "reward_std": 0.49663931131362915,
      "rewards/reward_correctness/mean": 0.4407552182674408,
      "rewards/reward_correctness/std": 0.49663934111595154,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999938011169434,
      "sampling/importance_sampling_ratio/min": 0.0287653636187315,
      "sampling/sampling_logp_difference/max": 3.5485832691192627,
      "sampling/sampling_logp_difference/mean": 0.010647393763065338,
      "step": 51,
      "step_time": 286.852661145851
    },
    {
      "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": 0.01302083395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3002.0,
      "completions/mean_length": 652.33984375,
      "completions/mean_terminated_length": 620.5264282226562,
      "completions/min_length": 116.0,
      "completions/min_terminated_length": 116.0,
      "entropy": 0.11450612859334797,
      "epoch": 0.7647058823529411,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.11777736125448364,
      "learning_rate": 2.258556242532317e-06,
      "loss": 0.0067,
      "num_tokens": 60706655.0,
      "reward": 0.4309895932674408,
      "reward_std": 0.49537593126296997,
      "rewards/reward_correctness/mean": 0.4309895932674408,
      "rewards/reward_correctness/std": 0.49537593126296997,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9964901804924011,
      "sampling/importance_sampling_ratio/min": 1.4169495366936997e-16,
      "sampling/sampling_logp_difference/max": 36.492855072021484,
      "sampling/sampling_logp_difference/mean": 0.04059432074427605,
      "step": 52,
      "step_time": 334.9317989842966
    },
    {
      "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": 0.013671875,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2735.0,
      "completions/mean_length": 650.4303588867188,
      "completions/mean_terminated_length": 616.8640747070312,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "entropy": 0.12498974212212488,
      "epoch": 0.7794117647058824,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.09996053573488595,
      "learning_rate": 2.2294848325548066e-06,
      "loss": 0.0031,
      "num_tokens": 61861008.0,
      "reward": 0.3795573115348816,
      "reward_std": 0.48543480038642883,
      "rewards/reward_correctness/mean": 0.3795572817325592,
      "rewards/reward_correctness/std": 0.4854348301887512,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999858736991882,
      "sampling/importance_sampling_ratio/min": 0.014357448555529118,
      "sampling/sampling_logp_difference/max": 4.243486404418945,
      "sampling/sampling_logp_difference/mean": 0.011073033325374126,
      "step": 53,
      "step_time": 285.01819738186896
    },
    {
      "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": 0.0162760429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3031.0,
      "completions/mean_length": 631.8841552734375,
      "completions/mean_terminated_length": 591.5115966796875,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "entropy": 0.10903365240665153,
      "epoch": 0.7941176470588235,
      "frac_reward_zero_std": 0.4375,
      "grad_norm": 0.10210749738034491,
      "learning_rate": 2.200080166242961e-06,
      "loss": 0.0049,
      "num_tokens": 62982650.0,
      "reward": 0.4563802182674408,
      "reward_std": 0.4982558488845825,
      "rewards/reward_correctness/mean": 0.4563802182674408,
      "rewards/reward_correctness/std": 0.4982558786869049,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999967098236084,
      "sampling/importance_sampling_ratio/min": 0.011410914361476898,
      "sampling/sampling_logp_difference/max": 4.473185062408447,
      "sampling/sampling_logp_difference/mean": 0.009591399691998959,
      "step": 54,
      "step_time": 292.70618091057986
    },
    {
      "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": 3072.0,
      "completions/max_terminated_length": 2644.0,
      "completions/mean_length": 683.1940307617188,
      "completions/mean_terminated_length": 645.2764282226562,
      "completions/min_length": 157.0,
      "completions/min_terminated_length": 157.0,
      "entropy": 0.1221256356802769,
      "epoch": 0.8088235294117647,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.09445385336737874,
      "learning_rate": 2.1703591539467283e-06,
      "loss": 0.0127,
      "num_tokens": 64207992.0,
      "reward": 0.380859375,
      "reward_std": 0.4857562184333801,
      "rewards/reward_correctness/mean": 0.380859375,
      "rewards/reward_correctness/std": 0.4857562482357025,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9998800158500671,
      "sampling/importance_sampling_ratio/min": 5.5687261919956654e-05,
      "sampling/sampling_logp_difference/max": 9.795759201049805,
      "sampling/sampling_logp_difference/mean": 0.010790005326271057,
      "step": 55,
      "step_time": 324.3310802578926
    },
    {
      "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": 0.0182291679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2970.0,
      "completions/mean_length": 641.8776245117188,
      "completions/mean_terminated_length": 596.8978881835938,
      "completions/min_length": 89.0,
      "completions/min_terminated_length": 89.0,
      "entropy": 0.11542422254569829,
      "epoch": 0.8235294117647058,
      "frac_reward_zero_std": 0.5859375,
      "grad_norm": 0.08674985428565889,
      "learning_rate": 2.140338887943686e-06,
      "loss": 0.0166,
      "num_tokens": 65349976.0,
      "reward": 0.47265625,
      "reward_std": 0.499414324760437,
      "rewards/reward_correctness/mean": 0.47265625,
      "rewards/reward_correctness/std": 0.4994143545627594,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000611543655396,
      "sampling/importance_sampling_ratio/min": 0.024185113608837128,
      "sampling/sampling_logp_difference/max": 3.722018003463745,
      "sampling/sampling_logp_difference/mean": 0.010218578390777111,
      "step": 56,
      "step_time": 296.37975679989904
    },
    {
      "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": 0.010416666977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2546.0,
      "completions/mean_length": 611.0390625,
      "completions/mean_terminated_length": 585.1342163085938,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "entropy": 0.11121287627611309,
      "epoch": 0.8382352941176471,
      "frac_reward_zero_std": 0.4296875,
      "grad_norm": 0.10503693257109341,
      "learning_rate": 2.110036632609435e-06,
      "loss": 0.019,
      "num_tokens": 66426784.0,
      "reward": 0.4993489682674408,
      "reward_std": 0.5001624226570129,
      "rewards/reward_correctness/mean": 0.4993489682674408,
      "rewards/reward_correctness/std": 0.5001624226570129,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999729990959167,
      "sampling/importance_sampling_ratio/min": 0.0030124566983431578,
      "sampling/sampling_logp_difference/max": 5.804999351501465,
      "sampling/sampling_logp_difference/mean": 0.009916255250573158,
      "step": 57,
      "step_time": 296.1852462873794
    },
    {
      "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": 0.01692708395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3022.0,
      "completions/mean_length": 698.658203125,
      "completions/mean_terminated_length": 657.792724609375,
      "completions/min_length": 129.0,
      "completions/min_terminated_length": 129.0,
      "entropy": 0.11595404433319345,
      "epoch": 0.8529411764705882,
      "frac_reward_zero_std": 0.4765625,
      "grad_norm": 0.09551442970026573,
      "learning_rate": 2.0794698144890156e-06,
      "loss": 0.0119,
      "num_tokens": 67683415.0,
      "reward": 0.427734375,
      "reward_std": 0.49491122364997864,
      "rewards/reward_correctness/mean": 0.427734375,
      "rewards/reward_correctness/std": 0.494911253452301,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9960780143737793,
      "sampling/importance_sampling_ratio/min": 1.8462416315266454e-14,
      "sampling/sampling_logp_difference/max": 31.62303924560547,
      "sampling/sampling_logp_difference/mean": 0.043669816106557846,
      "step": 58,
      "step_time": 339.7408985039219
    },
    {
      "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": 0.01302083395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2949.0,
      "completions/mean_length": 649.1393432617188,
      "completions/mean_terminated_length": 617.1754760742188,
      "completions/min_length": 137.0,
      "completions/min_terminated_length": 137.0,
      "entropy": 0.11917942861327901,
      "epoch": 0.8676470588235294,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.10106429850918942,
      "learning_rate": 2.048656012275064e-06,
      "loss": -0.0005,
      "num_tokens": 68856377.0,
      "reward": 0.3697916865348816,
      "reward_std": 0.48290538787841797,
      "rewards/reward_correctness/mean": 0.3697916567325592,
      "rewards/reward_correctness/std": 0.48290541768074036,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999811053276062,
      "sampling/importance_sampling_ratio/min": 0.09501548111438751,
      "sampling/sampling_logp_difference/max": 2.353715419769287,
      "sampling/sampling_logp_difference/mean": 0.010620813816785812,
      "step": 59,
      "step_time": 312.87796821817756
    },
    {
      "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": 0.029296875,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3066.0,
      "completions/mean_length": 748.1400146484375,
      "completions/mean_terminated_length": 678.0033569335938,
      "completions/min_length": 123.0,
      "completions/min_terminated_length": 123.0,
      "entropy": 0.11793089081766084,
      "epoch": 0.8823529411764706,
      "frac_reward_zero_std": 0.5703125,
      "grad_norm": 0.09465256634616137,
      "learning_rate": 2.017612946698471e-06,
      "loss": 0.0129,
      "num_tokens": 70163368.0,
      "reward": 0.3795573115348816,
      "reward_std": 0.48543480038642883,
      "rewards/reward_correctness/mean": 0.3795572817325592,
      "rewards/reward_correctness/std": 0.4854348599910736,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999856948852539,
      "sampling/importance_sampling_ratio/min": 0.016788704320788383,
      "sampling/sampling_logp_difference/max": 4.0870490074157715,
      "sampling/sampling_logp_difference/mean": 0.010650983080267906,
      "step": 60,
      "step_time": 309.0038933083415
    },
    {
      "epoch": 0.8823529411764706,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.023809524519102915,
      "eval_completions/max_length": 1401.5833333333333,
      "eval_completions/max_terminated_length": 1113.4404761904761,
      "eval_completions/mean_length": 614.5436699276879,
      "eval_completions/mean_terminated_length": 554.2809684390113,
      "eval_completions/min_length": 249.04761904761904,
      "eval_completions/min_terminated_length": 249.04761904761904,
      "eval_entropy": 0.10374800328697477,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 70163368.0,
      "eval_reward": 0.5000000094019231,
      "eval_reward_std": 0.4882907420396805,
      "eval_rewards/reward_correctness/mean": 0.5000000058540276,
      "eval_rewards/reward_correctness/std": 0.48829073849178495,
      "eval_runtime": 832.1384,
      "eval_samples_per_second": 0.601,
      "eval_sampling/importance_sampling_ratio/max": 3.0,
      "eval_sampling/importance_sampling_ratio/mean": 0.9949375064600081,
      "eval_sampling/importance_sampling_ratio/min": 0.5374763036767641,
      "eval_sampling/sampling_logp_difference/max": 2.227688502697718,
      "eval_sampling/sampling_logp_difference/mean": 0.027424226071508156,
      "eval_steps_per_second": 0.101,
      "step": 60
    },
    {
      "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": 0.0260416679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3042.0,
      "completions/mean_length": 666.3509521484375,
      "completions/mean_terminated_length": 602.0287475585938,
      "completions/min_length": 140.0,
      "completions/min_terminated_length": 140.0,
      "entropy": 0.11709785764105618,
      "epoch": 0.8970588235294118,
      "frac_reward_zero_std": 0.5625,
      "grad_norm": 0.09724587353427003,
      "learning_rate": 1.9863584703373534e-06,
      "loss": 0.0079,
      "num_tokens": 71365839.0,
      "reward": 0.41015625,
      "reward_std": 0.4920220375061035,
      "rewards/reward_correctness/mean": 0.41015625,
      "rewards/reward_correctness/std": 0.4920220673084259,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9959919452667236,
      "sampling/importance_sampling_ratio/min": 1.5554298034592538e-18,
      "sampling/sampling_logp_difference/max": 41.00477981567383,
      "sampling/sampling_logp_difference/mean": 0.044307515025138855,
      "step": 61,
      "step_time": 317.7361774868332
    },
    {
      "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": 0.0221354179084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3005.0,
      "completions/mean_length": 715.015625,
      "completions/mean_terminated_length": 661.6617431640625,
      "completions/min_length": 154.0,
      "completions/min_terminated_length": 154.0,
      "entropy": 0.11619744420750067,
      "epoch": 0.9117647058823529,
      "frac_reward_zero_std": 0.4453125,
      "grad_norm": 0.10403468878995832,
      "learning_rate": 1.954910557350202e-06,
      "loss": 0.0137,
      "num_tokens": 72637971.0,
      "reward": 0.3997395932674408,
      "reward_std": 0.49000421166419983,
      "rewards/reward_correctness/mean": 0.3997395932674408,
      "rewards/reward_correctness/std": 0.4900042414665222,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000476837158203,
      "sampling/importance_sampling_ratio/min": 0.028073355555534363,
      "sampling/sampling_logp_difference/max": 3.57293438911438,
      "sampling/sampling_logp_difference/mean": 0.010643940418958664,
      "step": 62,
      "step_time": 316.263129406143
    },
    {
      "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": 0.014322916977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3012.0,
      "completions/mean_length": 668.5345458984375,
      "completions/mean_terminated_length": 633.6096801757812,
      "completions/min_length": 151.0,
      "completions/min_terminated_length": 151.0,
      "entropy": 0.11588948150165379,
      "epoch": 0.9264705882352942,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.10204530460670029,
      "learning_rate": 1.9232872931391114e-06,
      "loss": 0.0135,
      "num_tokens": 73838996.0,
      "reward": 0.3515625,
      "reward_std": 0.4776136577129364,
      "rewards/reward_correctness/mean": 0.3515625,
      "rewards/reward_correctness/std": 0.4776136875152588,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999632239341736,
      "sampling/importance_sampling_ratio/min": 0.0017267765942960978,
      "sampling/sampling_logp_difference/max": 6.361498832702637,
      "sampling/sampling_logp_difference/mean": 0.010555064305663109,
      "step": 63,
      "step_time": 296.9185470682569
    },
    {
      "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": 3072.0,
      "completions/max_terminated_length": 2800.0,
      "completions/mean_length": 668.0931396484375,
      "completions/mean_terminated_length": 629.9358520507812,
      "completions/min_length": 97.0,
      "completions/min_terminated_length": 97.0,
      "entropy": 0.11724561610026285,
      "epoch": 0.9411764705882353,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.09705948499787383,
      "learning_rate": 1.8915068639490344e-06,
      "loss": 0.009,
      "num_tokens": 75027991.0,
      "reward": 0.4264323115348816,
      "reward_std": 0.4947192072868347,
      "rewards/reward_correctness/mean": 0.4264322817325592,
      "rewards/reward_correctness/std": 0.4947192370891571,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999983906745911,
      "sampling/importance_sampling_ratio/min": 0.02149857021868229,
      "sampling/sampling_logp_difference/max": 3.839768886566162,
      "sampling/sampling_logp_difference/mean": 0.010617914609611034,
      "step": 64,
      "step_time": 313.9419138771482
    },
    {
      "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": 0.0201822929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2956.0,
      "completions/mean_length": 672.75390625,
      "completions/mean_terminated_length": 623.4869995117188,
      "completions/min_length": 159.0,
      "completions/min_terminated_length": 159.0,
      "entropy": 0.11667637288337573,
      "epoch": 0.9558823529411765,
      "frac_reward_zero_std": 0.484375,
      "grad_norm": 0.09957814700163607,
      "learning_rate": 1.8595875464090389e-06,
      "loss": 0.0099,
      "num_tokens": 76231045.0,
      "reward": 0.4661458432674408,
      "reward_std": 0.4990150034427643,
      "rewards/reward_correctness/mean": 0.4661458432674408,
      "rewards/reward_correctness/std": 0.49901503324508667,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000202655792236,
      "sampling/importance_sampling_ratio/min": 0.012628194876015186,
      "sampling/sampling_logp_difference/max": 4.371823310852051,
      "sampling/sampling_logp_difference/mean": 0.010494709014892578,
      "step": 65,
      "step_time": 301.15293512120843
    },
    {
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2905.0,
      "completions/mean_length": 651.5514526367188,
      "completions/mean_terminated_length": 604.9734497070312,
      "completions/min_length": 152.0,
      "completions/min_terminated_length": 152.0,
      "entropy": 0.11085655010538176,
      "epoch": 0.9705882352941176,
      "frac_reward_zero_std": 0.515625,
      "grad_norm": 0.0997805532923189,
      "learning_rate": 1.8275476970215906e-06,
      "loss": 0.0033,
      "num_tokens": 77403500.0,
      "reward": 0.4140625,
      "reward_std": 0.4927197992801666,
      "rewards/reward_correctness/mean": 0.4140625,
      "rewards/reward_correctness/std": 0.4927197992801666,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999725222587585,
      "sampling/importance_sampling_ratio/min": 0.11216503381729126,
      "sampling/sampling_logp_difference/max": 2.18778395652771,
      "sampling/sampling_logp_difference/mean": 0.010185856372117996,
      "step": 66,
      "step_time": 305.88186718709767
    },
    {
      "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": 0.012369791977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3066.0,
      "completions/mean_length": 635.2025146484375,
      "completions/mean_terminated_length": 604.6822509765625,
      "completions/min_length": 111.0,
      "completions/min_terminated_length": 111.0,
      "entropy": 0.11111450218595564,
      "epoch": 0.9852941176470589,
      "frac_reward_zero_std": 0.4609375,
      "grad_norm": 0.1004527266463445,
      "learning_rate": 1.7954057416059002e-06,
      "loss": 0.0031,
      "num_tokens": 78520915.0,
      "reward": 0.4296875,
      "reward_std": 0.49519267678260803,
      "rewards/reward_correctness/mean": 0.4296875,
      "rewards/reward_correctness/std": 0.4951927065849304,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999885559082031,
      "sampling/importance_sampling_ratio/min": 0.01389921735972166,
      "sampling/sampling_logp_difference/max": 4.275922775268555,
      "sampling/sampling_logp_difference/mean": 0.009849893860518932,
      "step": 67,
      "step_time": 289.09394181659445
    },
    {
      "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": 0.006510416977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2954.0,
      "completions/mean_length": 591.7311401367188,
      "completions/mean_terminated_length": 575.4777221679688,
      "completions/min_length": 171.0,
      "completions/min_terminated_length": 171.0,
      "entropy": 0.11097419285215437,
      "epoch": 1.0,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.10381810303339536,
      "learning_rate": 1.7631801647014034e-06,
      "loss": 0.0062,
      "num_tokens": 79582934.0,
      "reward": 0.443359375,
      "reward_std": 0.49694323539733887,
      "rewards/reward_correctness/mean": 0.443359375,
      "rewards/reward_correctness/std": 0.49694326519966125,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999539256095886,
      "sampling/importance_sampling_ratio/min": 0.032169442623853683,
      "sampling/sampling_logp_difference/max": 3.4367382526397705,
      "sampling/sampling_logp_difference/mean": 0.009978776797652245,
      "step": 68,
      "step_time": 277.24001722969115
    },
    {
      "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": 0.02669270895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2924.0,
      "completions/mean_length": 658.923828125,
      "completions/mean_terminated_length": 592.8555297851562,
      "completions/min_length": 112.0,
      "completions/min_terminated_length": 112.0,
      "entropy": 0.11762749508488923,
      "epoch": 1.0147058823529411,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.09955363266557021,
      "learning_rate": 1.7308894989374766e-06,
      "loss": 0.0168,
      "num_tokens": 80758973.0,
      "reward": 0.404296875,
      "reward_std": 0.4909152686595917,
      "rewards/reward_correctness/mean": 0.404296875,
      "rewards/reward_correctness/std": 0.49091529846191406,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000025987625122,
      "sampling/importance_sampling_ratio/min": 0.06638438999652863,
      "sampling/sampling_logp_difference/max": 2.7122933864593506,
      "sampling/sampling_logp_difference/mean": 0.010275790467858315,
      "step": 69,
      "step_time": 303.03809165675193
    },
    {
      "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": 0.02473958395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2971.0,
      "completions/mean_length": 651.80859375,
      "completions/mean_terminated_length": 590.4152221679688,
      "completions/min_length": 132.0,
      "completions/min_terminated_length": 132.0,
      "entropy": 0.10798334499122575,
      "epoch": 1.0294117647058822,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.09818298129014348,
      "learning_rate": 1.6985523143754952e-06,
      "loss": 0.0079,
      "num_tokens": 81899183.0,
      "reward": 0.4453125,
      "reward_std": 0.4971621334552765,
      "rewards/reward_correctness/mean": 0.4453125,
      "rewards/reward_correctness/std": 0.4971621334552765,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000056028366089,
      "sampling/importance_sampling_ratio/min": 0.0010057517793029547,
      "sampling/sampling_logp_difference/max": 6.90201997756958,
      "sampling/sampling_logp_difference/mean": 0.009786490350961685,
      "step": 70,
      "step_time": 286.29961240198463
    },
    {
      "epoch": 1.0294117647058822,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.023809524519102915,
      "eval_completions/max_length": 1280.6190476190477,
      "eval_completions/max_terminated_length": 1000.1904761904761,
      "eval_completions/mean_length": 570.4266034080869,
      "eval_completions/mean_terminated_length": 510.04386247907365,
      "eval_completions/min_length": 233.58333333333334,
      "eval_completions/min_terminated_length": 233.58333333333334,
      "eval_entropy": 0.09882747022701162,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 81899183.0,
      "eval_reward": 0.513888899591707,
      "eval_reward_std": 0.4880865791014263,
      "eval_rewards/reward_correctness/mean": 0.5138888939150742,
      "eval_rewards/reward_correctness/std": 0.4880865748439516,
      "eval_runtime": 762.0449,
      "eval_samples_per_second": 0.656,
      "eval_sampling/importance_sampling_ratio/max": 2.9982360771724155,
      "eval_sampling/importance_sampling_ratio/mean": 0.9953853126083102,
      "eval_sampling/importance_sampling_ratio/min": 0.5366758298838422,
      "eval_sampling/sampling_logp_difference/max": 2.222159587201618,
      "eval_sampling/sampling_logp_difference/mean": 0.026387135365179608,
      "eval_steps_per_second": 0.11,
      "step": 70
    },
    {
      "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": 0.029296875,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2916.0,
      "completions/mean_length": 677.2819213867188,
      "completions/mean_terminated_length": 605.0067138671875,
      "completions/min_length": 113.0,
      "completions/min_terminated_length": 113.0,
      "entropy": 0.11027820646995679,
      "epoch": 1.0441176470588236,
      "frac_reward_zero_std": 0.515625,
      "grad_norm": 0.09778757444472032,
      "learning_rate": 1.6661872078293582e-06,
      "loss": 0.0019,
      "num_tokens": 83093292.0,
      "reward": 0.408203125,
      "reward_std": 0.4916611611843109,
      "rewards/reward_correctness/mean": 0.408203125,
      "rewards/reward_correctness/std": 0.4916611909866333,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999990463256836,
      "sampling/importance_sampling_ratio/min": 0.012234962545335293,
      "sampling/sampling_logp_difference/max": 4.4034576416015625,
      "sampling/sampling_logp_difference/mean": 0.010071806609630585,
      "step": 71,
      "step_time": 285.97908545611426
    },
    {
      "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": 0.0201822929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 678.1654052734375,
      "completions/mean_terminated_length": 628.8571166992188,
      "completions/min_length": 133.0,
      "completions/min_terminated_length": 133.0,
      "entropy": 0.12014149565948173,
      "epoch": 1.0588235294117647,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.0977574538696417,
      "learning_rate": 1.6338127921706424e-06,
      "loss": 0.0086,
      "num_tokens": 84306890.0,
      "reward": 0.3509114682674408,
      "reward_std": 0.4774106740951538,
      "rewards/reward_correctness/mean": 0.3509114682674408,
      "rewards/reward_correctness/std": 0.4774107038974762,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999822974205017,
      "sampling/importance_sampling_ratio/min": 1.1412609524086292e-07,
      "sampling/sampling_logp_difference/max": 15.9859619140625,
      "sampling/sampling_logp_difference/mean": 0.011068622581660748,
      "step": 72,
      "step_time": 297.21736324578524
    },
    {
      "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": 0.012369791977107525,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3060.0,
      "completions/mean_length": 608.6224365234375,
      "completions/mean_terminated_length": 577.82861328125,
      "completions/min_length": 71.0,
      "completions/min_terminated_length": 71.0,
      "entropy": 0.1028754398575984,
      "epoch": 1.0735294117647058,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.11441655651882668,
      "learning_rate": 1.6014476856245056e-06,
      "loss": 0.0136,
      "num_tokens": 85395286.0,
      "reward": 0.4733073115348816,
      "reward_std": 0.49944958090782166,
      "rewards/reward_correctness/mean": 0.4733072817325592,
      "rewards/reward_correctness/std": 0.49944961071014404,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000224113464355,
      "sampling/importance_sampling_ratio/min": 5.602986675512511e-06,
      "sampling/sampling_logp_difference/max": 12.09221076965332,
      "sampling/sampling_logp_difference/mean": 0.009659310802817345,
      "step": 73,
      "step_time": 304.63233517389745
    },
    {
      "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": 0.0201822929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3070.0,
      "completions/mean_length": 628.5150146484375,
      "completions/mean_terminated_length": 578.1840209960938,
      "completions/min_length": 96.0,
      "completions/min_terminated_length": 96.0,
      "entropy": 0.11295660870382562,
      "epoch": 1.088235294117647,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.10090089187503627,
      "learning_rate": 1.5691105010625233e-06,
      "loss": 0.0161,
      "num_tokens": 86510997.0,
      "reward": 0.458984375,
      "reward_std": 0.49847716093063354,
      "rewards/reward_correctness/mean": 0.458984375,
      "rewards/reward_correctness/std": 0.49847716093063354,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000795125961304,
      "sampling/importance_sampling_ratio/min": 0.008855166845023632,
      "sampling/sampling_logp_difference/max": 4.726754188537598,
      "sampling/sampling_logp_difference/mean": 0.010539291426539421,
      "step": 74,
      "step_time": 302.7705778395757
    },
    {
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2897.0,
      "completions/mean_length": 601.7467651367188,
      "completions/mean_terminated_length": 554.2103271484375,
      "completions/min_length": 85.0,
      "completions/min_terminated_length": 85.0,
      "entropy": 0.10929441236658022,
      "epoch": 1.1029411764705883,
      "frac_reward_zero_std": 0.59375,
      "grad_norm": 0.10200692192130892,
      "learning_rate": 1.536819835298597e-06,
      "loss": -0.0001,
      "num_tokens": 87608464.0,
      "reward": 0.484375,
      "reward_std": 0.49991855025291443,
      "rewards/reward_correctness/mean": 0.484375,
      "rewards/reward_correctness/std": 0.49991855025291443,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9957535862922668,
      "sampling/importance_sampling_ratio/min": 1.449367869865286e-16,
      "sampling/sampling_logp_difference/max": 36.47023391723633,
      "sampling/sampling_logp_difference/mean": 0.04288832098245621,
      "step": 75,
      "step_time": 311.1924736010842
    },
    {
      "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": 0.0234375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2985.0,
      "completions/mean_length": 660.0189208984375,
      "completions/mean_terminated_length": 602.13134765625,
      "completions/min_length": 100.0,
      "completions/min_terminated_length": 100.0,
      "entropy": 0.1125782448798418,
      "epoch": 1.1176470588235294,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.10450062970730634,
      "learning_rate": 1.5045942583941002e-06,
      "loss": 0.0069,
      "num_tokens": 88781529.0,
      "reward": 0.4166666865348816,
      "reward_std": 0.4931671619415283,
      "rewards/reward_correctness/mean": 0.4166666567325592,
      "rewards/reward_correctness/std": 0.4931672215461731,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999983310699463,
      "sampling/importance_sampling_ratio/min": 0.011418849229812622,
      "sampling/sampling_logp_difference/max": 4.472489833831787,
      "sampling/sampling_logp_difference/mean": 0.010256472043693066,
      "step": 76,
      "step_time": 301.73715012473986
    },
    {
      "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": 0.0182291679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3036.0,
      "completions/mean_length": 644.708984375,
      "completions/mean_terminated_length": 599.9515991210938,
      "completions/min_length": 104.0,
      "completions/min_terminated_length": 104.0,
      "entropy": 0.1032532652898226,
      "epoch": 1.1323529411764706,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.10228459845612378,
      "learning_rate": 1.4724523029784097e-06,
      "loss": 0.0058,
      "num_tokens": 89906982.0,
      "reward": 0.484375,
      "reward_std": 0.49991855025291443,
      "rewards/reward_correctness/mean": 0.484375,
      "rewards/reward_correctness/std": 0.49991855025291443,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9998574256896973,
      "sampling/importance_sampling_ratio/min": 0.0222180038690567,
      "sampling/sampling_logp_difference/max": 3.806852340698242,
      "sampling/sampling_logp_difference/mean": 0.009653305634856224,
      "step": 77,
      "step_time": 317.5641336161643
    },
    {
      "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": 0.0260416679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2996.0,
      "completions/mean_length": 738.216796875,
      "completions/mean_terminated_length": 675.816162109375,
      "completions/min_length": 99.0,
      "completions/min_terminated_length": 99.0,
      "entropy": 0.10691302624763921,
      "epoch": 1.1470588235294117,
      "frac_reward_zero_std": 0.4765625,
      "grad_norm": 0.09954934875420117,
      "learning_rate": 1.4404124535909613e-06,
      "loss": 0.0088,
      "num_tokens": 91189983.0,
      "reward": 0.43359375,
      "reward_std": 0.49573197960853577,
      "rewards/reward_correctness/mean": 0.43359375,
      "rewards/reward_correctness/std": 0.49573197960853577,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999540448188782,
      "sampling/importance_sampling_ratio/min": 0.006017227657139301,
      "sampling/sampling_logp_difference/max": 5.113128662109375,
      "sampling/sampling_logp_difference/mean": 0.009943715296685696,
      "step": 78,
      "step_time": 295.76658773003146
    },
    {
      "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": 3072.0,
      "completions/max_terminated_length": 2872.0,
      "completions/mean_length": 699.9069213867188,
      "completions/mean_terminated_length": 623.3877563476562,
      "completions/min_length": 78.0,
      "completions/min_terminated_length": 78.0,
      "entropy": 0.10646833421196789,
      "epoch": 1.161764705882353,
      "frac_reward_zero_std": 0.4453125,
      "grad_norm": 0.10932517333699619,
      "learning_rate": 1.4084931360509656e-06,
      "loss": 0.014,
      "num_tokens": 92420284.0,
      "reward": 0.5221354365348816,
      "reward_std": 0.49967241287231445,
      "rewards/reward_correctness/mean": 0.5221354365348816,
      "rewards/reward_correctness/std": 0.49967247247695923,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000180006027222,
      "sampling/importance_sampling_ratio/min": 0.025688737630844116,
      "sampling/sampling_logp_difference/max": 3.6617026329040527,
      "sampling/sampling_logp_difference/mean": 0.009844161570072174,
      "step": 79,
      "step_time": 317.65286510577425
    },
    {
      "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": 0.0279947929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2855.0,
      "completions/mean_length": 709.154296875,
      "completions/mean_terminated_length": 641.101806640625,
      "completions/min_length": 135.0,
      "completions/min_terminated_length": 135.0,
      "entropy": 0.10584194795228541,
      "epoch": 1.1764705882352942,
      "frac_reward_zero_std": 0.5390625,
      "grad_norm": 0.09227654592425696,
      "learning_rate": 1.376712706860888e-06,
      "loss": 0.0091,
      "num_tokens": 93671677.0,
      "reward": 0.3932291865348816,
      "reward_std": 0.48862606287002563,
      "rewards/reward_correctness/mean": 0.3932291567325592,
      "rewards/reward_correctness/std": 0.488626092672348,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999242424964905,
      "sampling/importance_sampling_ratio/min": 0.011996806599199772,
      "sampling/sampling_logp_difference/max": 4.423114776611328,
      "sampling/sampling_logp_difference/mean": 0.009747643955051899,
      "step": 80,
      "step_time": 302.8400612468831
    },
    {
      "epoch": 1.1764705882352942,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.027777778605620067,
      "eval_completions/max_length": 1370.3809523809523,
      "eval_completions/max_terminated_length": 1022.8809523809524,
      "eval_completions/mean_length": 592.8194589160737,
      "eval_completions/mean_terminated_length": 521.4313621520996,
      "eval_completions/min_length": 234.01190476190476,
      "eval_completions/min_terminated_length": 234.01190476190476,
      "eval_entropy": 0.0995181476076444,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 93671677.0,
      "eval_reward": 0.5119047732580275,
      "eval_reward_std": 0.5026668728817076,
      "eval_rewards/reward_correctness/mean": 0.5119047675813947,
      "eval_rewards/reward_correctness/std": 0.5026668714625495,
      "eval_runtime": 818.4358,
      "eval_samples_per_second": 0.611,
      "eval_sampling/importance_sampling_ratio/max": 2.9984878642218455,
      "eval_sampling/importance_sampling_ratio/mean": 0.9950423446439561,
      "eval_sampling/importance_sampling_ratio/min": 0.5204418023682332,
      "eval_sampling/sampling_logp_difference/max": 2.3195427627790544,
      "eval_sampling/sampling_logp_difference/mean": 0.026423310683596702,
      "eval_steps_per_second": 0.103,
      "step": 80
    },
    {
      "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": 0.025390625,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2966.0,
      "completions/mean_length": 725.4323120117188,
      "completions/mean_terminated_length": 664.29931640625,
      "completions/min_length": 115.0,
      "completions/min_terminated_length": 115.0,
      "entropy": 0.11524771642871201,
      "epoch": 1.1911764705882353,
      "frac_reward_zero_std": 0.578125,
      "grad_norm": 0.09536117201504443,
      "learning_rate": 1.3450894426497986e-06,
      "loss": 0.002,
      "num_tokens": 94931357.0,
      "reward": 0.3997395932674408,
      "reward_std": 0.49000421166419983,
      "rewards/reward_correctness/mean": 0.3997395932674408,
      "rewards/reward_correctness/std": 0.4900042414665222,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000006079673767,
      "sampling/importance_sampling_ratio/min": 0.018972357735037804,
      "sampling/sampling_logp_difference/max": 3.9647722244262695,
      "sampling/sampling_logp_difference/mean": 0.010577967390418053,
      "step": 81,
      "step_time": 311.70652225101367
    },
    {
      "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": 0.02734375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2805.0,
      "completions/mean_length": 665.19140625,
      "completions/mean_terminated_length": 597.5300903320312,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "entropy": 0.10638129588915035,
      "epoch": 1.2058823529411764,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.10873786908331745,
      "learning_rate": 1.313641529662647e-06,
      "loss": 0.014,
      "num_tokens": 96100355.0,
      "reward": 0.4505208432674408,
      "reward_std": 0.4977077841758728,
      "rewards/reward_correctness/mean": 0.4505208432674408,
      "rewards/reward_correctness/std": 0.4977078139781952,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999487400054932,
      "sampling/importance_sampling_ratio/min": 0.003440916072577238,
      "sampling/sampling_logp_difference/max": 5.672017574310303,
      "sampling/sampling_logp_difference/mean": 0.009805039502680302,
      "step": 82,
      "step_time": 290.32841725600883
    },
    {
      "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": 0.0162760429084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2964.0,
      "completions/mean_length": 650.9525146484375,
      "completions/mean_terminated_length": 611.2316284179688,
      "completions/min_length": 130.0,
      "completions/min_terminated_length": 130.0,
      "entropy": 0.10257759928936139,
      "epoch": 1.2205882352941178,
      "frac_reward_zero_std": 0.5078125,
      "grad_norm": 0.10011774111650357,
      "learning_rate": 1.2823870533015295e-06,
      "loss": 0.0071,
      "num_tokens": 97261378.0,
      "reward": 0.4739583432674408,
      "reward_std": 0.49948397278785706,
      "rewards/reward_correctness/mean": 0.4739583432674408,
      "rewards/reward_correctness/std": 0.49948397278785706,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999834895133972,
      "sampling/importance_sampling_ratio/min": 0.051053013652563095,
      "sampling/sampling_logp_difference/max": 3.1086947917938232,
      "sampling/sampling_logp_difference/mean": 0.009559324942529202,
      "step": 83,
      "step_time": 314.8325279019773
    },
    {
      "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": 0.01888020895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3036.0,
      "completions/mean_length": 677.6198120117188,
      "completions/mean_terminated_length": 631.54345703125,
      "completions/min_length": 134.0,
      "completions/min_terminated_length": 134.0,
      "entropy": 0.11533306317869574,
      "epoch": 1.2352941176470589,
      "frac_reward_zero_std": 0.5546875,
      "grad_norm": 0.09282478295845689,
      "learning_rate": 1.2513439877249363e-06,
      "loss": 0.004,
      "num_tokens": 98480102.0,
      "reward": 0.4225260615348816,
      "reward_std": 0.49412214756011963,
      "rewards/reward_correctness/mean": 0.4225260317325592,
      "rewards/reward_correctness/std": 0.4941222369670868,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000012993812561,
      "sampling/importance_sampling_ratio/min": 0.051133543252944946,
      "sampling/sampling_logp_difference/max": 2.9733145236968994,
      "sampling/sampling_logp_difference/mean": 0.01072630099952221,
      "step": 84,
      "step_time": 317.55150324339047
    },
    {
      "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": 0.03059895895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2932.0,
      "completions/mean_length": 699.4720458984375,
      "completions/mean_terminated_length": 624.5836181640625,
      "completions/min_length": 120.0,
      "completions/min_terminated_length": 120.0,
      "entropy": 0.11002266616560519,
      "epoch": 1.25,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.09594840299692999,
      "learning_rate": 1.220530185510985e-06,
      "loss": 0.0126,
      "num_tokens": 99716383.0,
      "reward": 0.4700520932674408,
      "reward_std": 0.49926483631134033,
      "rewards/reward_correctness/mean": 0.4700520932674408,
      "rewards/reward_correctness/std": 0.4992648661136627,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999629855155945,
      "sampling/importance_sampling_ratio/min": 0.042671963572502136,
      "sampling/sampling_logp_difference/max": 3.1542131900787354,
      "sampling/sampling_logp_difference/mean": 0.010363861918449402,
      "step": 85,
      "step_time": 306.6277994834818
    },
    {
      "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": 0.02473958395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3045.0,
      "completions/mean_length": 699.0983276367188,
      "completions/mean_terminated_length": 638.904541015625,
      "completions/min_length": 91.0,
      "completions/min_terminated_length": 91.0,
      "entropy": 0.11227407562546432,
      "epoch": 1.2647058823529411,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.0976507340489589,
      "learning_rate": 1.189963367390565e-06,
      "loss": 0.0028,
      "num_tokens": 100929470.0,
      "reward": 0.4270833432674408,
      "reward_std": 0.4948156476020813,
      "rewards/reward_correctness/mean": 0.4270833432674408,
      "rewards/reward_correctness/std": 0.4948156774044037,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000653266906738,
      "sampling/importance_sampling_ratio/min": 0.00828573852777481,
      "sampling/sampling_logp_difference/max": 4.793219566345215,
      "sampling/sampling_logp_difference/mean": 0.010349968448281288,
      "step": 86,
      "step_time": 309.40160133549944
    },
    {
      "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": 0.033203125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3067.0,
      "completions/mean_length": 759.8939208984375,
      "completions/mean_terminated_length": 680.816162109375,
      "completions/min_length": 143.0,
      "completions/min_terminated_length": 143.0,
      "entropy": 0.10843936848687008,
      "epoch": 1.2794117647058822,
      "frac_reward_zero_std": 0.5390625,
      "grad_norm": 0.089055654394311,
      "learning_rate": 1.159661112056314e-06,
      "loss": 0.0081,
      "num_tokens": 102291799.0,
      "reward": 0.3932291865348816,
      "reward_std": 0.48862606287002563,
      "rewards/reward_correctness/mean": 0.3932291567325592,
      "rewards/reward_correctness/std": 0.488626092672348,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999238848686218,
      "sampling/importance_sampling_ratio/min": 0.0069398460909724236,
      "sampling/sampling_logp_difference/max": 4.970475673675537,
      "sampling/sampling_logp_difference/mean": 0.009973037987947464,
      "step": 87,
      "step_time": 333.1185615430586
    },
    {
      "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": 0.0234375,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3020.0,
      "completions/mean_length": 676.6627807617188,
      "completions/mean_terminated_length": 619.1746826171875,
      "completions/min_length": 161.0,
      "completions/min_terminated_length": 161.0,
      "entropy": 0.10631966090295464,
      "epoch": 1.2941176470588236,
      "frac_reward_zero_std": 0.4609375,
      "grad_norm": 0.10902086573519744,
      "learning_rate": 1.1296408460532715e-06,
      "loss": 0.0173,
      "num_tokens": 103498073.0,
      "reward": 0.46875,
      "reward_std": 0.4991849958896637,
      "rewards/reward_correctness/mean": 0.46875,
      "rewards/reward_correctness/std": 0.4991849958896637,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999716877937317,
      "sampling/importance_sampling_ratio/min": 0.04054650664329529,
      "sampling/sampling_logp_difference/max": 3.205305576324463,
      "sampling/sampling_logp_difference/mean": 0.00969459768384695,
      "step": 88,
      "step_time": 307.3499052450061
    },
    {
      "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": 0.0279947929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2906.0,
      "completions/mean_length": 706.0625,
      "completions/mean_terminated_length": 637.9209594726562,
      "completions/min_length": 81.0,
      "completions/min_terminated_length": 81.0,
      "entropy": 0.10058472969103605,
      "epoch": 1.3088235294117647,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.09902012748669947,
      "learning_rate": 1.0999198337570392e-06,
      "loss": 0.0137,
      "num_tokens": 104734205.0,
      "reward": 0.4596354365348816,
      "reward_std": 0.4985303282737732,
      "rewards/reward_correctness/mean": 0.4596354067325592,
      "rewards/reward_correctness/std": 0.4985303580760956,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999986290931702,
      "sampling/importance_sampling_ratio/min": 7.468889816664159e-05,
      "sampling/sampling_logp_difference/max": 9.502179145812988,
      "sampling/sampling_logp_difference/mean": 0.00929464865475893,
      "step": 89,
      "step_time": 299.34720146795735
    },
    {
      "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": 0.01953125,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2999.0,
      "completions/mean_length": 693.0866088867188,
      "completions/mean_terminated_length": 645.9681396484375,
      "completions/min_length": 155.0,
      "completions/min_terminated_length": 155.0,
      "entropy": 0.10605936858337373,
      "epoch": 1.3235294117647058,
      "frac_reward_zero_std": 0.53125,
      "grad_norm": 0.09893613436677237,
      "learning_rate": 1.0705151674451938e-06,
      "loss": 0.0013,
      "num_tokens": 105954954.0,
      "reward": 0.412109375,
      "reward_std": 0.49237489700317383,
      "rewards/reward_correctness/mean": 0.412109375,
      "rewards/reward_correctness/std": 0.4923749268054962,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000171661376953,
      "sampling/importance_sampling_ratio/min": 0.0009838553378358483,
      "sampling/sampling_logp_difference/max": 6.924031734466553,
      "sampling/sampling_logp_difference/mean": 0.010026591829955578,
      "step": 90,
      "step_time": 320.5724582099356
    },
    {
      "epoch": 1.3235294117647058,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.01984127043258576,
      "eval_completions/max_length": 1227.952380952381,
      "eval_completions/max_terminated_length": 1012.6428571428571,
      "eval_completions/mean_length": 574.4107317243304,
      "eval_completions/mean_terminated_length": 524.4579500470843,
      "eval_completions/min_length": 231.54761904761904,
      "eval_completions/min_terminated_length": 231.54761904761904,
      "eval_entropy": 0.09793427143068541,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 105954954.0,
      "eval_reward": 0.5218254084743205,
      "eval_reward_std": 0.49357138325770694,
      "eval_rewards/reward_correctness/mean": 0.5218254020881086,
      "eval_rewards/reward_correctness/std": 0.4935713772262846,
      "eval_runtime": 727.7944,
      "eval_samples_per_second": 0.687,
      "eval_sampling/importance_sampling_ratio/max": 2.9972963560195196,
      "eval_sampling/importance_sampling_ratio/mean": 0.9947727024555206,
      "eval_sampling/importance_sampling_ratio/min": 0.5220564539943423,
      "eval_sampling/sampling_logp_difference/max": 2.271739939848582,
      "eval_sampling/sampling_logp_difference/mean": 0.025922782874355715,
      "eval_steps_per_second": 0.115,
      "step": 90
    },
    {
      "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": 0.02669270895421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2971.0,
      "completions/mean_length": 670.478515625,
      "completions/mean_terminated_length": 604.6173706054688,
      "completions/min_length": 168.0,
      "completions/min_terminated_length": 168.0,
      "entropy": 0.11070733214728534,
      "epoch": 1.3382352941176472,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.10384112383517705,
      "learning_rate": 1.0414437574676832e-06,
      "loss": 0.0096,
      "num_tokens": 107175105.0,
      "reward": 0.4850260615348816,
      "reward_std": 0.4999384880065918,
      "rewards/reward_correctness/mean": 0.4850260317325592,
      "rewards/reward_correctness/std": 0.4999385178089142,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999462962150574,
      "sampling/importance_sampling_ratio/min": 0.015210184268653393,
      "sampling/sampling_logp_difference/max": 4.185790061950684,
      "sampling/sampling_logp_difference/mean": 0.010488501749932766,
      "step": 91,
      "step_time": 318.08187717711553
    },
    {
      "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": 0.0572916679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2974.0,
      "completions/mean_length": 771.5859375,
      "completions/mean_terminated_length": 632.3377075195312,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "entropy": 0.1037912426982075,
      "epoch": 1.3529411764705883,
      "frac_reward_zero_std": 0.5234375,
      "grad_norm": 0.09178918055813894,
      "learning_rate": 1.0127223225218379e-06,
      "loss": 0.0206,
      "num_tokens": 108526425.0,
      "reward": 0.4505208432674408,
      "reward_std": 0.4977077841758728,
      "rewards/reward_correctness/mean": 0.4505208432674408,
      "rewards/reward_correctness/std": 0.4977078139781952,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000383853912354,
      "sampling/importance_sampling_ratio/min": 0.0019341822480782866,
      "sampling/sampling_logp_difference/max": 6.24807071685791,
      "sampling/sampling_logp_difference/mean": 0.009867710061371326,
      "step": 92,
      "step_time": 316.0852893306874
    },
    {
      "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": 0.02473958395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3047.0,
      "completions/mean_length": 672.578125,
      "completions/mean_terminated_length": 611.8264770507812,
      "completions/min_length": 136.0,
      "completions/min_terminated_length": 136.0,
      "entropy": 0.10452087642624974,
      "epoch": 1.3676470588235294,
      "frac_reward_zero_std": 0.625,
      "grad_norm": 0.0979449353074455,
      "learning_rate": 9.843673800376037e-07,
      "loss": 0.0097,
      "num_tokens": 109731417.0,
      "reward": 0.5006510615348816,
      "reward_std": 0.5001624226570129,
      "rewards/reward_correctness/mean": 0.5006510615348816,
      "rewards/reward_correctness/std": 0.5001624226570129,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9955242276191711,
      "sampling/importance_sampling_ratio/min": 2.6044188559093452e-17,
      "sampling/sampling_logp_difference/max": 38.186737060546875,
      "sampling/sampling_logp_difference/mean": 0.04697582125663757,
      "step": 93,
      "step_time": 322.82180776353925
    },
    {
      "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": 0.0182291679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2877.0,
      "completions/mean_length": 635.4290771484375,
      "completions/mean_terminated_length": 590.1876831054688,
      "completions/min_length": 97.0,
      "completions/min_terminated_length": 97.0,
      "entropy": 0.10279296265798621,
      "epoch": 1.3823529411764706,
      "frac_reward_zero_std": 0.546875,
      "grad_norm": 0.10426534258129598,
      "learning_rate": 9.563952366785246e-07,
      "loss": 0.005,
      "num_tokens": 110869904.0,
      "reward": 0.453125,
      "reward_std": 0.4979599714279175,
      "rewards/reward_correctness/mean": 0.453125,
      "rewards/reward_correctness/std": 0.49796000123023987,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.000075340270996,
      "sampling/importance_sampling_ratio/min": 0.010155136696994305,
      "sampling/sampling_logp_difference/max": 4.589775562286377,
      "sampling/sampling_logp_difference/mean": 0.009832991287112236,
      "step": 94,
      "step_time": 315.47392428759485
    },
    {
      "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": 0.0338541679084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 740.6784057617188,
      "completions/mean_terminated_length": 659.07275390625,
      "completions/min_length": 123.0,
      "completions/min_terminated_length": 123.0,
      "entropy": 0.10438565909862518,
      "epoch": 1.3970588235294117,
      "frac_reward_zero_std": 0.46875,
      "grad_norm": 0.10174293989185926,
      "learning_rate": 9.288219789639276e-07,
      "loss": 0.0078,
      "num_tokens": 112162002.0,
      "reward": 0.3971354365348816,
      "reward_std": 0.48946380615234375,
      "rewards/reward_correctness/mean": 0.3971354067325592,
      "rewards/reward_correctness/std": 0.4894638657569885,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000094175338745,
      "sampling/importance_sampling_ratio/min": 0.05274541676044464,
      "sampling/sampling_logp_difference/max": 2.9422783851623535,
      "sampling/sampling_logp_difference/mean": 0.009899480268359184,
      "step": 95,
      "step_time": 319.02526410855353
    },
    {
      "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": 0.0279947929084301,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3001.0,
      "completions/mean_length": 696.400390625,
      "completions/mean_terminated_length": 627.9805908203125,
      "completions/min_length": 112.0,
      "completions/min_terminated_length": 112.0,
      "entropy": 0.1095525169512257,
      "epoch": 1.4117647058823528,
      "frac_reward_zero_std": 0.4921875,
      "grad_norm": 0.09956779342472952,
      "learning_rate": 9.016634640177203e-07,
      "loss": 0.0032,
      "num_tokens": 113383389.0,
      "reward": 0.4700520932674408,
      "reward_std": 0.49926483631134033,
      "rewards/reward_correctness/mean": 0.4700520932674408,
      "rewards/reward_correctness/std": 0.4992648661136627,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.999976634979248,
      "sampling/importance_sampling_ratio/min": 0.018746379762887955,
      "sampling/sampling_logp_difference/max": 3.976754665374756,
      "sampling/sampling_logp_difference/mean": 0.010300280526280403,
      "step": 96,
      "step_time": 296.686714746058
    },
    {
      "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": 0.0442708358168602,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2918.0,
      "completions/mean_length": 707.390625,
      "completions/mean_terminated_length": 597.8582763671875,
      "completions/min_length": 109.0,
      "completions/min_terminated_length": 109.0,
      "entropy": 0.10301207873271778,
      "epoch": 1.4264705882352942,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.10752603822364419,
      "learning_rate": 8.74935310449101e-07,
      "loss": 0.0101,
      "num_tokens": 114641661.0,
      "reward": 0.4459635615348816,
      "reward_std": 0.4972333610057831,
      "rewards/reward_correctness/mean": 0.4459635317325592,
      "rewards/reward_correctness/std": 0.49723339080810547,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9956812858581543,
      "sampling/importance_sampling_ratio/min": 1.156828547549599e-17,
      "sampling/sampling_logp_difference/max": 38.99826431274414,
      "sampling/sampling_logp_difference/mean": 0.05035912245512009,
      "step": 97,
      "step_time": 362.47534032771364
    },
    {
      "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": 0.02473958395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 2888.0,
      "completions/mean_length": 693.869140625,
      "completions/mean_terminated_length": 633.542724609375,
      "completions/min_length": 138.0,
      "completions/min_terminated_length": 138.0,
      "entropy": 0.10327525227330625,
      "epoch": 1.4411764705882353,
      "frac_reward_zero_std": 0.453125,
      "grad_norm": 0.11138329626839688,
      "learning_rate": 8.486528893704481e-07,
      "loss": 0.0123,
      "num_tokens": 115854972.0,
      "reward": 0.4765625,
      "reward_std": 0.4996130168437958,
      "rewards/reward_correctness/mean": 0.4765625,
      "rewards/reward_correctness/std": 0.49961304664611816,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000249147415161,
      "sampling/importance_sampling_ratio/min": 0.003647557459771633,
      "sampling/sampling_logp_difference/max": 5.613697528839111,
      "sampling/sampling_logp_difference/mean": 0.009877603501081467,
      "step": 98,
      "step_time": 316.61871033906937
    },
    {
      "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": 0.02473958395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3030.0,
      "completions/mean_length": 681.708984375,
      "completions/mean_terminated_length": 621.0740966796875,
      "completions/min_length": 94.0,
      "completions/min_terminated_length": 94.0,
      "entropy": 0.09731233183993027,
      "epoch": 1.4558823529411764,
      "frac_reward_zero_std": 0.5390625,
      "grad_norm": 0.09440221134197979,
      "learning_rate": 8.228313155575304e-07,
      "loss": 0.0093,
      "num_tokens": 117048021.0,
      "reward": 0.4459635615348816,
      "reward_std": 0.4972333312034607,
      "rewards/reward_correctness/mean": 0.4459635317325592,
      "rewards/reward_correctness/std": 0.49723339080810547,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 1.0000076293945312,
      "sampling/importance_sampling_ratio/min": 0.025448348373174667,
      "sampling/sampling_logp_difference/max": 3.6711044311523438,
      "sampling/sampling_logp_difference/mean": 0.0093543641269207,
      "step": 99,
      "step_time": 289.0866700396873
    },
    {
      "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": 0.02864583395421505,
      "completions/max_length": 3072.0,
      "completions/max_terminated_length": 3048.0,
      "completions/mean_length": 721.3880615234375,
      "completions/mean_terminated_length": 652.0676879882812,
      "completions/min_length": 84.0,
      "completions/min_terminated_length": 84.0,
      "entropy": 0.10687656450318173,
      "epoch": 1.4705882352941178,
      "frac_reward_zero_std": 0.5,
      "grad_norm": 0.10480703467405905,
      "learning_rate": 7.97485438757144e-07,
      "loss": 0.0068,
      "num_tokens": 118314965.0,
      "reward": 0.404296875,
      "reward_std": 0.4909152686595917,
      "rewards/reward_correctness/mean": 0.404296875,
      "rewards/reward_correctness/std": 0.49091529846191406,
      "sampling/importance_sampling_ratio/max": 3.0,
      "sampling/importance_sampling_ratio/mean": 0.9999392628669739,
      "sampling/importance_sampling_ratio/min": 0.018848691135644913,
      "sampling/sampling_logp_difference/max": 3.9713118076324463,
      "sampling/sampling_logp_difference/mean": 0.01010206714272499,
      "step": 100,
      "step_time": 316.2152795935981
    },
    {
      "epoch": 1.4705882352941178,
      "eval_clip_ratio/high_max": 0.0,
      "eval_clip_ratio/high_mean": 0.0,
      "eval_clip_ratio/low_mean": 0.0,
      "eval_clip_ratio/low_min": 0.0,
      "eval_clip_ratio/region_mean": 0.0,
      "eval_completions/clipped_ratio": 0.029761905648878643,
      "eval_completions/max_length": 1368.107142857143,
      "eval_completions/max_terminated_length": 995.8095238095239,
      "eval_completions/mean_length": 603.3313642229352,
      "eval_completions/mean_terminated_length": 527.4833472115653,
      "eval_completions/min_length": 238.27380952380952,
      "eval_completions/min_terminated_length": 238.27380952380952,
      "eval_entropy": 0.09634455040629421,
      "eval_frac_reward_zero_std": 1.0,
      "eval_loss": 0.0,
      "eval_num_tokens": 118314965.0,
      "eval_reward": 0.52976191646996,
      "eval_reward_std": 0.4925737629334132,
      "eval_rewards/reward_correctness/mean": 0.5297619107933271,
      "eval_rewards/reward_correctness/std": 0.4925737593855177,
      "eval_runtime": 817.1205,
      "eval_samples_per_second": 0.612,
      "eval_sampling/importance_sampling_ratio/max": 2.998494170960926,
      "eval_sampling/importance_sampling_ratio/mean": 0.9951962956360408,
      "eval_sampling/importance_sampling_ratio/min": 0.5286199062885273,
      "eval_sampling/sampling_logp_difference/max": 2.220592745712825,
      "eval_sampling/sampling_logp_difference/mean": 0.02555376207012506,
      "eval_steps_per_second": 0.103,
      "step": 100
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 136,
  "num_input_tokens_seen": 118314965,
  "num_train_epochs": 2,
  "save_steps": 10,
  "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": 2,
  "trial_name": null,
  "trial_params": null
}