RLCR-hotpot / trainer_state.json
Byanka's picture
Model save
d8c3758 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.9984,
"eval_steps": 50,
"global_step": 312,
"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": 0.03251953125,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 1501.4,
"completions/mean_length": 261.44921875,
"completions/mean_terminated_length": 218.59018249511718,
"completions/min_length": 2.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.016,
"grad_norm": 0.04527007043361664,
"learning_rate": 3.1249999999999997e-07,
"loss": 0.0817,
"num_tokens": 17521272.0,
"reward": 0.6406179428100586,
"reward_std": 0.49268757104873656,
"rewards/accuracy_reward": 0.2216796875,
"rewards/brier_reward": 0.37469087839126586,
"rewards/confidence_one_or_zero": 0.26728515625,
"rewards/format_reward": 0.68486328125,
"rewards/mean_confidence_reward": 0.7439393520355224,
"step": 5
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.03271484375,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 1502.0,
"completions/mean_length": 256.41748046875,
"completions/mean_terminated_length": 213.14424743652344,
"completions/min_length": 2.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.032,
"grad_norm": 0.13754118978977203,
"learning_rate": 6.249999999999999e-07,
"loss": 0.0794,
"num_tokens": 35247339.0,
"reward": 0.6558520913124084,
"reward_std": 0.46240503787994386,
"rewards/accuracy_reward": 0.2123046875,
"rewards/brier_reward": 0.37713180780410765,
"rewards/confidence_one_or_zero": 0.27294921875,
"rewards/format_reward": 0.722265625,
"rewards/mean_confidence_reward": 0.7526058673858642,
"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": 0.023046875,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 1481.6,
"completions/mean_length": 217.0734375,
"completions/mean_terminated_length": 186.05869750976564,
"completions/min_length": 2.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.048,
"grad_norm": 0.07561526447534561,
"learning_rate": 9.374999999999999e-07,
"loss": 0.074,
"num_tokens": 52518907.0,
"reward": 0.7986767530441284,
"reward_std": 0.38383460640907285,
"rewards/accuracy_reward": 0.26533203125,
"rewards/brier_reward": 0.4729374527931213,
"rewards/confidence_one_or_zero": 0.252734375,
"rewards/format_reward": 0.85908203125,
"rewards/mean_confidence_reward": 0.7597096920013428,
"step": 15
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0064453125,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 1263.0,
"completions/mean_length": 151.86923828125,
"completions/mean_terminated_length": 142.8934539794922,
"completions/min_length": 2.0,
"completions/min_terminated_length": 2.0,
"epoch": 0.064,
"grad_norm": 0.016455456614494324,
"learning_rate": 1e-06,
"loss": 0.0127,
"num_tokens": 68992448.0,
"reward": 0.935137140750885,
"reward_std": 0.29728189706802366,
"rewards/accuracy_reward": 0.33779296875,
"rewards/brier_reward": 0.5799404501914978,
"rewards/confidence_one_or_zero": 0.1966796875,
"rewards/format_reward": 0.9525390625,
"rewards/mean_confidence_reward": 0.7396757483482361,
"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": 0.0021484375,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 923.6,
"completions/mean_length": 128.238671875,
"completions/mean_terminated_length": 125.20802764892578,
"completions/min_length": 11.4,
"completions/min_terminated_length": 11.4,
"epoch": 0.08,
"grad_norm": 0.05269403010606766,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 85238764.0,
"reward": 1.0060295939445496,
"reward_std": 0.2248561441898346,
"rewards/accuracy_reward": 0.35712890625,
"rewards/brier_reward": 0.6683077096939087,
"rewards/confidence_one_or_zero": 0.09072265625,
"rewards/format_reward": 0.98662109375,
"rewards/mean_confidence_reward": 0.6470906853675842,
"step": 25
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00224609375,
"completions/max_length": 1306.2,
"completions/max_terminated_length": 937.4,
"completions/mean_length": 133.3306640625,
"completions/mean_terminated_length": 130.1755615234375,
"completions/min_length": 10.8,
"completions/min_terminated_length": 10.8,
"epoch": 0.096,
"grad_norm": 0.11787062138319016,
"learning_rate": 1e-06,
"loss": 0.0043,
"num_tokens": 101648678.0,
"reward": 1.0414881467819215,
"reward_std": 0.18446856439113618,
"rewards/accuracy_reward": 0.3625,
"rewards/brier_reward": 0.7302408814430237,
"rewards/confidence_one_or_zero": 0.04970703125,
"rewards/format_reward": 0.990234375,
"rewards/mean_confidence_reward": 0.5119157314300538,
"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": 0.00146484375,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 1072.2,
"completions/mean_length": 143.08466796875,
"completions/mean_terminated_length": 141.0427459716797,
"completions/min_length": 2.6,
"completions/min_terminated_length": 2.6,
"epoch": 0.112,
"grad_norm": 0.03638599067926407,
"learning_rate": 1e-06,
"loss": 0.004,
"num_tokens": 118223337.0,
"reward": 1.0774429082870483,
"reward_std": 0.13774871528148652,
"rewards/accuracy_reward": 0.39619140625,
"rewards/brier_reward": 0.7643576622009277,
"rewards/confidence_one_or_zero": 0.05927734375,
"rewards/format_reward": 0.9943359375,
"rewards/mean_confidence_reward": 0.37070313692092893,
"step": 35
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00048828125,
"completions/max_length": 913.2,
"completions/max_terminated_length": 723.4,
"completions/mean_length": 151.192578125,
"completions/mean_terminated_length": 150.51485290527344,
"completions/min_length": 43.2,
"completions/min_terminated_length": 43.2,
"epoch": 0.128,
"grad_norm": 0.008543482981622219,
"learning_rate": 1e-06,
"loss": 0.0007,
"num_tokens": 134688221.0,
"reward": 1.0826910495758058,
"reward_std": 0.10313712060451508,
"rewards/accuracy_reward": 0.405078125,
"rewards/brier_reward": 0.7624517321586609,
"rewards/confidence_one_or_zero": 0.0619140625,
"rewards/format_reward": 0.9978515625,
"rewards/mean_confidence_reward": 0.31029855012893676,
"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": 0.000390625,
"completions/max_length": 899.6,
"completions/max_terminated_length": 555.0,
"completions/mean_length": 155.73017578125,
"completions/mean_terminated_length": 155.19070434570312,
"completions/min_length": 57.0,
"completions/min_terminated_length": 57.0,
"epoch": 0.144,
"grad_norm": 0.010719305835664272,
"learning_rate": 1e-06,
"loss": 0.0005,
"num_tokens": 151233330.0,
"reward": 1.1194574117660523,
"reward_std": 0.10182622820138931,
"rewards/accuracy_reward": 0.4982421875,
"rewards/brier_reward": 0.7418437123298645,
"rewards/confidence_one_or_zero": 0.04111328125,
"rewards/format_reward": 0.998828125,
"rewards/mean_confidence_reward": 0.3395689487457275,
"step": 45
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.00078125,
"completions/max_length": 635.4,
"completions/max_terminated_length": 433.6,
"completions/mean_length": 156.94189453125,
"completions/mean_terminated_length": 155.86400146484374,
"completions/min_length": 61.0,
"completions/min_terminated_length": 61.0,
"epoch": 0.16,
"grad_norm": 0.02116214483976364,
"learning_rate": 1e-06,
"loss": 0.0021,
"num_tokens": 167861343.0,
"reward": 1.1079134941101074,
"reward_std": 0.10681741833686828,
"rewards/accuracy_reward": 0.45244140625,
"rewards/brier_reward": 0.7644589900970459,
"rewards/confidence_one_or_zero": 0.02294921875,
"rewards/format_reward": 0.99892578125,
"rewards/mean_confidence_reward": 0.3910071313381195,
"step": 50
},
{
"epoch": 0.16,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 351.75,
"eval_completions/max_terminated_length": 351.75,
"eval_completions/mean_length": 158.1922149658203,
"eval_completions/mean_terminated_length": 158.1922149658203,
"eval_completions/min_length": 87.0,
"eval_completions/min_terminated_length": 87.0,
"eval_loss": 0.0,
"eval_num_tokens": 167861343.0,
"eval_reward": 1.0662100315093994,
"eval_reward_std": 0.21462075412273407,
"eval_rewards/accuracy_reward": 0.357421875,
"eval_rewards/brier_reward": 0.774997279047966,
"eval_rewards/confidence_one_or_zero": 0.013671875,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.4146093651652336,
"eval_runtime": 21.8666,
"eval_samples_per_second": 22.866,
"eval_steps_per_second": 0.183,
"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": 0.000390625,
"completions/max_length": 915.6,
"completions/max_terminated_length": 474.8,
"completions/mean_length": 159.457421875,
"completions/mean_terminated_length": 158.91932067871093,
"completions/min_length": 45.8,
"completions/min_terminated_length": 45.8,
"epoch": 0.176,
"grad_norm": 0.009603990241885185,
"learning_rate": 1e-06,
"loss": 0.0014,
"num_tokens": 184731307.0,
"reward": 1.1049310684204101,
"reward_std": 0.11344930976629257,
"rewards/accuracy_reward": 0.441796875,
"rewards/brier_reward": 0.769236147403717,
"rewards/confidence_one_or_zero": 0.01845703125,
"rewards/format_reward": 0.998828125,
"rewards/mean_confidence_reward": 0.4387405276298523,
"step": 55
},
{
"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": 0.0005859375,
"completions/max_length": 995.2,
"completions/max_terminated_length": 543.0,
"completions/mean_length": 163.6873046875,
"completions/mean_terminated_length": 162.88194580078124,
"completions/min_length": 77.4,
"completions/min_terminated_length": 77.4,
"epoch": 0.192,
"grad_norm": 0.007208758499473333,
"learning_rate": 1e-06,
"loss": 0.0024,
"num_tokens": 201222281.0,
"reward": 1.117084002494812,
"reward_std": 0.11234763264656067,
"rewards/accuracy_reward": 0.4650390625,
"rewards/brier_reward": 0.7699091911315918,
"rewards/confidence_one_or_zero": 0.0126953125,
"rewards/format_reward": 0.99921875,
"rewards/mean_confidence_reward": 0.47076983451843263,
"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": 0.000390625,
"completions/max_length": 695.4,
"completions/max_terminated_length": 507.2,
"completions/mean_length": 167.3060546875,
"completions/mean_terminated_length": 166.7700225830078,
"completions/min_length": 62.8,
"completions/min_terminated_length": 62.8,
"epoch": 0.208,
"grad_norm": 0.010833042673766613,
"learning_rate": 1e-06,
"loss": 0.0016,
"num_tokens": 217967719.0,
"reward": 1.1401428937911988,
"reward_std": 0.11306387037038804,
"rewards/accuracy_reward": 0.5169921875,
"rewards/brier_reward": 0.7639761686325073,
"rewards/confidence_one_or_zero": 0.00966796875,
"rewards/format_reward": 0.99931640625,
"rewards/mean_confidence_reward": 0.48724169135093687,
"step": 65
},
{
"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": 9.765625e-05,
"completions/max_length": 730.4,
"completions/max_terminated_length": 518.4,
"completions/mean_length": 166.8869140625,
"completions/mean_terminated_length": 166.7537078857422,
"completions/min_length": 81.8,
"completions/min_terminated_length": 81.8,
"epoch": 0.224,
"grad_norm": 0.00990777276456356,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 234829825.0,
"reward": 1.117465901374817,
"reward_std": 0.11148046851158142,
"rewards/accuracy_reward": 0.4650390625,
"rewards/brier_reward": 0.7703800439834595,
"rewards/confidence_one_or_zero": 0.0109375,
"rewards/format_reward": 0.99951171875,
"rewards/mean_confidence_reward": 0.48857617378234863,
"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": 0.0005859375,
"completions/max_length": 1110.6,
"completions/max_terminated_length": 485.6,
"completions/mean_length": 173.951171875,
"completions/mean_terminated_length": 173.1517333984375,
"completions/min_length": 81.6,
"completions/min_terminated_length": 81.6,
"epoch": 0.24,
"grad_norm": 0.012188726104795933,
"learning_rate": 1e-06,
"loss": 0.0026,
"num_tokens": 251862765.0,
"reward": 1.1460721731185912,
"reward_std": 0.1271799236536026,
"rewards/accuracy_reward": 0.52431640625,
"rewards/brier_reward": 0.7685105323791503,
"rewards/confidence_one_or_zero": 0.00810546875,
"rewards/format_reward": 0.99931640625,
"rewards/mean_confidence_reward": 0.5015208303928376,
"step": 75
},
{
"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": 9.765625e-05,
"completions/max_length": 779.0,
"completions/max_terminated_length": 570.2,
"completions/mean_length": 171.87490234375,
"completions/mean_terminated_length": 171.741259765625,
"completions/min_length": 83.8,
"completions/min_terminated_length": 83.8,
"epoch": 0.256,
"grad_norm": 0.00730907404795289,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 268677580.0,
"reward": 1.13645658493042,
"reward_std": 0.11724818050861359,
"rewards/accuracy_reward": 0.49716796875,
"rewards/brier_reward": 0.7759395122528077,
"rewards/confidence_one_or_zero": 0.0126953125,
"rewards/format_reward": 0.9998046875,
"rewards/mean_confidence_reward": 0.5036328196525574,
"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": 0.00048828125,
"completions/max_length": 863.2,
"completions/max_terminated_length": 460.8,
"completions/mean_length": 177.0607421875,
"completions/mean_terminated_length": 176.3964630126953,
"completions/min_length": 57.6,
"completions/min_terminated_length": 57.6,
"epoch": 0.272,
"grad_norm": 0.009313421323895454,
"learning_rate": 1e-06,
"loss": 0.0016,
"num_tokens": 285456378.0,
"reward": 1.1312544345855713,
"reward_std": 0.11844458729028702,
"rewards/accuracy_reward": 0.49052734375,
"rewards/brier_reward": 0.7730547547340393,
"rewards/confidence_one_or_zero": 0.0130859375,
"rewards/format_reward": 0.99892578125,
"rewards/mean_confidence_reward": 0.4925888657569885,
"step": 85
},
{
"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": 0.0001953125,
"completions/max_length": 890.4,
"completions/max_terminated_length": 446.2,
"completions/mean_length": 173.14130859375,
"completions/mean_terminated_length": 172.87517395019532,
"completions/min_length": 66.2,
"completions/min_terminated_length": 66.2,
"epoch": 0.288,
"grad_norm": 0.021724838763475418,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 302187521.0,
"reward": 1.1318594217300415,
"reward_std": 0.11806258261203766,
"rewards/accuracy_reward": 0.4912109375,
"rewards/brier_reward": 0.7728975296020508,
"rewards/confidence_one_or_zero": 0.01240234375,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.4906367301940918,
"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": 0.0005859375,
"completions/max_length": 1014.6,
"completions/max_terminated_length": 857.8,
"completions/mean_length": 176.462890625,
"completions/mean_terminated_length": 175.66776123046876,
"completions/min_length": 62.4,
"completions/min_terminated_length": 62.4,
"epoch": 0.304,
"grad_norm": 0.006397055462002754,
"learning_rate": 1e-06,
"loss": 0.0018,
"num_tokens": 318924453.0,
"reward": 1.1344104766845704,
"reward_std": 0.11458559930324555,
"rewards/accuracy_reward": 0.4953125,
"rewards/brier_reward": 0.7745817184448243,
"rewards/confidence_one_or_zero": 0.01767578125,
"rewards/format_reward": 0.99892578125,
"rewards/mean_confidence_reward": 0.47802149057388305,
"step": 95
},
{
"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": 0.0009765625,
"completions/max_length": 1325.6,
"completions/max_terminated_length": 647.8,
"completions/mean_length": 176.624609375,
"completions/mean_terminated_length": 175.29497375488282,
"completions/min_length": 66.0,
"completions/min_terminated_length": 66.0,
"epoch": 0.32,
"grad_norm": 0.13792400062084198,
"learning_rate": 1e-06,
"loss": 0.0028,
"num_tokens": 335821793.0,
"reward": 1.1494127273559571,
"reward_std": 0.10425310283899307,
"rewards/accuracy_reward": 0.51708984375,
"rewards/brier_reward": 0.7831019043922425,
"rewards/confidence_one_or_zero": 0.01220703125,
"rewards/format_reward": 0.9986328125,
"rewards/mean_confidence_reward": 0.4873457133769989,
"step": 100
},
{
"epoch": 0.32,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 408.0,
"eval_completions/max_terminated_length": 408.0,
"eval_completions/mean_length": 175.2724952697754,
"eval_completions/mean_terminated_length": 175.2724952697754,
"eval_completions/min_length": 94.75,
"eval_completions/min_terminated_length": 94.75,
"eval_loss": 0.0,
"eval_num_tokens": 335821793.0,
"eval_reward": 1.0855459570884705,
"eval_reward_std": 0.25649269297719,
"eval_rewards/accuracy_reward": 0.392578125,
"eval_rewards/brier_reward": 0.7785128951072693,
"eval_rewards/confidence_one_or_zero": 0.021484375,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.47496095299720764,
"eval_runtime": 22.7529,
"eval_samples_per_second": 21.975,
"eval_steps_per_second": 0.176,
"step": 100
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.001171875,
"completions/max_length": 1462.2,
"completions/max_terminated_length": 1011.0,
"completions/mean_length": 175.45908203125,
"completions/mean_terminated_length": 173.86156005859374,
"completions/min_length": 49.6,
"completions/min_terminated_length": 49.6,
"epoch": 0.336,
"grad_norm": 0.017879005521535873,
"learning_rate": 1e-06,
"loss": 0.0028,
"num_tokens": 352340926.0,
"reward": 1.1430244445800781,
"reward_std": 0.12031411826610565,
"rewards/accuracy_reward": 0.51044921875,
"rewards/brier_reward": 0.7780401229858398,
"rewards/confidence_one_or_zero": 0.015625,
"rewards/format_reward": 0.99755859375,
"rewards/mean_confidence_reward": 0.4848623156547546,
"step": 105
},
{
"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": 0.0017578125,
"completions/max_length": 1461.4,
"completions/max_terminated_length": 807.2,
"completions/mean_length": 180.69208984375,
"completions/mean_terminated_length": 178.3111328125,
"completions/min_length": 1.0,
"completions/min_terminated_length": 1.0,
"epoch": 0.352,
"grad_norm": 0.039632029831409454,
"learning_rate": 1e-06,
"loss": 0.0025,
"num_tokens": 369451629.0,
"reward": 1.1116411447525025,
"reward_std": 0.12863886207342148,
"rewards/accuracy_reward": 0.4513671875,
"rewards/brier_reward": 0.7790430426597595,
"rewards/confidence_one_or_zero": 0.01787109375,
"rewards/format_reward": 0.99287109375,
"rewards/mean_confidence_reward": 0.45101758241653445,
"step": 110
},
{
"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": 0.00048828125,
"completions/max_length": 1182.2,
"completions/max_terminated_length": 745.6,
"completions/mean_length": 183.9287109375,
"completions/mean_terminated_length": 183.26663513183593,
"completions/min_length": 7.2,
"completions/min_terminated_length": 7.2,
"epoch": 0.368,
"grad_norm": 0.03524937480688095,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 386400531.0,
"reward": 1.1271000146865844,
"reward_std": 0.11355644613504409,
"rewards/accuracy_reward": 0.475,
"rewards/brier_reward": 0.7824217438697815,
"rewards/confidence_one_or_zero": 0.01640625,
"rewards/format_reward": 0.99677734375,
"rewards/mean_confidence_reward": 0.465801864862442,
"step": 115
},
{
"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": 0.001171875,
"completions/max_length": 1536.0,
"completions/max_terminated_length": 745.8,
"completions/mean_length": 184.33974609375,
"completions/mean_terminated_length": 182.75419921875,
"completions/min_length": 62.2,
"completions/min_terminated_length": 62.2,
"epoch": 0.384,
"grad_norm": 0.017022427171468735,
"learning_rate": 1e-06,
"loss": 0.004,
"num_tokens": 403144682.0,
"reward": 1.1463651657104492,
"reward_std": 0.12106073200702668,
"rewards/accuracy_reward": 0.506640625,
"rewards/brier_reward": 0.7902879238128662,
"rewards/confidence_one_or_zero": 0.02177734375,
"rewards/format_reward": 0.99580078125,
"rewards/mean_confidence_reward": 0.46812206506729126,
"step": 120
},
{
"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": 0.00126953125,
"completions/max_length": 1335.0,
"completions/max_terminated_length": 753.8,
"completions/mean_length": 184.96728515625,
"completions/mean_terminated_length": 183.24875183105468,
"completions/min_length": 60.2,
"completions/min_terminated_length": 60.2,
"epoch": 0.4,
"grad_norm": 0.017970727756619453,
"learning_rate": 1e-06,
"loss": 0.0022,
"num_tokens": 420075195.0,
"reward": 1.1312126159667968,
"reward_std": 0.12662589848041533,
"rewards/accuracy_reward": 0.48759765625,
"rewards/brier_reward": 0.7792211294174194,
"rewards/confidence_one_or_zero": 0.0115234375,
"rewards/format_reward": 0.99560546875,
"rewards/mean_confidence_reward": 0.5128134965896607,
"step": 125
},
{
"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": 0.001171875,
"completions/max_length": 1323.8,
"completions/max_terminated_length": 799.8,
"completions/mean_length": 190.72177734375,
"completions/mean_terminated_length": 189.14300231933595,
"completions/min_length": 83.2,
"completions/min_terminated_length": 83.2,
"epoch": 0.416,
"grad_norm": 0.01795245334506035,
"learning_rate": 1e-06,
"loss": 0.0039,
"num_tokens": 436909370.0,
"reward": 1.1423231840133667,
"reward_std": 0.12082898765802383,
"rewards/accuracy_reward": 0.49794921875,
"rewards/brier_reward": 0.7881609320640564,
"rewards/confidence_one_or_zero": 0.00966796875,
"rewards/format_reward": 0.99853515625,
"rewards/mean_confidence_reward": 0.5266152441501617,
"step": 130
},
{
"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": 0.0005859375,
"completions/max_length": 1316.6,
"completions/max_terminated_length": 564.4,
"completions/mean_length": 189.984765625,
"completions/mean_terminated_length": 189.19561157226562,
"completions/min_length": 65.4,
"completions/min_terminated_length": 65.4,
"epoch": 0.432,
"grad_norm": 0.12305645644664764,
"learning_rate": 1e-06,
"loss": 0.0022,
"num_tokens": 453869150.0,
"reward": 1.167378830909729,
"reward_std": 0.11343645602464676,
"rewards/accuracy_reward": 0.53447265625,
"rewards/brier_reward": 0.8010651707649231,
"rewards/confidence_one_or_zero": 0.00986328125,
"rewards/format_reward": 0.99921875,
"rewards/mean_confidence_reward": 0.5294726729393006,
"step": 135
},
{
"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": 0.000390625,
"completions/max_length": 1255.2,
"completions/max_terminated_length": 826.8,
"completions/mean_length": 201.4958984375,
"completions/mean_terminated_length": 200.97523803710936,
"completions/min_length": 76.6,
"completions/min_terminated_length": 76.6,
"epoch": 0.448,
"grad_norm": 0.01579131931066513,
"learning_rate": 1e-06,
"loss": 0.0017,
"num_tokens": 470885268.0,
"reward": 1.1469329595565796,
"reward_std": 0.1118901401758194,
"rewards/accuracy_reward": 0.49921875,
"rewards/brier_reward": 0.7957204103469848,
"rewards/confidence_one_or_zero": 0.0072265625,
"rewards/format_reward": 0.99892578125,
"rewards/mean_confidence_reward": 0.5328691601753235,
"step": 140
},
{
"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": 0.0001953125,
"completions/max_length": 736.6,
"completions/max_terminated_length": 513.0,
"completions/mean_length": 211.05078125,
"completions/mean_terminated_length": 210.79130859375,
"completions/min_length": 81.8,
"completions/min_terminated_length": 81.8,
"epoch": 0.464,
"grad_norm": 0.011122009716928005,
"learning_rate": 1e-06,
"loss": 0.0008,
"num_tokens": 488217244.0,
"reward": 1.1156266927719116,
"reward_std": 0.10946736484766006,
"rewards/accuracy_reward": 0.4498046875,
"rewards/brier_reward": 0.781838345527649,
"rewards/confidence_one_or_zero": 0.009765625,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5123164117336273,
"step": 145
},
{
"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": 0.0001953125,
"completions/max_length": 999.8,
"completions/max_terminated_length": 586.0,
"completions/mean_length": 209.76787109375,
"completions/mean_terminated_length": 209.50829467773437,
"completions/min_length": 94.0,
"completions/min_terminated_length": 94.0,
"epoch": 0.48,
"grad_norm": 0.011441366747021675,
"learning_rate": 1e-06,
"loss": 0.0006,
"num_tokens": 505413299.0,
"reward": 1.1479370832443236,
"reward_std": 0.10615950524806976,
"rewards/accuracy_reward": 0.5056640625,
"rewards/brier_reward": 0.7905020475387573,
"rewards/confidence_one_or_zero": 0.01357421875,
"rewards/format_reward": 0.99970703125,
"rewards/mean_confidence_reward": 0.49698535799980165,
"step": 150
},
{
"epoch": 0.48,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 454.75,
"eval_completions/max_terminated_length": 454.75,
"eval_completions/mean_length": 213.07852935791016,
"eval_completions/mean_terminated_length": 213.07852935791016,
"eval_completions/min_length": 122.75,
"eval_completions/min_terminated_length": 122.75,
"eval_loss": 0.0,
"eval_num_tokens": 505413299.0,
"eval_reward": 1.1056718528270721,
"eval_reward_std": 0.2651190534234047,
"eval_rewards/accuracy_reward": 0.41015625,
"eval_rewards/brier_reward": 0.8011865168809891,
"eval_rewards/confidence_one_or_zero": 0.005859375,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.4764648526906967,
"eval_runtime": 25.1582,
"eval_samples_per_second": 19.874,
"eval_steps_per_second": 0.159,
"step": 150
},
{
"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": 575.8,
"completions/max_terminated_length": 575.8,
"completions/mean_length": 212.034765625,
"completions/mean_terminated_length": 212.034765625,
"completions/min_length": 104.2,
"completions/min_terminated_length": 104.2,
"epoch": 0.496,
"grad_norm": 0.0033662207424640656,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 522892375.0,
"reward": 1.1667110204696656,
"reward_std": 0.1078746810555458,
"rewards/accuracy_reward": 0.5400390625,
"rewards/brier_reward": 0.7934796333312988,
"rewards/confidence_one_or_zero": 0.01025390625,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.5157265901565552,
"step": 155
},
{
"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": 0.0001953125,
"completions/max_length": 924.6,
"completions/max_terminated_length": 532.0,
"completions/mean_length": 210.298828125,
"completions/mean_terminated_length": 210.0409729003906,
"completions/min_length": 97.4,
"completions/min_terminated_length": 97.4,
"epoch": 0.512,
"grad_norm": 0.012034610845148563,
"learning_rate": 1e-06,
"loss": 0.0012,
"num_tokens": 540191499.0,
"reward": 1.1708500623703002,
"reward_std": 0.11257555186748505,
"rewards/accuracy_reward": 0.5369140625,
"rewards/brier_reward": 0.8050778865814209,
"rewards/confidence_one_or_zero": 0.00703125,
"rewards/format_reward": 0.99970703125,
"rewards/mean_confidence_reward": 0.5358349800109863,
"step": 160
},
{
"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": 0.00029296875,
"completions/max_length": 1004.8,
"completions/max_terminated_length": 661.4,
"completions/mean_length": 207.5166015625,
"completions/mean_terminated_length": 207.12822265625,
"completions/min_length": 95.0,
"completions/min_terminated_length": 95.0,
"epoch": 0.528,
"grad_norm": 0.015975475311279297,
"learning_rate": 1e-06,
"loss": 0.001,
"num_tokens": 557346005.0,
"reward": 1.1701508998870849,
"reward_std": 0.10177138149738311,
"rewards/accuracy_reward": 0.5318359375,
"rewards/brier_reward": 0.8088555216789246,
"rewards/confidence_one_or_zero": 0.00927734375,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5180878877639771,
"step": 165
},
{
"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": 9.765625e-05,
"completions/max_length": 768.2,
"completions/max_terminated_length": 551.8,
"completions/mean_length": 209.73466796875,
"completions/mean_terminated_length": 209.60549926757812,
"completions/min_length": 101.0,
"completions/min_terminated_length": 101.0,
"epoch": 0.544,
"grad_norm": 0.0062843854539096355,
"learning_rate": 1e-06,
"loss": 0.0003,
"num_tokens": 574657272.0,
"reward": 1.1808686017990113,
"reward_std": 0.10841628313064575,
"rewards/accuracy_reward": 0.561328125,
"rewards/brier_reward": 0.8007986664772033,
"rewards/confidence_one_or_zero": 0.00556640625,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5436318397521973,
"step": 170
},
{
"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": 0.00029296875,
"completions/max_length": 918.4,
"completions/max_terminated_length": 479.4,
"completions/mean_length": 208.92236328125,
"completions/mean_terminated_length": 208.53166809082032,
"completions/min_length": 104.0,
"completions/min_terminated_length": 104.0,
"epoch": 0.56,
"grad_norm": 0.012134869582951069,
"learning_rate": 1e-06,
"loss": 0.0014,
"num_tokens": 591618045.0,
"reward": 1.1645673036575317,
"reward_std": 0.09899833053350449,
"rewards/accuracy_reward": 0.51962890625,
"rewards/brier_reward": 0.809797728061676,
"rewards/confidence_one_or_zero": 0.0025390625,
"rewards/format_reward": 0.99970703125,
"rewards/mean_confidence_reward": 0.5343828201293945,
"step": 175
},
{
"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": 0.000390625,
"completions/max_length": 1103.4,
"completions/max_terminated_length": 482.4,
"completions/mean_length": 208.23701171875,
"completions/mean_terminated_length": 207.72026672363282,
"completions/min_length": 101.0,
"completions/min_terminated_length": 101.0,
"epoch": 0.576,
"grad_norm": 0.014361800625920296,
"learning_rate": 1e-06,
"loss": 0.0014,
"num_tokens": 608937016.0,
"reward": 1.153262686729431,
"reward_std": 0.09391747117042541,
"rewards/accuracy_reward": 0.50830078125,
"rewards/brier_reward": 0.7986142754554748,
"rewards/confidence_one_or_zero": 0.0078125,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5241601765155792,
"step": 180
},
{
"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": 0.000390625,
"completions/max_length": 1119.4,
"completions/max_terminated_length": 530.6,
"completions/mean_length": 206.19326171875,
"completions/mean_terminated_length": 205.6750518798828,
"completions/min_length": 102.8,
"completions/min_terminated_length": 102.8,
"epoch": 0.592,
"grad_norm": 0.008724602870643139,
"learning_rate": 1e-06,
"loss": 0.0015,
"num_tokens": 626216147.0,
"reward": 1.160974383354187,
"reward_std": 0.0895046427845955,
"rewards/accuracy_reward": 0.51416015625,
"rewards/brier_reward": 0.8081783413887024,
"rewards/confidence_one_or_zero": 0.0060546875,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5143349647521973,
"step": 185
},
{
"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": 9.765625e-05,
"completions/max_length": 663.0,
"completions/max_terminated_length": 556.8,
"completions/mean_length": 205.84892578125,
"completions/mean_terminated_length": 205.7189514160156,
"completions/min_length": 100.0,
"completions/min_terminated_length": 100.0,
"epoch": 0.608,
"grad_norm": 0.008571009151637554,
"learning_rate": 1e-06,
"loss": 0.0005,
"num_tokens": 643323528.0,
"reward": 1.170970630645752,
"reward_std": 0.0794813334941864,
"rewards/accuracy_reward": 0.5240234375,
"rewards/brier_reward": 0.8181121468544006,
"rewards/confidence_one_or_zero": 0.00751953125,
"rewards/format_reward": 0.9998046875,
"rewards/mean_confidence_reward": 0.48931640982627866,
"step": 190
},
{
"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": 573.0,
"completions/max_terminated_length": 573.0,
"completions/mean_length": 204.29423828125,
"completions/mean_terminated_length": 204.29423828125,
"completions/min_length": 99.4,
"completions/min_terminated_length": 99.4,
"epoch": 0.624,
"grad_norm": 0.006929480005055666,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 660759405.0,
"reward": 1.1707647800445558,
"reward_std": 0.08714393228292465,
"rewards/accuracy_reward": 0.5267578125,
"rewards/brier_reward": 0.8147708535194397,
"rewards/confidence_one_or_zero": 0.00556640625,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.4799599587917328,
"step": 195
},
{
"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": 0.00048828125,
"completions/max_length": 876.8,
"completions/max_terminated_length": 465.2,
"completions/mean_length": 204.7708984375,
"completions/mean_terminated_length": 204.12095947265624,
"completions/min_length": 101.4,
"completions/min_terminated_length": 101.4,
"epoch": 0.64,
"grad_norm": 0.006568757817149162,
"learning_rate": 1e-06,
"loss": 0.0015,
"num_tokens": 678198947.0,
"reward": 1.188051414489746,
"reward_std": 0.08363442420959473,
"rewards/accuracy_reward": 0.56201171875,
"rewards/brier_reward": 0.8145784258842468,
"rewards/confidence_one_or_zero": 0.0029296875,
"rewards/format_reward": 0.99951171875,
"rewards/mean_confidence_reward": 0.5187148451805115,
"step": 200
},
{
"epoch": 0.64,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 362.5,
"eval_completions/max_terminated_length": 362.5,
"eval_completions/mean_length": 202.62493133544922,
"eval_completions/mean_terminated_length": 202.62493133544922,
"eval_completions/min_length": 116.25,
"eval_completions/min_terminated_length": 116.25,
"eval_loss": 0.0,
"eval_num_tokens": 678198947.0,
"eval_reward": 1.1124199032783508,
"eval_reward_std": 0.2762097716331482,
"eval_rewards/accuracy_reward": 0.41796875,
"eval_rewards/brier_reward": 0.8068701177835464,
"eval_rewards/confidence_one_or_zero": 0.001953125,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.5002929642796516,
"eval_runtime": 20.9461,
"eval_samples_per_second": 23.871,
"eval_steps_per_second": 0.191,
"step": 200
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.000390625,
"completions/max_length": 1170.6,
"completions/max_terminated_length": 699.2,
"completions/mean_length": 200.96318359375,
"completions/mean_terminated_length": 200.4418518066406,
"completions/min_length": 99.0,
"completions/min_terminated_length": 99.0,
"epoch": 0.656,
"grad_norm": 0.0061631170101463795,
"learning_rate": 1e-06,
"loss": 0.0011,
"num_tokens": 695113354.0,
"reward": 1.1497117042541505,
"reward_std": 0.09324304014444351,
"rewards/accuracy_reward": 0.50517578125,
"rewards/brier_reward": 0.7946373224258423,
"rewards/confidence_one_or_zero": 0.00361328125,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5214970707893372,
"step": 205
},
{
"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": 9.765625e-05,
"completions/max_length": 666.0,
"completions/max_terminated_length": 434.4,
"completions/mean_length": 199.86826171875,
"completions/mean_terminated_length": 199.73834228515625,
"completions/min_length": 101.8,
"completions/min_terminated_length": 101.8,
"epoch": 0.672,
"grad_norm": 0.005808599293231964,
"learning_rate": 1e-06,
"loss": 0.0006,
"num_tokens": 712073445.0,
"reward": 1.168108344078064,
"reward_std": 0.08731473982334137,
"rewards/accuracy_reward": 0.5232421875,
"rewards/brier_reward": 0.8130711436271667,
"rewards/confidence_one_or_zero": 0.003515625,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.5128886938095093,
"step": 210
},
{
"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": 9.765625e-05,
"completions/max_length": 675.2,
"completions/max_terminated_length": 629.4,
"completions/mean_length": 205.4484375,
"completions/mean_terminated_length": 205.31842651367188,
"completions/min_length": 104.0,
"completions/min_terminated_length": 104.0,
"epoch": 0.688,
"grad_norm": 0.005605604965239763,
"learning_rate": 1e-06,
"loss": 0.0003,
"num_tokens": 729131157.0,
"reward": 1.1653831958770753,
"reward_std": 0.09237445890903473,
"rewards/accuracy_reward": 0.5251953125,
"rewards/brier_reward": 0.8056677460670472,
"rewards/confidence_one_or_zero": 0.0005859375,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.5064306616783142,
"step": 215
},
{
"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": 411.0,
"completions/max_terminated_length": 411.0,
"completions/mean_length": 202.72197265625,
"completions/mean_terminated_length": 202.72197265625,
"completions/min_length": 104.2,
"completions/min_terminated_length": 104.2,
"epoch": 0.704,
"grad_norm": 0.0074613383039832115,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 746073174.0,
"reward": 1.176201581954956,
"reward_std": 0.07743649333715438,
"rewards/accuracy_reward": 0.53837890625,
"rewards/brier_reward": 0.8140231966972351,
"rewards/confidence_one_or_zero": 0.00517578125,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.5200888633728027,
"step": 220
},
{
"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": 656.2,
"completions/max_terminated_length": 656.2,
"completions/mean_length": 202.00751953125,
"completions/mean_terminated_length": 202.00751953125,
"completions/min_length": 99.4,
"completions/min_terminated_length": 99.4,
"epoch": 0.72,
"grad_norm": 0.008289781399071217,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 763151587.0,
"reward": 1.186832022666931,
"reward_std": 0.08442019075155258,
"rewards/accuracy_reward": 0.55576171875,
"rewards/brier_reward": 0.817901360988617,
"rewards/confidence_one_or_zero": 0.0025390625,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.5230859518051147,
"step": 225
},
{
"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": 0.0005859375,
"completions/max_length": 647.4,
"completions/max_terminated_length": 516.6,
"completions/mean_length": 205.50849609375,
"completions/mean_terminated_length": 204.73146057128906,
"completions/min_length": 97.6,
"completions/min_terminated_length": 97.6,
"epoch": 0.736,
"grad_norm": 0.008065270259976387,
"learning_rate": 1e-06,
"loss": 0.0016,
"num_tokens": 780195578.0,
"reward": 1.1865583658218384,
"reward_std": 0.08369777351617813,
"rewards/accuracy_reward": 0.56474609375,
"rewards/brier_reward": 0.8089555978775025,
"rewards/confidence_one_or_zero": 0.003125,
"rewards/format_reward": 0.9994140625,
"rewards/mean_confidence_reward": 0.5270449399948121,
"step": 230
},
{
"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": 606.4,
"completions/max_terminated_length": 606.4,
"completions/mean_length": 206.17548828125,
"completions/mean_terminated_length": 206.17548828125,
"completions/min_length": 102.2,
"completions/min_terminated_length": 102.2,
"epoch": 0.752,
"grad_norm": 0.009517704136669636,
"learning_rate": 1e-06,
"loss": 0.0003,
"num_tokens": 797534015.0,
"reward": 1.1857917547225951,
"reward_std": 0.08198632448911666,
"rewards/accuracy_reward": 0.5568359375,
"rewards/brier_reward": 0.8151372194290161,
"rewards/confidence_one_or_zero": 0.005859375,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5247080326080322,
"step": 235
},
{
"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": 0.000390625,
"completions/max_length": 1174.4,
"completions/max_terminated_length": 875.4,
"completions/mean_length": 215.8525390625,
"completions/mean_terminated_length": 215.33623962402345,
"completions/min_length": 103.4,
"completions/min_terminated_length": 103.4,
"epoch": 0.768,
"grad_norm": 0.02318732999265194,
"learning_rate": 1e-06,
"loss": 0.0013,
"num_tokens": 814677049.0,
"reward": 1.1634629249572754,
"reward_std": 0.08838685750961303,
"rewards/accuracy_reward": 0.50859375,
"rewards/brier_reward": 0.8190146684646606,
"rewards/confidence_one_or_zero": 0.00322265625,
"rewards/format_reward": 0.99931640625,
"rewards/mean_confidence_reward": 0.535248053073883,
"step": 240
},
{
"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": 467.2,
"completions/max_terminated_length": 467.2,
"completions/mean_length": 216.6044921875,
"completions/mean_terminated_length": 216.6044921875,
"completions/min_length": 108.2,
"completions/min_terminated_length": 108.2,
"epoch": 0.784,
"grad_norm": 0.02482554130256176,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 832069447.0,
"reward": 1.1771643161773682,
"reward_std": 0.09045170843601227,
"rewards/accuracy_reward": 0.55361328125,
"rewards/brier_reward": 0.8010073304176331,
"rewards/confidence_one_or_zero": 0.00498046875,
"rewards/format_reward": 0.99970703125,
"rewards/mean_confidence_reward": 0.5427734732627869,
"step": 245
},
{
"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": 497.2,
"completions/max_terminated_length": 497.2,
"completions/mean_length": 218.06123046875,
"completions/mean_terminated_length": 218.06123046875,
"completions/min_length": 113.6,
"completions/min_terminated_length": 113.6,
"epoch": 0.8,
"grad_norm": 0.005663494113832712,
"learning_rate": 1e-06,
"loss": 0.0002,
"num_tokens": 849312954.0,
"reward": 1.2057418823242188,
"reward_std": 0.08701288551092148,
"rewards/accuracy_reward": 0.587109375,
"rewards/brier_reward": 0.8243734002113342,
"rewards/confidence_one_or_zero": 0.0033203125,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.5567070245742798,
"step": 250
},
{
"epoch": 0.8,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 408.0,
"eval_completions/max_terminated_length": 408.0,
"eval_completions/mean_length": 220.35998153686523,
"eval_completions/mean_terminated_length": 220.35998153686523,
"eval_completions/min_length": 134.5,
"eval_completions/min_terminated_length": 134.5,
"eval_loss": 0.0,
"eval_num_tokens": 849312954.0,
"eval_reward": 1.118152379989624,
"eval_reward_std": 0.28543028980493546,
"eval_rewards/accuracy_reward": 0.435546875,
"eval_rewards/brier_reward": 0.8007568567991257,
"eval_rewards/confidence_one_or_zero": 0.001953125,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.5190429389476776,
"eval_runtime": 22.0758,
"eval_samples_per_second": 22.649,
"eval_steps_per_second": 0.181,
"step": 250
},
{
"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": 425.8,
"completions/max_terminated_length": 425.8,
"completions/mean_length": 218.22822265625,
"completions/mean_terminated_length": 218.22822265625,
"completions/min_length": 110.4,
"completions/min_terminated_length": 110.4,
"epoch": 0.816,
"grad_norm": 0.021623145788908005,
"learning_rate": 1e-06,
"loss": 0.0003,
"num_tokens": 866646779.0,
"reward": 1.1943198680877685,
"reward_std": 0.08455176651477814,
"rewards/accuracy_reward": 0.5806640625,
"rewards/brier_reward": 0.808072280883789,
"rewards/confidence_one_or_zero": 0.00185546875,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.5697656273841858,
"step": 255
},
{
"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": 9.765625e-05,
"completions/max_length": 688.2,
"completions/max_terminated_length": 466.0,
"completions/mean_length": 219.43544921875,
"completions/mean_terminated_length": 219.30695190429688,
"completions/min_length": 112.6,
"completions/min_terminated_length": 112.6,
"epoch": 0.832,
"grad_norm": 0.010291030630469322,
"learning_rate": 1e-06,
"loss": 0.0003,
"num_tokens": 883902150.0,
"reward": 1.1886512756347656,
"reward_std": 0.08867516815662384,
"rewards/accuracy_reward": 0.552734375,
"rewards/brier_reward": 0.8246647596359253,
"rewards/confidence_one_or_zero": 0.001953125,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.5912207126617431,
"step": 260
},
{
"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": 444.2,
"completions/max_terminated_length": 444.2,
"completions/mean_length": 219.04970703125,
"completions/mean_terminated_length": 219.04970703125,
"completions/min_length": 118.6,
"completions/min_terminated_length": 118.6,
"epoch": 0.848,
"grad_norm": 0.007072034757584333,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 901159587.0,
"reward": 1.1735345602035523,
"reward_std": 0.08179984986782074,
"rewards/accuracy_reward": 0.53017578125,
"rewards/brier_reward": 0.8168921947479248,
"rewards/confidence_one_or_zero": 0.001171875,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.5902080178260803,
"step": 265
},
{
"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": 9.765625e-05,
"completions/max_length": 655.6,
"completions/max_terminated_length": 498.2,
"completions/mean_length": 225.3056640625,
"completions/mean_terminated_length": 225.17782287597657,
"completions/min_length": 124.0,
"completions/min_terminated_length": 124.0,
"epoch": 0.864,
"grad_norm": 0.009159094654023647,
"learning_rate": 1e-06,
"loss": 0.0004,
"num_tokens": 918453533.0,
"reward": 1.199086856842041,
"reward_std": 0.08750579506158829,
"rewards/accuracy_reward": 0.5861328125,
"rewards/brier_reward": 0.8121374368667602,
"rewards/confidence_one_or_zero": 0.00283203125,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.552389633655548,
"step": 270
},
{
"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": 488.2,
"completions/max_terminated_length": 488.2,
"completions/mean_length": 225.61953125,
"completions/mean_terminated_length": 225.61953125,
"completions/min_length": 117.4,
"completions/min_terminated_length": 117.4,
"epoch": 0.88,
"grad_norm": 0.013025188818573952,
"learning_rate": 1e-06,
"loss": -0.0002,
"num_tokens": 935910949.0,
"reward": 1.1634913206100463,
"reward_std": 0.08096154034137726,
"rewards/accuracy_reward": 0.508984375,
"rewards/brier_reward": 0.8179973006248474,
"rewards/confidence_one_or_zero": 0.001953125,
"rewards/format_reward": 1.0,
"rewards/mean_confidence_reward": 0.5161884605884552,
"step": 275
},
{
"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": 0.00087890625,
"completions/max_length": 1493.2,
"completions/max_terminated_length": 1388.4,
"completions/mean_length": 230.47041015625,
"completions/mean_terminated_length": 229.32456970214844,
"completions/min_length": 111.4,
"completions/min_terminated_length": 111.4,
"epoch": 0.896,
"grad_norm": 0.013893580064177513,
"learning_rate": 1e-06,
"loss": 0.002,
"num_tokens": 953381814.0,
"reward": 1.171729063987732,
"reward_std": 0.07993723750114441,
"rewards/accuracy_reward": 0.53193359375,
"rewards/brier_reward": 0.8125000953674316,
"rewards/confidence_one_or_zero": 0.00244140625,
"rewards/format_reward": 0.9990234375,
"rewards/mean_confidence_reward": 0.5306591987609863,
"step": 280
},
{
"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": 0.00185546875,
"completions/max_length": 1531.0,
"completions/max_terminated_length": 1445.0,
"completions/mean_length": 230.8126953125,
"completions/mean_terminated_length": 228.377685546875,
"completions/min_length": 82.0,
"completions/min_terminated_length": 82.0,
"epoch": 0.912,
"grad_norm": 0.036593444645404816,
"learning_rate": 1e-06,
"loss": 0.0051,
"num_tokens": 970796632.0,
"reward": 1.1708195447921752,
"reward_std": 0.09443000853061675,
"rewards/accuracy_reward": 0.529296875,
"rewards/brier_reward": 0.8148803591728211,
"rewards/confidence_one_or_zero": 0.00146484375,
"rewards/format_reward": 0.9974609375,
"rewards/mean_confidence_reward": 0.5297421932220459,
"step": 285
},
{
"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": 0.00048828125,
"completions/max_length": 1076.0,
"completions/max_terminated_length": 947.4,
"completions/mean_length": 224.44541015625,
"completions/mean_terminated_length": 223.80473937988282,
"completions/min_length": 112.2,
"completions/min_terminated_length": 112.2,
"epoch": 0.928,
"grad_norm": 0.032313406467437744,
"learning_rate": 1e-06,
"loss": 0.0014,
"num_tokens": 988121769.0,
"reward": 1.1626015663146974,
"reward_std": 0.08359554558992385,
"rewards/accuracy_reward": 0.52421875,
"rewards/brier_reward": 0.8017645239830017,
"rewards/confidence_one_or_zero": 9.765625e-05,
"rewards/format_reward": 0.99921875,
"rewards/mean_confidence_reward": 0.5579541087150574,
"step": 290
},
{
"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": 9.765625e-05,
"completions/max_length": 1049.2,
"completions/max_terminated_length": 1025.4,
"completions/mean_length": 229.90673828125,
"completions/mean_terminated_length": 229.7791961669922,
"completions/min_length": 100.6,
"completions/min_terminated_length": 100.6,
"epoch": 0.944,
"grad_norm": 0.03256411850452423,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 1005451438.0,
"reward": 1.1718948125839233,
"reward_std": 0.09756192564964294,
"rewards/accuracy_reward": 0.53125,
"rewards/brier_reward": 0.8127339124679566,
"rewards/confidence_one_or_zero": 0.00029296875,
"rewards/format_reward": 0.9998046875,
"rewards/mean_confidence_reward": 0.5793359398841857,
"step": 295
},
{
"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": 664.8,
"completions/max_terminated_length": 664.8,
"completions/mean_length": 237.31884765625,
"completions/mean_terminated_length": 237.31884765625,
"completions/min_length": 110.4,
"completions/min_terminated_length": 110.4,
"epoch": 0.96,
"grad_norm": 0.005703456234186888,
"learning_rate": 1e-06,
"loss": -0.0003,
"num_tokens": 1022821903.0,
"reward": 1.1647626161575317,
"reward_std": 0.07187836617231369,
"rewards/accuracy_reward": 0.51796875,
"rewards/brier_reward": 0.8118483901023865,
"rewards/confidence_one_or_zero": 0.00029296875,
"rewards/format_reward": 0.99970703125,
"rewards/mean_confidence_reward": 0.5718408465385437,
"step": 300
},
{
"epoch": 0.96,
"eval_completions/clipped_ratio": 0.0,
"eval_completions/max_length": 407.0,
"eval_completions/max_terminated_length": 407.0,
"eval_completions/mean_length": 245.04660415649414,
"eval_completions/mean_terminated_length": 245.04660415649414,
"eval_completions/min_length": 155.25,
"eval_completions/min_terminated_length": 155.25,
"eval_loss": 0.0,
"eval_num_tokens": 1022821903.0,
"eval_reward": 1.12132129073143,
"eval_reward_std": 0.27718404680490494,
"eval_rewards/accuracy_reward": 0.43359375,
"eval_rewards/brier_reward": 0.8090478628873825,
"eval_rewards/confidence_one_or_zero": 0.001953125,
"eval_rewards/format_reward": 1.0,
"eval_rewards/mean_confidence_reward": 0.508496105670929,
"eval_runtime": 22.3953,
"eval_samples_per_second": 22.326,
"eval_steps_per_second": 0.179,
"step": 300
},
{
"clip_ratio/high_max": 0.0,
"clip_ratio/high_mean": 0.0,
"clip_ratio/low_mean": 0.0,
"clip_ratio/low_min": 0.0,
"clip_ratio/region_mean": 0.0,
"completions/clipped_ratio": 0.0001953125,
"completions/max_length": 980.4,
"completions/max_terminated_length": 564.8,
"completions/mean_length": 244.62080078125,
"completions/mean_terminated_length": 244.36865234375,
"completions/min_length": 118.8,
"completions/min_terminated_length": 118.8,
"epoch": 0.976,
"grad_norm": 0.01138161402195692,
"learning_rate": 1e-06,
"loss": 0.0007,
"num_tokens": 1040187940.0,
"reward": 1.1814841508865357,
"reward_std": 0.07296017110347748,
"rewards/accuracy_reward": 0.5513671875,
"rewards/brier_reward": 0.8119907259941102,
"rewards/confidence_one_or_zero": 0.0001953125,
"rewards/format_reward": 0.999609375,
"rewards/mean_confidence_reward": 0.5193847775459289,
"step": 305
},
{
"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": 682.0,
"completions/max_terminated_length": 682.0,
"completions/mean_length": 238.13291015625,
"completions/mean_terminated_length": 238.13291015625,
"completions/min_length": 127.2,
"completions/min_terminated_length": 127.2,
"epoch": 0.992,
"grad_norm": 0.006512052845209837,
"learning_rate": 1e-06,
"loss": 0.0001,
"num_tokens": 1057754901.0,
"reward": 1.1691525459289551,
"reward_std": 0.06929974779486656,
"rewards/accuracy_reward": 0.5232421875,
"rewards/brier_reward": 0.8151596069335938,
"rewards/confidence_one_or_zero": 0.00087890625,
"rewards/format_reward": 0.99990234375,
"rewards/mean_confidence_reward": 0.4931289255619049,
"step": 310
},
{
"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": 559.5,
"completions/max_terminated_length": 559.5,
"completions/mean_length": 235.83663940429688,
"completions/mean_terminated_length": 235.83663940429688,
"completions/min_length": 122.0,
"completions/min_terminated_length": 122.0,
"epoch": 0.9984,
"num_tokens": 1064720670.0,
"reward": 1.1648695468902588,
"reward_std": 0.073847196996212,
"rewards/accuracy_reward": 0.54248046875,
"rewards/brier_reward": 0.787746012210846,
"rewards/confidence_one_or_zero": 0.000244140625,
"rewards/format_reward": 0.99951171875,
"rewards/mean_confidence_reward": 0.5165649354457855,
"step": 312,
"total_flos": 0.0,
"train_loss": 0.005198169167087121,
"train_runtime": 92786.4477,
"train_samples_per_second": 0.216,
"train_steps_per_second": 0.003
}
],
"logging_steps": 5,
"max_steps": 312,
"num_input_tokens_seen": 1064720670,
"num_train_epochs": 1,
"save_steps": 60,
"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": 8,
"trial_name": null,
"trial_params": null
}