grpo-checkpoints-better / trainer_state.json
SamMikaelson's picture
Upload folder using huggingface_hub
ef895e2 verified
raw
history blame
39 kB
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.02406821620134782,
"eval_steps": 500,
"global_step": 350,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"completion_length": 422.65,
"completions/clipped_ratio": 0.025,
"completions/max_length": 835.6,
"completions/max_terminated_length": 738.6,
"completions/mean_length": 422.65,
"completions/mean_terminated_length": 395.875,
"completions/min_length": 131.9,
"completions/min_terminated_length": 131.9,
"epoch": 0.0006876633200385092,
"frac_reward_zero_std": 0.3,
"grad_norm": 2.2393298149108887,
"kl": 1.021408660709858,
"learning_rate": 4.722222222222222e-06,
"loss": 0.001,
"num_tokens": 26778.0,
"reward": 4.25,
"reward_std": 0.35469383001327515,
"rewards/check_coherence/mean": 0.5625,
"rewards/check_coherence/std": 0.3129152894020081,
"rewards/check_response_quality/mean": 2.0875,
"rewards/check_response_quality/std": 0.17355985641479493,
"rewards/match_format_approximately/mean": 0.6,
"rewards/match_format_approximately/std": 0.15773502588272095,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 10
},
{
"completion_length": 334.6,
"completions/clipped_ratio": 0.025,
"completions/max_length": 824.9,
"completions/max_terminated_length": 713.2,
"completions/mean_length": 334.6,
"completions/mean_terminated_length": 299.05833435058594,
"completions/min_length": 50.2,
"completions/min_terminated_length": 50.2,
"epoch": 0.0013753266400770184,
"frac_reward_zero_std": 0.1,
"grad_norm": 5.82886266708374,
"kl": 1.5751606613397597,
"learning_rate": 2.3750000000000003e-07,
"loss": 0.0016,
"num_tokens": 51474.0,
"reward": 4.4375,
"reward_std": 0.4622055947780609,
"rewards/check_coherence/mean": 0.4875,
"rewards/check_coherence/std": 0.33231321573257444,
"rewards/check_response_quality/mean": 2.2125,
"rewards/check_response_quality/std": 0.2633414089679718,
"rewards/match_format_approximately/mean": 0.7375,
"rewards/match_format_approximately/std": 0.21933756470680238,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 20
},
{
"completion_length": 299.05,
"completions/clipped_ratio": 0.025,
"completions/max_length": 684.5,
"completions/max_terminated_length": 582.1,
"completions/mean_length": 299.05,
"completions/mean_terminated_length": 268.775,
"completions/min_length": 12.0,
"completions/min_terminated_length": 12.0,
"epoch": 0.0020629899601155273,
"frac_reward_zero_std": 0.0,
"grad_norm": 1.6133402585983276,
"kl": 1.2915641874074937,
"learning_rate": 4.1428571428571435e-06,
"loss": 0.0013,
"num_tokens": 75404.0,
"reward": 4.5625,
"reward_std": 0.6107304871082306,
"rewards/check_coherence/mean": 0.6125,
"rewards/check_coherence/std": 0.403445702791214,
"rewards/check_response_quality/mean": 2.2125,
"rewards/check_response_quality/std": 0.2752987265586853,
"rewards/match_format_approximately/mean": 0.7375,
"rewards/match_format_approximately/std": 0.2404700517654419,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 30
},
{
"completion_length": 347.025,
"completions/clipped_ratio": 0.075,
"completions/max_length": 862.6,
"completions/max_terminated_length": 509.7,
"completions/mean_length": 347.025,
"completions/mean_terminated_length": 242.34166870117187,
"completions/min_length": 58.0,
"completions/min_terminated_length": 58.0,
"epoch": 0.0027506532801540367,
"frac_reward_zero_std": 0.1,
"grad_norm": 2.190549850463867,
"kl": 0.8691788390278816,
"learning_rate": 4.936507936507937e-06,
"loss": 0.0009,
"num_tokens": 102089.0,
"reward": 4.775,
"reward_std": 0.5918105363845825,
"rewards/check_coherence/mean": 0.85,
"rewards/check_coherence/std": 0.345650315284729,
"rewards/check_response_quality/mean": 2.2,
"rewards/check_response_quality/std": 0.20347774028778076,
"rewards/match_format_approximately/mean": 0.725,
"rewards/match_format_approximately/std": 0.16547005176544188,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 40
},
{
"completion_length": 278.275,
"completions/clipped_ratio": 0.025,
"completions/max_length": 695.3,
"completions/max_terminated_length": 556.3,
"completions/mean_length": 278.275,
"completions/mean_terminated_length": 240.375,
"completions/min_length": 19.8,
"completions/min_terminated_length": 19.8,
"epoch": 0.0034383166001925457,
"frac_reward_zero_std": 0.0,
"grad_norm": 1.1459013223648071,
"kl": 0.7370456486940384,
"learning_rate": 4.793650793650794e-06,
"loss": 0.0007,
"num_tokens": 125192.0,
"reward": 4.7375,
"reward_std": 0.9061064124107361,
"rewards/check_coherence/mean": 0.8625,
"rewards/check_coherence/std": 0.4861203670501709,
"rewards/check_response_quality/mean": 2.175,
"rewards/check_response_quality/std": 0.34289742112159727,
"rewards/match_format_approximately/mean": 0.725,
"rewards/match_format_approximately/std": 0.2654700517654419,
"rewards/match_format_exactly/mean": 0.975,
"rewards/match_format_exactly/std": 0.05,
"step": 50
},
{
"completion_length": 127.625,
"completions/clipped_ratio": 0.0,
"completions/max_length": 249.7,
"completions/max_terminated_length": 249.7,
"completions/mean_length": 127.625,
"completions/mean_terminated_length": 127.625,
"completions/min_length": 11.6,
"completions/min_terminated_length": 11.6,
"epoch": 0.004125979920231055,
"frac_reward_zero_std": 0.1,
"grad_norm": 1.772578477859497,
"kl": 1.2247508466243744,
"learning_rate": 4.634920634920635e-06,
"loss": 0.0012,
"num_tokens": 140377.0,
"reward": 5.075,
"reward_std": 0.583526349067688,
"rewards/check_coherence/mean": 0.925,
"rewards/check_coherence/std": 0.4689477920532227,
"rewards/check_response_quality/mean": 2.325,
"rewards/check_response_quality/std": 0.17886751294136047,
"rewards/match_format_approximately/mean": 0.825,
"rewards/match_format_approximately/std": 0.17886751294136047,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 60
},
{
"completion_length": 161.4,
"completions/clipped_ratio": 0.0,
"completions/max_length": 423.5,
"completions/max_terminated_length": 423.5,
"completions/mean_length": 161.4,
"completions/mean_terminated_length": 161.4,
"completions/min_length": 17.8,
"completions/min_terminated_length": 17.8,
"epoch": 0.004813643240269564,
"frac_reward_zero_std": 0.0,
"grad_norm": 3.128584861755371,
"kl": 0.8261611372232437,
"learning_rate": 4.476190476190477e-06,
"loss": 0.0008,
"num_tokens": 156317.0,
"reward": 5.1625,
"reward_std": 0.6853798747062683,
"rewards/check_coherence/mean": 1.025,
"rewards/check_coherence/std": 0.4077350258827209,
"rewards/check_response_quality/mean": 2.3125,
"rewards/check_response_quality/std": 0.21582483053207396,
"rewards/match_format_approximately/mean": 0.825,
"rewards/match_format_approximately/std": 0.2,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 70
},
{
"completion_length": 120.125,
"completions/clipped_ratio": 0.0,
"completions/max_length": 248.3,
"completions/max_terminated_length": 248.3,
"completions/mean_length": 120.125,
"completions/mean_terminated_length": 120.125,
"completions/min_length": 16.5,
"completions/min_terminated_length": 16.5,
"epoch": 0.0055013065603080735,
"frac_reward_zero_std": 0.4,
"grad_norm": 2.667182683944702,
"kl": 0.8017235696315765,
"learning_rate": 4.317460317460318e-06,
"loss": 0.0008,
"num_tokens": 171822.0,
"reward": 5.4,
"reward_std": 0.41746232509613035,
"rewards/check_coherence/mean": 1.2625,
"rewards/check_coherence/std": 0.2978713572025299,
"rewards/check_response_quality/mean": 2.3125,
"rewards/check_response_quality/std": 0.11582483053207397,
"rewards/match_format_approximately/mean": 0.825,
"rewards/match_format_approximately/std": 0.1,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 80
},
{
"completion_length": 164.875,
"completions/clipped_ratio": 0.0,
"completions/max_length": 490.9,
"completions/max_terminated_length": 490.9,
"completions/mean_length": 164.875,
"completions/mean_terminated_length": 164.875,
"completions/min_length": 13.5,
"completions/min_terminated_length": 13.5,
"epoch": 0.006188969880346582,
"frac_reward_zero_std": 0.0,
"grad_norm": 1.8317680358886719,
"kl": 0.8454075694084168,
"learning_rate": 4.158730158730159e-06,
"loss": 0.0008,
"num_tokens": 189509.0,
"reward": 5.325,
"reward_std": 0.7779198408126831,
"rewards/check_coherence/mean": 1.15,
"rewards/check_coherence/std": 0.47320507764816283,
"rewards/check_response_quality/mean": 2.3375,
"rewards/check_response_quality/std": 0.21933756470680238,
"rewards/match_format_approximately/mean": 0.8375,
"rewards/match_format_approximately/std": 0.21933756470680238,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 90
},
{
"completion_length": 128.325,
"completions/clipped_ratio": 0.0,
"completions/max_length": 322.5,
"completions/max_terminated_length": 322.5,
"completions/mean_length": 128.325,
"completions/mean_terminated_length": 128.325,
"completions/min_length": 46.3,
"completions/min_terminated_length": 46.3,
"epoch": 0.006876633200385091,
"frac_reward_zero_std": 0.3,
"grad_norm": 0.003549874061718583,
"kl": 0.820201675593853,
"learning_rate": 4.000000000000001e-06,
"loss": 0.0008,
"num_tokens": 205998.0,
"reward": 5.45,
"reward_std": 0.5343478560447693,
"rewards/check_coherence/mean": 1.225,
"rewards/check_coherence/std": 0.3809401035308838,
"rewards/check_response_quality/mean": 2.3625,
"rewards/check_response_quality/std": 0.10386751294136047,
"rewards/match_format_approximately/mean": 0.8625,
"rewards/match_format_approximately/std": 0.10386751294136047,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 100
},
{
"completion_length": 147.575,
"completions/clipped_ratio": 0.0,
"completions/max_length": 452.6,
"completions/max_terminated_length": 452.6,
"completions/mean_length": 147.575,
"completions/mean_terminated_length": 147.575,
"completions/min_length": 8.8,
"completions/min_terminated_length": 8.8,
"epoch": 0.007564296520423601,
"frac_reward_zero_std": 0.1,
"grad_norm": 2.094895124435425,
"kl": 1.165158998966217,
"learning_rate": 3.857142857142858e-06,
"loss": 0.0012,
"num_tokens": 224325.0,
"reward": 5.3,
"reward_std": 0.8748959302902222,
"rewards/check_coherence/mean": 1.2125,
"rewards/check_coherence/std": 0.38273502588272096,
"rewards/check_response_quality/mean": 2.2875,
"rewards/check_response_quality/std": 0.28898603916168214,
"rewards/match_format_approximately/mean": 0.825,
"rewards/match_format_approximately/std": 0.22320507764816283,
"rewards/match_format_exactly/mean": 0.975,
"rewards/match_format_exactly/std": 0.05,
"step": 110
},
{
"completion_length": 93.475,
"completions/clipped_ratio": 0.0,
"completions/max_length": 232.9,
"completions/max_terminated_length": 232.9,
"completions/mean_length": 93.475,
"completions/mean_terminated_length": 93.475,
"completions/min_length": 8.8,
"completions/min_terminated_length": 8.8,
"epoch": 0.00825195984046211,
"frac_reward_zero_std": 0.3,
"grad_norm": 1.8320742845535278,
"kl": 0.8784003466367721,
"learning_rate": 3.6984126984126987e-06,
"loss": 0.0009,
"num_tokens": 238872.0,
"reward": 5.6125,
"reward_std": 0.5023834943771363,
"rewards/check_coherence/mean": 1.3875,
"rewards/check_coherence/std": 0.1978713572025299,
"rewards/check_response_quality/mean": 2.3625,
"rewards/check_response_quality/std": 0.18273502588272095,
"rewards/match_format_approximately/mean": 0.8625,
"rewards/match_format_approximately/std": 0.18273502588272095,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 120
},
{
"completion_length": 63.425,
"completions/clipped_ratio": 0.0,
"completions/max_length": 180.7,
"completions/max_terminated_length": 180.7,
"completions/mean_length": 63.425,
"completions/mean_terminated_length": 63.425,
"completions/min_length": 9.8,
"completions/min_terminated_length": 9.8,
"epoch": 0.008939623160500619,
"frac_reward_zero_std": 0.5,
"grad_norm": 1.388489007949829,
"kl": 0.855793622136116,
"learning_rate": 3.53968253968254e-06,
"loss": 0.0009,
"num_tokens": 252433.0,
"reward": 5.75,
"reward_std": 0.4457427144050598,
"rewards/check_coherence/mean": 1.4,
"rewards/check_coherence/std": 0.2,
"rewards/check_response_quality/mean": 2.425,
"rewards/check_response_quality/std": 0.12886751294136048,
"rewards/match_format_approximately/mean": 0.925,
"rewards/match_format_approximately/std": 0.12886751294136048,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 130
},
{
"completion_length": 41.35,
"completions/clipped_ratio": 0.0,
"completions/max_length": 127.7,
"completions/max_terminated_length": 127.7,
"completions/mean_length": 41.35,
"completions/mean_terminated_length": 41.35,
"completions/min_length": 7.7,
"completions/min_terminated_length": 7.7,
"epoch": 0.009627286480539128,
"frac_reward_zero_std": 0.0,
"grad_norm": 2.7645645141601562,
"kl": 1.5349723994731903,
"learning_rate": 3.3809523809523814e-06,
"loss": 0.0015,
"num_tokens": 264579.0,
"reward": 5.0875,
"reward_std": 1.1462337374687195,
"rewards/check_coherence/mean": 1.15,
"rewards/check_coherence/std": 0.48867512941360475,
"rewards/check_response_quality/mean": 2.1875,
"rewards/check_response_quality/std": 0.3914190471172333,
"rewards/match_format_approximately/mean": 0.85,
"rewards/match_format_approximately/std": 0.20773502588272094,
"rewards/match_format_exactly/mean": 0.9,
"rewards/match_format_exactly/std": 0.1154700517654419,
"step": 140
},
{
"completion_length": 46.625,
"completions/clipped_ratio": 0.0,
"completions/max_length": 135.3,
"completions/max_terminated_length": 135.3,
"completions/mean_length": 46.625,
"completions/mean_terminated_length": 46.625,
"completions/min_length": 10.0,
"completions/min_terminated_length": 10.0,
"epoch": 0.010314949800577638,
"frac_reward_zero_std": 0.4,
"grad_norm": 0.36925405263900757,
"kl": 1.2759959518909454,
"learning_rate": 3.2222222222222227e-06,
"loss": 0.0013,
"num_tokens": 278280.0,
"reward": 5.7,
"reward_std": 0.4686140716075897,
"rewards/check_coherence/mean": 1.3875,
"rewards/check_coherence/std": 0.225,
"rewards/check_response_quality/mean": 2.4,
"rewards/check_response_quality/std": 0.15173887014389037,
"rewards/match_format_approximately/mean": 0.9125,
"rewards/match_format_approximately/std": 0.13273502588272096,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 150
},
{
"completion_length": 37.85,
"completions/clipped_ratio": 0.0,
"completions/max_length": 108.9,
"completions/max_terminated_length": 108.9,
"completions/mean_length": 37.85,
"completions/mean_terminated_length": 37.85,
"completions/min_length": 10.3,
"completions/min_terminated_length": 10.3,
"epoch": 0.011002613120616147,
"frac_reward_zero_std": 0.5,
"grad_norm": 0.003718956606462598,
"kl": 1.1356925666332245,
"learning_rate": 3.063492063492064e-06,
"loss": 0.0011,
"num_tokens": 289698.0,
"reward": 5.8625,
"reward_std": 0.275,
"rewards/check_coherence/mean": 1.4,
"rewards/check_coherence/std": 0.2,
"rewards/check_response_quality/mean": 2.475,
"rewards/check_response_quality/std": 0.05,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 160
},
{
"completion_length": 24.475,
"completions/clipped_ratio": 0.0,
"completions/max_length": 58.6,
"completions/max_terminated_length": 58.6,
"completions/mean_length": 24.475,
"completions/mean_terminated_length": 24.475,
"completions/min_length": 10.1,
"completions/min_terminated_length": 10.1,
"epoch": 0.011690276440654656,
"frac_reward_zero_std": 0.8,
"grad_norm": 0.0026451745070517063,
"kl": 1.1320037961006164,
"learning_rate": 2.9047619047619053e-06,
"loss": 0.0011,
"num_tokens": 302601.0,
"reward": 5.9,
"reward_std": 0.14574271440505981,
"rewards/check_coherence/mean": 1.425,
"rewards/check_coherence/std": 0.10773502588272095,
"rewards/check_response_quality/mean": 2.4875,
"rewards/check_response_quality/std": 0.025,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 170
},
{
"completion_length": 65.375,
"completions/clipped_ratio": 0.025,
"completions/max_length": 216.0,
"completions/max_terminated_length": 67.0,
"completions/mean_length": 65.375,
"completions/mean_terminated_length": 26.866666793823242,
"completions/min_length": 9.4,
"completions/min_terminated_length": 9.4,
"epoch": 0.012377939760693164,
"frac_reward_zero_std": 0.3,
"grad_norm": 2.4312984943389893,
"kl": 1.0950138330459596,
"learning_rate": 2.7460317460317466e-06,
"loss": 0.0011,
"num_tokens": 317912.0,
"reward": 5.775,
"reward_std": 0.3957427144050598,
"rewards/check_coherence/mean": 1.4,
"rewards/check_coherence/std": 0.2,
"rewards/check_response_quality/mean": 2.4375,
"rewards/check_response_quality/std": 0.10386751294136047,
"rewards/match_format_approximately/mean": 0.9375,
"rewards/match_format_approximately/std": 0.10386751294136047,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 180
},
{
"completion_length": 16.275,
"completions/clipped_ratio": 0.0,
"completions/max_length": 25.2,
"completions/max_terminated_length": 25.2,
"completions/mean_length": 16.275,
"completions/mean_terminated_length": 16.275,
"completions/min_length": 9.8,
"completions/min_terminated_length": 9.8,
"epoch": 0.013065603080731673,
"frac_reward_zero_std": 0.6,
"grad_norm": 0.004265956114977598,
"kl": 1.1890327751636505,
"learning_rate": 2.587301587301588e-06,
"loss": 0.0012,
"num_tokens": 332055.0,
"reward": 5.8875,
"reward_std": 0.18273502588272095,
"rewards/check_coherence/mean": 1.3875,
"rewards/check_coherence/std": 0.18273502588272095,
"rewards/check_response_quality/mean": 2.5,
"rewards/check_response_quality/std": 0.0,
"rewards/match_format_approximately/mean": 1.0,
"rewards/match_format_approximately/std": 0.0,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 190
},
{
"completion_length": 37.9,
"completions/clipped_ratio": 0.0,
"completions/max_length": 112.8,
"completions/max_terminated_length": 112.8,
"completions/mean_length": 37.9,
"completions/mean_terminated_length": 37.9,
"completions/min_length": 8.1,
"completions/min_terminated_length": 8.1,
"epoch": 0.013753266400770183,
"frac_reward_zero_std": 0.4,
"grad_norm": 16.73357582092285,
"kl": 1.1858276724815369,
"learning_rate": 2.428571428571429e-06,
"loss": 0.0012,
"num_tokens": 344875.0,
"reward": 5.75,
"reward_std": 0.4457427144050598,
"rewards/check_coherence/mean": 1.35,
"rewards/check_coherence/std": 0.25773502588272096,
"rewards/check_response_quality/mean": 2.45,
"rewards/check_response_quality/std": 0.1,
"rewards/match_format_approximately/mean": 0.95,
"rewards/match_format_approximately/std": 0.1,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 200
},
{
"completion_length": 57.225,
"completions/clipped_ratio": 0.0,
"completions/max_length": 142.9,
"completions/max_terminated_length": 142.9,
"completions/mean_length": 57.225,
"completions/mean_terminated_length": 57.225,
"completions/min_length": 9.4,
"completions/min_terminated_length": 9.4,
"epoch": 0.014440929720808692,
"frac_reward_zero_std": 0.6,
"grad_norm": 0.006797971669584513,
"kl": 1.0238984107971192,
"learning_rate": 2.26984126984127e-06,
"loss": 0.001,
"num_tokens": 359308.0,
"reward": 5.7625,
"reward_std": 0.2946484684944153,
"rewards/check_coherence/mean": 1.375,
"rewards/check_coherence/std": 0.20773502588272094,
"rewards/check_response_quality/mean": 2.4375,
"rewards/check_response_quality/std": 0.06582483053207397,
"rewards/match_format_approximately/mean": 0.95,
"rewards/match_format_approximately/std": 0.05,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 210
},
{
"completion_length": 60.225,
"completions/clipped_ratio": 0.0,
"completions/max_length": 198.7,
"completions/max_terminated_length": 198.7,
"completions/mean_length": 60.225,
"completions/mean_terminated_length": 60.225,
"completions/min_length": 10.7,
"completions/min_terminated_length": 10.7,
"epoch": 0.015128593040847202,
"frac_reward_zero_std": 0.4,
"grad_norm": 1.2490910291671753,
"kl": 1.139722502231598,
"learning_rate": 2.1111111111111114e-06,
"loss": 0.0011,
"num_tokens": 372089.0,
"reward": 5.8,
"reward_std": 0.4,
"rewards/check_coherence/mean": 1.375,
"rewards/check_coherence/std": 0.25,
"rewards/check_response_quality/mean": 2.4625,
"rewards/check_response_quality/std": 0.075,
"rewards/match_format_approximately/mean": 0.9625,
"rewards/match_format_approximately/std": 0.075,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 220
},
{
"completion_length": 30.075,
"completions/clipped_ratio": 0.0,
"completions/max_length": 84.5,
"completions/max_terminated_length": 84.5,
"completions/mean_length": 30.075,
"completions/mean_terminated_length": 30.075,
"completions/min_length": 9.8,
"completions/min_terminated_length": 9.8,
"epoch": 0.01581625636088571,
"frac_reward_zero_std": 0.7,
"grad_norm": 0.004948398098349571,
"kl": 1.1260765612125396,
"learning_rate": 1.9523809523809527e-06,
"loss": 0.0011,
"num_tokens": 384276.0,
"reward": 5.875,
"reward_std": 0.1728713572025299,
"rewards/check_coherence/mean": 1.4375,
"rewards/check_coherence/std": 0.0978713572025299,
"rewards/check_response_quality/mean": 2.4625,
"rewards/check_response_quality/std": 0.075,
"rewards/match_format_approximately/mean": 0.975,
"rewards/match_format_approximately/std": 0.05,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 230
},
{
"completion_length": 65.25,
"completions/clipped_ratio": 0.025,
"completions/max_length": 223.1,
"completions/max_terminated_length": 65.9,
"completions/mean_length": 65.25,
"completions/mean_terminated_length": 25.908333396911623,
"completions/min_length": 9.1,
"completions/min_terminated_length": 9.1,
"epoch": 0.01650391968092422,
"frac_reward_zero_std": 0.6,
"grad_norm": 0.0049158609472215176,
"kl": 1.044507622718811,
"learning_rate": 1.7936507936507938e-06,
"loss": 0.001,
"num_tokens": 399686.0,
"reward": 5.8375,
"reward_std": 0.325,
"rewards/check_coherence/mean": 1.45,
"rewards/check_coherence/std": 0.1,
"rewards/check_response_quality/mean": 2.4375,
"rewards/check_response_quality/std": 0.125,
"rewards/match_format_approximately/mean": 0.95,
"rewards/match_format_approximately/std": 0.1,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 240
},
{
"completion_length": 34.25,
"completions/clipped_ratio": 0.0,
"completions/max_length": 97.6,
"completions/max_terminated_length": 97.6,
"completions/mean_length": 34.25,
"completions/mean_terminated_length": 34.25,
"completions/min_length": 9.6,
"completions/min_terminated_length": 9.6,
"epoch": 0.01719158300096273,
"frac_reward_zero_std": 0.6,
"grad_norm": 0.0035998751409351826,
"kl": 1.131579464673996,
"learning_rate": 1.6349206349206351e-06,
"loss": 0.0011,
"num_tokens": 412936.0,
"reward": 5.875,
"reward_std": 0.25,
"rewards/check_coherence/mean": 1.45,
"rewards/check_coherence/std": 0.1,
"rewards/check_response_quality/mean": 2.4625,
"rewards/check_response_quality/std": 0.075,
"rewards/match_format_approximately/mean": 0.9625,
"rewards/match_format_approximately/std": 0.075,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 250
},
{
"completion_length": 44.525,
"completions/clipped_ratio": 0.0,
"completions/max_length": 138.9,
"completions/max_terminated_length": 138.9,
"completions/mean_length": 44.525,
"completions/mean_terminated_length": 44.525,
"completions/min_length": 9.5,
"completions/min_terminated_length": 9.5,
"epoch": 0.017879246321001237,
"frac_reward_zero_std": 0.8,
"grad_norm": 0.003912392072379589,
"kl": 1.0597735822200776,
"learning_rate": 1.4761904761904762e-06,
"loss": 0.0011,
"num_tokens": 427709.0,
"reward": 5.9375,
"reward_std": 0.125,
"rewards/check_coherence/mean": 1.475,
"rewards/check_coherence/std": 0.05,
"rewards/check_response_quality/mean": 2.475,
"rewards/check_response_quality/std": 0.05,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 260
},
{
"completion_length": 13.175,
"completions/clipped_ratio": 0.0,
"completions/max_length": 22.9,
"completions/max_terminated_length": 22.9,
"completions/mean_length": 13.175,
"completions/mean_terminated_length": 13.175,
"completions/min_length": 7.3,
"completions/min_terminated_length": 7.3,
"epoch": 0.01856690964103975,
"frac_reward_zero_std": 0.6,
"grad_norm": 4.749101638793945,
"kl": 1.352827501296997,
"learning_rate": 1.3174603174603175e-06,
"loss": 0.0014,
"num_tokens": 437888.0,
"reward": 5.7625,
"reward_std": 0.41304759979248046,
"rewards/check_coherence/mean": 1.4125,
"rewards/check_coherence/std": 0.175,
"rewards/check_response_quality/mean": 2.4125,
"rewards/check_response_quality/std": 0.14464847445487977,
"rewards/match_format_approximately/mean": 0.9625,
"rewards/match_format_approximately/std": 0.053867512941360475,
"rewards/match_format_exactly/mean": 0.975,
"rewards/match_format_exactly/std": 0.05,
"step": 270
},
{
"completion_length": 29.05,
"completions/clipped_ratio": 0.0,
"completions/max_length": 75.4,
"completions/max_terminated_length": 75.4,
"completions/mean_length": 29.05,
"completions/mean_terminated_length": 29.05,
"completions/min_length": 9.3,
"completions/min_terminated_length": 9.3,
"epoch": 0.019254572961078256,
"frac_reward_zero_std": 0.7,
"grad_norm": 0.0032444808166474104,
"kl": 1.0743911743164063,
"learning_rate": 1.1587301587301589e-06,
"loss": 0.0011,
"num_tokens": 450566.0,
"reward": 5.8875,
"reward_std": 0.225,
"rewards/check_coherence/mean": 1.4375,
"rewards/check_coherence/std": 0.125,
"rewards/check_response_quality/mean": 2.475,
"rewards/check_response_quality/std": 0.05,
"rewards/match_format_approximately/mean": 0.975,
"rewards/match_format_approximately/std": 0.05,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 280
},
{
"completion_length": 35.0,
"completions/clipped_ratio": 0.0,
"completions/max_length": 100.7,
"completions/max_terminated_length": 100.7,
"completions/mean_length": 35.0,
"completions/mean_terminated_length": 35.0,
"completions/min_length": 9.7,
"completions/min_terminated_length": 9.7,
"epoch": 0.019942236281116764,
"frac_reward_zero_std": 0.7,
"grad_norm": 4.597572326660156,
"kl": 9.488562166690826,
"learning_rate": 1.0000000000000002e-06,
"loss": 0.0095,
"num_tokens": 462366.0,
"reward": 5.875,
"reward_std": 0.20773502588272094,
"rewards/check_coherence/mean": 1.4,
"rewards/check_coherence/std": 0.15773502588272095,
"rewards/check_response_quality/mean": 2.4875,
"rewards/check_response_quality/std": 0.025,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 290
},
{
"completion_length": 16.875,
"completions/clipped_ratio": 0.0,
"completions/max_length": 31.8,
"completions/max_terminated_length": 31.8,
"completions/mean_length": 16.875,
"completions/mean_terminated_length": 16.875,
"completions/min_length": 8.6,
"completions/min_terminated_length": 8.6,
"epoch": 0.020629899601155275,
"frac_reward_zero_std": 0.8,
"grad_norm": 0.0025285985320806503,
"kl": 1.0927811563014984,
"learning_rate": 8.412698412698414e-07,
"loss": 0.0011,
"num_tokens": 473501.0,
"reward": 5.9375,
"reward_std": 0.125,
"rewards/check_coherence/mean": 1.5,
"rewards/check_coherence/std": 0.0,
"rewards/check_response_quality/mean": 2.4625,
"rewards/check_response_quality/std": 0.075,
"rewards/match_format_approximately/mean": 0.975,
"rewards/match_format_approximately/std": 0.05,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 300
},
{
"completion_length": 34.825,
"completions/clipped_ratio": 0.0,
"completions/max_length": 104.8,
"completions/max_terminated_length": 104.8,
"completions/mean_length": 34.825,
"completions/mean_terminated_length": 34.825,
"completions/min_length": 9.6,
"completions/min_terminated_length": 9.6,
"epoch": 0.021317562921193783,
"frac_reward_zero_std": 0.9,
"grad_norm": 0.004729569889605045,
"kl": 1.0428565800189973,
"learning_rate": 6.825396825396826e-07,
"loss": 0.001,
"num_tokens": 487282.0,
"reward": 5.975,
"reward_std": 0.05,
"rewards/check_coherence/mean": 1.5,
"rewards/check_coherence/std": 0.0,
"rewards/check_response_quality/mean": 2.4875,
"rewards/check_response_quality/std": 0.025,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 310
},
{
"completion_length": 28.475,
"completions/clipped_ratio": 0.0,
"completions/max_length": 81.7,
"completions/max_terminated_length": 81.7,
"completions/mean_length": 28.475,
"completions/mean_terminated_length": 28.475,
"completions/min_length": 9.0,
"completions/min_terminated_length": 9.0,
"epoch": 0.022005226241232294,
"frac_reward_zero_std": 0.8,
"grad_norm": 6.385797500610352,
"kl": 1.2766858220100403,
"learning_rate": 5.238095238095239e-07,
"loss": 0.0013,
"num_tokens": 500121.0,
"reward": 5.925,
"reward_std": 0.15,
"rewards/check_coherence/mean": 1.45,
"rewards/check_coherence/std": 0.1,
"rewards/check_response_quality/mean": 2.4875,
"rewards/check_response_quality/std": 0.025,
"rewards/match_format_approximately/mean": 0.9875,
"rewards/match_format_approximately/std": 0.025,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 320
},
{
"completion_length": 18.75,
"completions/clipped_ratio": 0.0,
"completions/max_length": 33.8,
"completions/max_terminated_length": 33.8,
"completions/mean_length": 18.75,
"completions/mean_terminated_length": 18.75,
"completions/min_length": 8.4,
"completions/min_terminated_length": 8.4,
"epoch": 0.0226928895612708,
"frac_reward_zero_std": 0.7,
"grad_norm": 0.010602269321680069,
"kl": 1.1651100397109986,
"learning_rate": 3.6507936507936514e-07,
"loss": 0.0012,
"num_tokens": 514331.0,
"reward": 5.8875,
"reward_std": 0.18273502588272095,
"rewards/check_coherence/mean": 1.4625,
"rewards/check_coherence/std": 0.075,
"rewards/check_response_quality/mean": 2.4625,
"rewards/check_response_quality/std": 0.053867512941360475,
"rewards/match_format_approximately/mean": 0.9625,
"rewards/match_format_approximately/std": 0.053867512941360475,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 330
},
{
"completion_length": 13.8,
"completions/clipped_ratio": 0.0,
"completions/max_length": 20.7,
"completions/max_terminated_length": 20.7,
"completions/mean_length": 13.8,
"completions/mean_terminated_length": 13.8,
"completions/min_length": 8.9,
"completions/min_terminated_length": 8.9,
"epoch": 0.023380552881309313,
"frac_reward_zero_std": 0.9,
"grad_norm": 0.007794269360601902,
"kl": 1.1532993257045745,
"learning_rate": 2.0634920634920635e-07,
"loss": 0.0012,
"num_tokens": 528547.0,
"reward": 5.975,
"reward_std": 0.05,
"rewards/check_coherence/mean": 1.475,
"rewards/check_coherence/std": 0.05,
"rewards/check_response_quality/mean": 2.5,
"rewards/check_response_quality/std": 0.0,
"rewards/match_format_approximately/mean": 1.0,
"rewards/match_format_approximately/std": 0.0,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 340
},
{
"completion_length": 13.9,
"completions/clipped_ratio": 0.0,
"completions/max_length": 21.0,
"completions/max_terminated_length": 21.0,
"completions/mean_length": 13.9,
"completions/mean_terminated_length": 13.9,
"completions/min_length": 9.0,
"completions/min_terminated_length": 9.0,
"epoch": 0.02406821620134782,
"frac_reward_zero_std": 0.8,
"grad_norm": 3.0694658756256104,
"kl": 1.2572884202003478,
"learning_rate": 4.7619047619047627e-08,
"loss": 0.0013,
"num_tokens": 540139.0,
"reward": 5.95,
"reward_std": 0.1,
"rewards/check_coherence/mean": 1.45,
"rewards/check_coherence/std": 0.1,
"rewards/check_response_quality/mean": 2.5,
"rewards/check_response_quality/std": 0.0,
"rewards/match_format_approximately/mean": 1.0,
"rewards/match_format_approximately/std": 0.0,
"rewards/match_format_exactly/mean": 1.0,
"rewards/match_format_exactly/std": 0.0,
"step": 350
}
],
"logging_steps": 10,
"max_steps": 350,
"num_input_tokens_seen": 540139,
"num_train_epochs": 1,
"save_steps": 10,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}