824 lines
34 KiB
JSON
824 lines
34 KiB
JSON
|
{
|
||
|
|
"best_global_step": null,
|
||
|
|
"best_metric": null,
|
||
|
|
"best_model_checkpoint": null,
|
||
|
|
"epoch": 19.666666666666668,
|
||
|
|
"eval_steps": 500,
|
||
|
|
"global_step": 20,
|
||
|
|
"is_hyper_param_search": false,
|
||
|
|
"is_local_process_zero": true,
|
||
|
|
"is_world_process_zero": true,
|
||
|
|
"log_history": [
|
||
|
|
{
|
||
|
|
"clip_ratio/high_max": 0.0,
|
||
|
|
"clip_ratio/high_mean": 0.0,
|
||
|
|
"clip_ratio/low_mean": 0.0,
|
||
|
|
"clip_ratio/low_min": 0.0,
|
||
|
|
"clip_ratio/region_mean": 0.0,
|
||
|
|
"completions/clipped_ratio": 0.0,
|
||
|
|
"completions/max_length": 388.0,
|
||
|
|
"completions/max_terminated_length": 388.0,
|
||
|
|
"completions/mean_length": 248.0,
|
||
|
|
"completions/mean_terminated_length": 248.0,
|
||
|
|
"completions/min_length": 105.0,
|
||
|
|
"completions/min_terminated_length": 105.0,
|
||
|
|
"epoch": 0.6666666666666666,
|
||
|
|
"grad_norm": 7.726290702819824,
|
||
|
|
"kl": 0.0,
|
||
|
|
"learning_rate": 0.0,
|
||
|
|
"loss": -0.1465,
|
||
|
|
"num_tokens": 1704.0,
|
||
|
|
"reward": 0.01887180097401142,
|
||
|
|
"reward_std": 0.0037353751249611378,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.01887180097401142,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.006438133772462606,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 1
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 270.0,
|
||
|
|
"completions/mean_length": 316.6666564941406,
|
||
|
|
"completions/mean_terminated_length": 251.0,
|
||
|
|
"completions/min_length": 232.0,
|
||
|
|
"completions/min_terminated_length": 232.0,
|
||
|
|
"epoch": 1.6666666666666665,
|
||
|
|
"grad_norm": 7.617258071899414,
|
||
|
|
"kl": 0.0,
|
||
|
|
"learning_rate": 1e-06,
|
||
|
|
"loss": -0.1153,
|
||
|
|
"num_tokens": 3229.0,
|
||
|
|
"reward": 0.043164871633052826,
|
||
|
|
"reward_std": 0.025065843015909195,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.022164873778820038,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.008109219372272491,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.020999999716877937,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.04200000315904617,
|
||
|
|
"step": 2
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 380.0,
|
||
|
|
"completions/max_terminated_length": 380.0,
|
||
|
|
"completions/mean_length": 268.3333435058594,
|
||
|
|
"completions/mean_terminated_length": 268.3333435058594,
|
||
|
|
"completions/min_length": 101.0,
|
||
|
|
"completions/min_terminated_length": 101.0,
|
||
|
|
"epoch": 2.6666666666666665,
|
||
|
|
"grad_norm": 8.109015464782715,
|
||
|
|
"kl": 0.0020800233469344676,
|
||
|
|
"learning_rate": 9.931806517013612e-07,
|
||
|
|
"loss": 0.177,
|
||
|
|
"num_tokens": 4821.0,
|
||
|
|
"reward": -0.3887316882610321,
|
||
|
|
"reward_std": 0.5801461935043335,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.018268316984176636,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.004313780926167965,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": -0.40700000524520874,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.8139999508857727,
|
||
|
|
"step": 3
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.6666666666666667,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 158.0,
|
||
|
|
"completions/mean_length": 351.3333435058594,
|
||
|
|
"completions/mean_terminated_length": 158.0,
|
||
|
|
"completions/min_length": 158.0,
|
||
|
|
"completions/min_terminated_length": 158.0,
|
||
|
|
"epoch": 3.6666666666666665,
|
||
|
|
"grad_norm": 8.73157024383545,
|
||
|
|
"kl": 0.004346272035036236,
|
||
|
|
"learning_rate": 9.729086208503173e-07,
|
||
|
|
"loss": 0.0175,
|
||
|
|
"num_tokens": 6530.0,
|
||
|
|
"reward": 0.016507050022482872,
|
||
|
|
"reward_std": 0.004461287520825863,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.01650705188512802,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.010719080455601215,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 4
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 434.0,
|
||
|
|
"completions/max_terminated_length": 434.0,
|
||
|
|
"completions/mean_length": 244.6666717529297,
|
||
|
|
"completions/mean_terminated_length": 244.6666717529297,
|
||
|
|
"completions/min_length": 97.0,
|
||
|
|
"completions/min_terminated_length": 97.0,
|
||
|
|
"epoch": 4.666666666666667,
|
||
|
|
"grad_norm": 8.055858612060547,
|
||
|
|
"kl": 0.00888643367215991,
|
||
|
|
"learning_rate": 9.397368756032444e-07,
|
||
|
|
"loss": 0.3139,
|
||
|
|
"num_tokens": 8122.0,
|
||
|
|
"reward": 0.028256595134735107,
|
||
|
|
"reward_std": 0.010700763203203678,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.028256593272089958,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.020540650933980942,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 249.0,
|
||
|
|
"completions/mean_length": 264.0,
|
||
|
|
"completions/mean_terminated_length": 172.0,
|
||
|
|
"completions/min_length": 95.0,
|
||
|
|
"completions/min_terminated_length": 95.0,
|
||
|
|
"epoch": 5.666666666666667,
|
||
|
|
"grad_norm": 8.429837226867676,
|
||
|
|
"kl": 0.014231715817004442,
|
||
|
|
"learning_rate": 8.945702546981968e-07,
|
||
|
|
"loss": 0.3539,
|
||
|
|
"num_tokens": 9543.0,
|
||
|
|
"reward": 0.03853300213813782,
|
||
|
|
"reward_std": 0.03958161920309067,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.03853300213813782,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.05091189593076706,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 6
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 371.0,
|
||
|
|
"completions/mean_length": 392.6666564941406,
|
||
|
|
"completions/mean_terminated_length": 365.0,
|
||
|
|
"completions/min_length": 359.0,
|
||
|
|
"completions/min_terminated_length": 359.0,
|
||
|
|
"epoch": 6.666666666666667,
|
||
|
|
"grad_norm": 6.220398426055908,
|
||
|
|
"kl": 0.021292249206453562,
|
||
|
|
"learning_rate": 8.386407858128706e-07,
|
||
|
|
"loss": 0.1064,
|
||
|
|
"num_tokens": 11475.0,
|
||
|
|
"reward": 0.010966416448354721,
|
||
|
|
"reward_std": 0.003916876390576363,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.010966415517032146,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.005275326315313578,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 7
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 413.0,
|
||
|
|
"completions/max_terminated_length": 413.0,
|
||
|
|
"completions/mean_length": 194.0,
|
||
|
|
"completions/mean_terminated_length": 194.0,
|
||
|
|
"completions/min_length": 83.0,
|
||
|
|
"completions/min_terminated_length": 83.0,
|
||
|
|
"epoch": 7.666666666666667,
|
||
|
|
"grad_norm": 6.589484214782715,
|
||
|
|
"kl": 0.013817875646054745,
|
||
|
|
"learning_rate": 7.734740790612136e-07,
|
||
|
|
"loss": 0.4696,
|
||
|
|
"num_tokens": 13017.0,
|
||
|
|
"reward": 0.02751215174794197,
|
||
|
|
"reward_std": 0.012231169268488884,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.02751215174794197,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.011579956859350204,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 8
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 399.0,
|
||
|
|
"completions/max_terminated_length": 399.0,
|
||
|
|
"completions/mean_length": 278.3333435058594,
|
||
|
|
"completions/mean_terminated_length": 278.3333435058594,
|
||
|
|
"completions/min_length": 100.0,
|
||
|
|
"completions/min_terminated_length": 100.0,
|
||
|
|
"epoch": 8.666666666666666,
|
||
|
|
"grad_norm": 10.296525001525879,
|
||
|
|
"kl": 0.029440555721521378,
|
||
|
|
"learning_rate": 7.008477123264847e-07,
|
||
|
|
"loss": 0.089,
|
||
|
|
"num_tokens": 14434.0,
|
||
|
|
"reward": 0.0298730731010437,
|
||
|
|
"reward_std": 0.0038347183726727962,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.0298730731010437,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.007508063223212957,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 9
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 233.0,
|
||
|
|
"completions/mean_length": 245.0,
|
||
|
|
"completions/mean_terminated_length": 143.5,
|
||
|
|
"completions/min_length": 54.0,
|
||
|
|
"completions/min_terminated_length": 54.0,
|
||
|
|
"epoch": 9.666666666666666,
|
||
|
|
"grad_norm": 10.521705627441406,
|
||
|
|
"kl": 0.033842616714537144,
|
||
|
|
"learning_rate": 6.227427435703995e-07,
|
||
|
|
"loss": 0.5265,
|
||
|
|
"num_tokens": 15713.0,
|
||
|
|
"reward": -0.028218472376465797,
|
||
|
|
"reward_std": 0.09047404676675797,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.028281521052122116,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.01116233877837658,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": -0.05649999901652336,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.11299999803304672,
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 314.0,
|
||
|
|
"completions/mean_length": 298.6666564941406,
|
||
|
|
"completions/mean_terminated_length": 224.0,
|
||
|
|
"completions/min_length": 134.0,
|
||
|
|
"completions/min_terminated_length": 134.0,
|
||
|
|
"epoch": 10.666666666666666,
|
||
|
|
"grad_norm": 6.642791748046875,
|
||
|
|
"kl": 0.025248159654438496,
|
||
|
|
"learning_rate": 5.412896727361662e-07,
|
||
|
|
"loss": 0.0622,
|
||
|
|
"num_tokens": 17296.0,
|
||
|
|
"reward": 0.024774959310889244,
|
||
|
|
"reward_std": 0.004647332243621349,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.019774960353970528,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.0032825965899974108,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.004999999888241291,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.009999999776482582,
|
||
|
|
"step": 11
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 324.0,
|
||
|
|
"completions/max_terminated_length": 324.0,
|
||
|
|
"completions/mean_length": 230.3333282470703,
|
||
|
|
"completions/mean_terminated_length": 230.3333282470703,
|
||
|
|
"completions/min_length": 174.0,
|
||
|
|
"completions/min_terminated_length": 174.0,
|
||
|
|
"epoch": 11.666666666666666,
|
||
|
|
"grad_norm": 7.2714314460754395,
|
||
|
|
"kl": 0.04418220743536949,
|
||
|
|
"learning_rate": 4.5871032726383385e-07,
|
||
|
|
"loss": 0.1436,
|
||
|
|
"num_tokens": 18736.0,
|
||
|
|
"reward": 0.026946237310767174,
|
||
|
|
"reward_std": 0.024818968027830124,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.026946237310767174,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.026517190039157867,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 12
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 360.0,
|
||
|
|
"completions/mean_length": 336.0,
|
||
|
|
"completions/mean_terminated_length": 280.0,
|
||
|
|
"completions/min_length": 200.0,
|
||
|
|
"completions/min_terminated_length": 200.0,
|
||
|
|
"epoch": 12.666666666666666,
|
||
|
|
"grad_norm": 8.894806861877441,
|
||
|
|
"kl": 0.04840276390314102,
|
||
|
|
"learning_rate": 3.772572564296004e-07,
|
||
|
|
"loss": 0.0025,
|
||
|
|
"num_tokens": 20504.0,
|
||
|
|
"reward": 0.027708034962415695,
|
||
|
|
"reward_std": 0.015193699859082699,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.01970803365111351,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.009420881047844887,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.00800000037997961,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.01600000075995922,
|
||
|
|
"step": 13
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 238.0,
|
||
|
|
"completions/max_terminated_length": 238.0,
|
||
|
|
"completions/mean_length": 129.6666717529297,
|
||
|
|
"completions/mean_terminated_length": 129.6666717529297,
|
||
|
|
"completions/min_length": 26.0,
|
||
|
|
"completions/min_terminated_length": 26.0,
|
||
|
|
"epoch": 13.666666666666666,
|
||
|
|
"grad_norm": 13.065749168395996,
|
||
|
|
"kl": 0.060946088284254074,
|
||
|
|
"learning_rate": 2.9915228767351535e-07,
|
||
|
|
"loss": 0.228,
|
||
|
|
"num_tokens": 21494.0,
|
||
|
|
"reward": 0.036238811910152435,
|
||
|
|
"reward_std": 0.0063911802135407925,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.036238811910152435,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.007549918722361326,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 14
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 387.0,
|
||
|
|
"completions/mean_length": 306.3333435058594,
|
||
|
|
"completions/mean_terminated_length": 235.5,
|
||
|
|
"completions/min_length": 84.0,
|
||
|
|
"completions/min_terminated_length": 84.0,
|
||
|
|
"epoch": 14.666666666666666,
|
||
|
|
"grad_norm": 5.861216068267822,
|
||
|
|
"kl": 0.028976112604141235,
|
||
|
|
"learning_rate": 2.2652592093878665e-07,
|
||
|
|
"loss": 0.264,
|
||
|
|
"num_tokens": 23257.0,
|
||
|
|
"reward": 0.016261670738458633,
|
||
|
|
"reward_std": 0.00623771408572793,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.016261670738458633,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.00987034197896719,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"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": 0.0,
|
||
|
|
"completions/max_length": 177.0,
|
||
|
|
"completions/max_terminated_length": 177.0,
|
||
|
|
"completions/mean_length": 130.3333282470703,
|
||
|
|
"completions/mean_terminated_length": 130.3333282470703,
|
||
|
|
"completions/min_length": 76.0,
|
||
|
|
"completions/min_terminated_length": 76.0,
|
||
|
|
"epoch": 15.666666666666666,
|
||
|
|
"grad_norm": 10.36058521270752,
|
||
|
|
"kl": 0.03464523144066334,
|
||
|
|
"learning_rate": 1.6135921418712955e-07,
|
||
|
|
"loss": 0.2436,
|
||
|
|
"num_tokens": 24434.0,
|
||
|
|
"reward": 0.044337280094623566,
|
||
|
|
"reward_std": 0.023607168346643448,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.044337283819913864,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.026401923969388008,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 16
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 231.0,
|
||
|
|
"completions/mean_length": 234.6666717529297,
|
||
|
|
"completions/mean_terminated_length": 128.0,
|
||
|
|
"completions/min_length": 25.0,
|
||
|
|
"completions/min_terminated_length": 25.0,
|
||
|
|
"epoch": 16.666666666666668,
|
||
|
|
"grad_norm": 7.803691864013672,
|
||
|
|
"kl": 0.04637394845485687,
|
||
|
|
"learning_rate": 1.0542974530180327e-07,
|
||
|
|
"loss": 0.3776,
|
||
|
|
"num_tokens": 25979.0,
|
||
|
|
"reward": 0.023612000048160553,
|
||
|
|
"reward_std": 0.014000779017806053,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.023612000048160553,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.012624057941138744,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 17
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 420.0,
|
||
|
|
"completions/max_terminated_length": 420.0,
|
||
|
|
"completions/mean_length": 278.3333435058594,
|
||
|
|
"completions/mean_terminated_length": 278.3333435058594,
|
||
|
|
"completions/min_length": 72.0,
|
||
|
|
"completions/min_terminated_length": 72.0,
|
||
|
|
"epoch": 17.666666666666668,
|
||
|
|
"grad_norm": 10.446023941040039,
|
||
|
|
"kl": 0.04021918121725321,
|
||
|
|
"learning_rate": 6.026312439675551e-08,
|
||
|
|
"loss": 0.119,
|
||
|
|
"num_tokens": 27508.0,
|
||
|
|
"reward": 0.014419719576835632,
|
||
|
|
"reward_std": 0.006932843942195177,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.014419719576835632,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.009299629367887974,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 18
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 0.33333333333333337,
|
||
|
|
"completions/max_length": 448.0,
|
||
|
|
"completions/max_terminated_length": 156.0,
|
||
|
|
"completions/mean_length": 238.0,
|
||
|
|
"completions/mean_terminated_length": 133.0,
|
||
|
|
"completions/min_length": 110.0,
|
||
|
|
"completions/min_terminated_length": 110.0,
|
||
|
|
"epoch": 18.666666666666668,
|
||
|
|
"grad_norm": 9.053911209106445,
|
||
|
|
"kl": 0.08691145107150078,
|
||
|
|
"learning_rate": 2.7091379149682682e-08,
|
||
|
|
"loss": 0.2805,
|
||
|
|
"num_tokens": 28962.0,
|
||
|
|
"reward": 0.020562436431646347,
|
||
|
|
"reward_std": 0.007597055751830339,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.020562436431646347,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.00773102231323719,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
||
|
|
"step": 19
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"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": 352.0,
|
||
|
|
"completions/max_terminated_length": 352.0,
|
||
|
|
"completions/mean_length": 260.0,
|
||
|
|
"completions/mean_terminated_length": 260.0,
|
||
|
|
"completions/min_length": 167.0,
|
||
|
|
"completions/min_terminated_length": 167.0,
|
||
|
|
"epoch": 19.666666666666668,
|
||
|
|
"grad_norm": 7.216071128845215,
|
||
|
|
"kl": 0.030871054157614708,
|
||
|
|
"learning_rate": 6.819348298638839e-09,
|
||
|
|
"loss": 0.0929,
|
||
|
|
"num_tokens": 30476.0,
|
||
|
|
"reward": -0.004711912013590336,
|
||
|
|
"reward_std": 0.025457601994276047,
|
||
|
|
"rewards/concensus_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/concensus_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/consensus_reward_func/mean": 0.0,
|
||
|
|
"rewards/consensus_reward_func/std": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/mean": 0.0,
|
||
|
|
"rewards/cumulative_reward_2/std": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/mean": 0.0,
|
||
|
|
"rewards/final_correctness_reward_func/std": 0.0,
|
||
|
|
"rewards/question_recreation_reward_func/mean": 0.013288087211549282,
|
||
|
|
"rewards/question_recreation_reward_func/std": 0.00169478973839432,
|
||
|
|
"rewards/soft_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/soft_format_reward_func/std": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/mean": 0.0,
|
||
|
|
"rewards/strict_format_reward_func/std": 0.0,
|
||
|
|
"rewards/xmlcount_reward_func/mean": -0.017999999225139618,
|
||
|
|
"rewards/xmlcount_reward_func/std": 0.035999998450279236,
|
||
|
|
"step": 20
|
||
|
|
},
|
||
|
|
{
|
||
|
|
"epoch": 19.666666666666668,
|
||
|
|
"step": 20,
|
||
|
|
"total_flos": 0.0,
|
||
|
|
"train_loss": 0.1802978585474193,
|
||
|
|
"train_runtime": 16451.9181,
|
||
|
|
"train_samples_per_second": 0.005,
|
||
|
|
"train_steps_per_second": 0.001
|
||
|
|
}
|
||
|
|
],
|
||
|
|
"logging_steps": 1,
|
||
|
|
"max_steps": 20,
|
||
|
|
"num_input_tokens_seen": 30476,
|
||
|
|
"num_train_epochs": 20,
|
||
|
|
"save_steps": 25,
|
||
|
|
"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": 2,
|
||
|
|
"trial_name": null,
|
||
|
|
"trial_params": null
|
||
|
|
}
|