Files
supergames-grpo/checkpoint-20/trainer_state.json
ModelHub XC a53def3b08 初始化项目,由ModelHub XC社区提供模型
Model: pvs333/supergames-grpo
Source: Original Platform
2026-06-16 08:06:17 +08:00

615 lines
22 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.1,
"eval_steps": 500,
"global_step": 20,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 478.0,
"completions/max_terminated_length": 478.0,
"completions/mean_length": 191.0,
"completions/mean_terminated_length": 191.0,
"completions/min_length": 66.0,
"completions/min_terminated_length": 66.0,
"entropy": 0.5566893219947815,
"epoch": 0.005,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.370361328125,
"learning_rate": 1e-06,
"loss": -0.24658073484897614,
"num_tokens": 10832.0,
"reward": -3.3000030517578125,
"reward_std": 85.62333679199219,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -3.4000015258789062,
"rewards/supergames_reward/std": 85.62333679199219,
"step": 1,
"step_time": 22.779711072013015
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 216.0,
"completions/max_terminated_length": 216.0,
"completions/mean_length": 131.5,
"completions/mean_terminated_length": 131.5,
"completions/min_length": 60.0,
"completions/min_terminated_length": 60.0,
"entropy": 0.6395841240882874,
"epoch": 0.01,
"frac_reward_zero_std": 0.0,
"grad_norm": 8.09774398803711,
"learning_rate": 9.95e-07,
"loss": 0.20567649602890015,
"num_tokens": 16404.0,
"reward": -12.422499656677246,
"reward_std": 7.134707450866699,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -12.522500038146973,
"rewards/supergames_reward/std": 7.134707927703857,
"step": 2,
"step_time": 9.960156448010821
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 287.0,
"completions/max_terminated_length": 287.0,
"completions/mean_length": 156.625,
"completions/mean_terminated_length": 156.625,
"completions/min_length": 77.0,
"completions/min_terminated_length": 77.0,
"entropy": 0.562222421169281,
"epoch": 0.015,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.9053616523742676,
"learning_rate": 9.9e-07,
"loss": 0.09602774679660797,
"num_tokens": 26953.0,
"reward": -74.94999694824219,
"reward_std": 70.73094177246094,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -75.0,
"rewards/supergames_reward/std": 70.71067810058594,
"step": 3,
"step_time": 13.686320498993155
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 268.0,
"completions/max_terminated_length": 268.0,
"completions/mean_length": 154.125,
"completions/mean_terminated_length": 154.125,
"completions/min_length": 53.0,
"completions/min_terminated_length": 53.0,
"entropy": 0.7323317527770996,
"epoch": 0.02,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.933310031890869,
"learning_rate": 9.849999999999999e-07,
"loss": 0.28249427676200867,
"num_tokens": 36514.0,
"reward": -40.7599983215332,
"reward_std": 81.73140716552734,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -40.80999755859375,
"rewards/supergames_reward/std": 81.68995666503906,
"step": 4,
"step_time": 12.805880262021674
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 401.0,
"completions/max_terminated_length": 401.0,
"completions/mean_length": 205.25,
"completions/mean_terminated_length": 205.25,
"completions/min_length": 115.0,
"completions/min_terminated_length": 115.0,
"entropy": 0.6167430281639099,
"epoch": 0.025,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.344135046005249,
"learning_rate": 9.8e-07,
"loss": 0.1433994024991989,
"num_tokens": 47476.0,
"reward": -97.6612548828125,
"reward_std": 6.3321428298950195,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -97.76124572753906,
"rewards/supergames_reward/std": 6.3321428298950195,
"step": 5,
"step_time": 18.789364666008623
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 276.0,
"completions/max_terminated_length": 276.0,
"completions/mean_length": 167.25,
"completions/mean_terminated_length": 167.25,
"completions/min_length": 73.0,
"completions/min_terminated_length": 73.0,
"entropy": 0.7583790421485901,
"epoch": 0.03,
"frac_reward_zero_std": 0.0,
"grad_norm": 5.172371864318848,
"learning_rate": 9.75e-07,
"loss": -0.12127404659986496,
"num_tokens": 55246.0,
"reward": -58.687503814697266,
"reward_std": 58.64059829711914,
"rewards/json_format_reward/mean": 0.07500000298023224,
"rewards/json_format_reward/std": 0.0707106813788414,
"rewards/supergames_reward/mean": -58.76250457763672,
"rewards/supergames_reward/std": 58.620460510253906,
"step": 6,
"step_time": 12.948570273991209
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 261.0,
"completions/max_terminated_length": 261.0,
"completions/mean_length": 133.875,
"completions/mean_terminated_length": 133.875,
"completions/min_length": 57.0,
"completions/min_terminated_length": 57.0,
"entropy": 0.7447654604911804,
"epoch": 0.035,
"frac_reward_zero_std": 0.0,
"grad_norm": 6.428595066070557,
"learning_rate": 9.7e-07,
"loss": 0.10135584324598312,
"num_tokens": 60837.0,
"reward": -32.45000076293945,
"reward_std": 41.754486083984375,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -32.5,
"rewards/supergames_reward/std": 41.66190719604492,
"step": 7,
"step_time": 12.058315072004916
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 414.0,
"completions/max_terminated_length": 414.0,
"completions/mean_length": 209.125,
"completions/mean_terminated_length": 209.125,
"completions/min_length": 104.0,
"completions/min_terminated_length": 104.0,
"entropy": 0.3860666751861572,
"epoch": 0.04,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.806255578994751,
"learning_rate": 9.649999999999999e-07,
"loss": -0.33238139748573303,
"num_tokens": 70838.0,
"reward": -80.50375366210938,
"reward_std": 38.82380294799805,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -80.60375213623047,
"rewards/supergames_reward/std": 38.82379913330078,
"step": 8,
"step_time": 19.323370319994865
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.25,
"completions/max_length": 512.0,
"completions/max_terminated_length": 199.0,
"completions/mean_length": 247.125,
"completions/mean_terminated_length": 158.83334350585938,
"completions/min_length": 106.0,
"completions/min_terminated_length": 106.0,
"entropy": 0.5816237330436707,
"epoch": 0.045,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.3209943771362305,
"learning_rate": 9.6e-07,
"loss": 0.36752766370773315,
"num_tokens": 82135.0,
"reward": -27.688751220703125,
"reward_std": 82.8358154296875,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -27.738750457763672,
"rewards/supergames_reward/std": 82.78590393066406,
"step": 9,
"step_time": 21.948575104994234
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 240.0,
"completions/max_terminated_length": 240.0,
"completions/mean_length": 153.0,
"completions/mean_terminated_length": 153.0,
"completions/min_length": 51.0,
"completions/min_terminated_length": 51.0,
"entropy": 0.6301521062850952,
"epoch": 0.05,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.125359058380127,
"learning_rate": 9.55e-07,
"loss": 0.06129350885748863,
"num_tokens": 89807.0,
"reward": -78.39624786376953,
"reward_std": 40.17042922973633,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -78.44625091552734,
"rewards/supergames_reward/std": 40.1396484375,
"step": 10,
"step_time": 9.098202993016457
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 133.0,
"completions/max_terminated_length": 133.0,
"completions/mean_length": 112.0,
"completions/mean_terminated_length": 112.0,
"completions/min_length": 68.0,
"completions/min_terminated_length": 68.0,
"entropy": 0.5521990656852722,
"epoch": 0.055,
"frac_reward_zero_std": 0.0,
"grad_norm": 6.616060256958008,
"learning_rate": 9.499999999999999e-07,
"loss": -0.006313305348157883,
"num_tokens": 95231.0,
"reward": -12.161249160766602,
"reward_std": 6.39578104019165,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -12.261249542236328,
"rewards/supergames_reward/std": 6.39578104019165,
"step": 11,
"step_time": 5.136311663984088
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 396.0,
"completions/max_terminated_length": 396.0,
"completions/mean_length": 214.875,
"completions/mean_terminated_length": 214.875,
"completions/min_length": 55.0,
"completions/min_terminated_length": 55.0,
"entropy": 0.5505905151367188,
"epoch": 0.06,
"frac_reward_zero_std": 0.0,
"grad_norm": 5.289583206176758,
"learning_rate": 9.45e-07,
"loss": -0.29802003502845764,
"num_tokens": 106238.0,
"reward": -91.66999816894531,
"reward_std": 23.2779541015625,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -91.77000427246094,
"rewards/supergames_reward/std": 23.277956008911133,
"step": 12,
"step_time": 14.945365622988902
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 269.0,
"completions/max_terminated_length": 269.0,
"completions/mean_length": 167.375,
"completions/mean_terminated_length": 167.375,
"completions/min_length": 113.0,
"completions/min_terminated_length": 113.0,
"entropy": 0.540415346622467,
"epoch": 0.065,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.858291149139404,
"learning_rate": 9.399999999999999e-07,
"loss": 0.24004128575325012,
"num_tokens": 116937.0,
"reward": -20.38249969482422,
"reward_std": 87.55204010009766,
"rewards/json_format_reward/mean": 0.07500000298023224,
"rewards/json_format_reward/std": 0.0707106813788414,
"rewards/supergames_reward/mean": -20.457500457763672,
"rewards/supergames_reward/std": 87.52605438232422,
"step": 13,
"step_time": 10.482285185018554
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 254.0,
"completions/max_terminated_length": 254.0,
"completions/mean_length": 156.875,
"completions/mean_terminated_length": 156.875,
"completions/min_length": 114.0,
"completions/min_terminated_length": 114.0,
"entropy": 0.7237679958343506,
"epoch": 0.07,
"frac_reward_zero_std": 0.0,
"grad_norm": 5.083980560302734,
"learning_rate": 9.35e-07,
"loss": -0.19025824964046478,
"num_tokens": 124640.0,
"reward": -51.397499084472656,
"reward_std": 54.486454010009766,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -51.49749755859375,
"rewards/supergames_reward/std": 54.486454010009766,
"step": 14,
"step_time": 9.597231683001155
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 236.0,
"completions/max_terminated_length": 236.0,
"completions/mean_length": 141.375,
"completions/mean_terminated_length": 141.375,
"completions/min_length": 91.0,
"completions/min_terminated_length": 91.0,
"entropy": 0.754334568977356,
"epoch": 0.075,
"frac_reward_zero_std": 0.0,
"grad_norm": 5.720428943634033,
"learning_rate": 9.3e-07,
"loss": -0.16830675303936005,
"num_tokens": 132219.0,
"reward": -56.226253509521484,
"reward_std": 60.115657806396484,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -56.32625198364258,
"rewards/supergames_reward/std": 60.11566162109375,
"step": 15,
"step_time": 8.998362872982398
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 209.0,
"completions/max_terminated_length": 209.0,
"completions/mean_length": 136.375,
"completions/mean_terminated_length": 136.375,
"completions/min_length": 97.0,
"completions/min_terminated_length": 97.0,
"entropy": 0.6280568838119507,
"epoch": 0.08,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.9681806564331055,
"learning_rate": 9.25e-07,
"loss": -0.01910916529595852,
"num_tokens": 139726.0,
"reward": -68.24000549316406,
"reward_std": 52.238914489746094,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -68.33999633789062,
"rewards/supergames_reward/std": 52.238914489746094,
"step": 16,
"step_time": 8.034480294008972
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 365.0,
"completions/max_terminated_length": 365.0,
"completions/mean_length": 211.875,
"completions/mean_terminated_length": 211.875,
"completions/min_length": 100.0,
"completions/min_terminated_length": 100.0,
"entropy": 0.899176836013794,
"epoch": 0.085,
"frac_reward_zero_std": 0.0,
"grad_norm": 6.691746711730957,
"learning_rate": 9.2e-07,
"loss": 0.25513291358947754,
"num_tokens": 147869.0,
"reward": -99.92500305175781,
"reward_std": 0.07070883363485336,
"rewards/json_format_reward/mean": 0.07500000298023224,
"rewards/json_format_reward/std": 0.0707106813788414,
"rewards/supergames_reward/mean": -100.0,
"rewards/supergames_reward/std": 0.0,
"step": 17,
"step_time": 13.655792869016295
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 387.0,
"completions/max_terminated_length": 387.0,
"completions/mean_length": 238.0,
"completions/mean_terminated_length": 238.0,
"completions/min_length": 87.0,
"completions/min_terminated_length": 87.0,
"entropy": 0.5596873760223389,
"epoch": 0.09,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.6711347103118896,
"learning_rate": 9.15e-07,
"loss": -0.2628335654735565,
"num_tokens": 159101.0,
"reward": -24.900001525878906,
"reward_std": 103.50983428955078,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -25.0,
"rewards/supergames_reward/std": 103.50983428955078,
"step": 18,
"step_time": 14.784329420013819
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.125,
"completions/max_length": 512.0,
"completions/max_terminated_length": 277.0,
"completions/mean_length": 219.75,
"completions/mean_terminated_length": 178.00001525878906,
"completions/min_length": 90.0,
"completions/min_terminated_length": 90.0,
"entropy": 0.9245517253875732,
"epoch": 0.095,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.55502986907959,
"learning_rate": 9.1e-07,
"loss": 0.13060179352760315,
"num_tokens": 167299.0,
"reward": -86.89125061035156,
"reward_std": 36.95603561401367,
"rewards/json_format_reward/mean": 0.05000000074505806,
"rewards/json_format_reward/std": 0.09258200973272324,
"rewards/supergames_reward/mean": -86.94125366210938,
"rewards/supergames_reward/std": 36.93572235107422,
"step": 19,
"step_time": 18.832684700988466
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 267.0,
"completions/max_terminated_length": 267.0,
"completions/mean_length": 177.25,
"completions/mean_terminated_length": 177.25,
"completions/min_length": 93.0,
"completions/min_terminated_length": 93.0,
"entropy": 0.7331254482269287,
"epoch": 0.1,
"frac_reward_zero_std": 0.0,
"grad_norm": 4.538497447967529,
"learning_rate": 9.05e-07,
"loss": -0.16418980062007904,
"num_tokens": 175141.0,
"reward": -68.52874755859375,
"reward_std": 58.48719787597656,
"rewards/json_format_reward/mean": 0.10000000149011612,
"rewards/json_format_reward/std": 0.0,
"rewards/supergames_reward/mean": -68.62875366210938,
"rewards/supergames_reward/std": 58.48720169067383,
"step": 20,
"step_time": 10.090975169994636
}
],
"logging_steps": 1,
"max_steps": 200,
"num_input_tokens_seen": 175141,
"num_train_epochs": 1,
"save_steps": 20,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 8,
"trial_name": null,
"trial_params": null
}