86dcda4584
Model: seeib/Qwen2.5-0.5B-Instruct-Gensyn-Swarm-prehistoric_gregarious_seahorse Source: Original Platform
824 lines
34 KiB
JSON
824 lines
34 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 2.2222222222222223,
|
|
"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": 63.0,
|
|
"completions/max_terminated_length": 63.0,
|
|
"completions/mean_length": 40.75,
|
|
"completions/mean_terminated_length": 40.75,
|
|
"completions/min_length": 31.0,
|
|
"completions/min_terminated_length": 31.0,
|
|
"epoch": 0.1111111111111111,
|
|
"grad_norm": 25.69437599182129,
|
|
"kl": 0.0,
|
|
"learning_rate": 0.0,
|
|
"loss": -0.1171,
|
|
"num_tokens": 675.0,
|
|
"reward": 0.5159521698951721,
|
|
"reward_std": 0.0688270628452301,
|
|
"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": 0.07845214009284973,
|
|
"rewards/question_recreation_reward_func/std": 0.04025953263044357,
|
|
"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": 0.4375,
|
|
"rewards/xmlcount_reward_func/std": 0.125,
|
|
"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": 0.0,
|
|
"completions/max_length": 104.0,
|
|
"completions/max_terminated_length": 104.0,
|
|
"completions/mean_length": 56.5,
|
|
"completions/mean_terminated_length": 56.5,
|
|
"completions/min_length": 24.0,
|
|
"completions/min_terminated_length": 24.0,
|
|
"epoch": 0.2222222222222222,
|
|
"grad_norm": 22.427824020385742,
|
|
"kl": 0.0,
|
|
"learning_rate": 1e-06,
|
|
"loss": 0.0564,
|
|
"num_tokens": 1413.0,
|
|
"reward": 0.5325700640678406,
|
|
"reward_std": 0.0006790067418478429,
|
|
"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": 0.032570019364356995,
|
|
"rewards/question_recreation_reward_func/std": 0.0007998892688192427,
|
|
"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": 0.5,
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
|
"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": 231.0,
|
|
"completions/max_terminated_length": 231.0,
|
|
"completions/mean_length": 129.5,
|
|
"completions/mean_terminated_length": 129.5,
|
|
"completions/min_length": 59.0,
|
|
"completions/min_terminated_length": 59.0,
|
|
"epoch": 0.3333333333333333,
|
|
"grad_norm": 18.40679168701172,
|
|
"kl": 0.002804491203278303,
|
|
"learning_rate": 9.931806517013612e-07,
|
|
"loss": 0.297,
|
|
"num_tokens": 2443.0,
|
|
"reward": 0.6747503876686096,
|
|
"reward_std": 0.1533203423023224,
|
|
"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": 0.20650038123130798,
|
|
"rewards/question_recreation_reward_func/std": 0.019515778869390488,
|
|
"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": 0.46825000643730164,
|
|
"rewards/xmlcount_reward_func/std": 0.21338286995887756,
|
|
"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": 0.0,
|
|
"completions/max_length": 110.0,
|
|
"completions/max_terminated_length": 110.0,
|
|
"completions/mean_length": 62.25,
|
|
"completions/mean_terminated_length": 62.25,
|
|
"completions/min_length": 23.0,
|
|
"completions/min_terminated_length": 23.0,
|
|
"epoch": 0.4444444444444444,
|
|
"grad_norm": 22.179134368896484,
|
|
"kl": 0.012003686744719744,
|
|
"learning_rate": 9.729086208503173e-07,
|
|
"loss": -0.0112,
|
|
"num_tokens": 3204.0,
|
|
"reward": 0.6682786345481873,
|
|
"reward_std": 0.016284530982375145,
|
|
"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": 0.04327860474586487,
|
|
"rewards/question_recreation_reward_func/std": 0.0372861810028553,
|
|
"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": 0.625,
|
|
"rewards/xmlcount_reward_func/std": 0.14433756470680237,
|
|
"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": 79.0,
|
|
"completions/max_terminated_length": 79.0,
|
|
"completions/mean_length": 54.75,
|
|
"completions/mean_terminated_length": 54.75,
|
|
"completions/min_length": 26.0,
|
|
"completions/min_terminated_length": 26.0,
|
|
"epoch": 0.5555555555555556,
|
|
"grad_norm": 19.958255767822266,
|
|
"kl": 0.011007877299562097,
|
|
"learning_rate": 9.397368756032444e-07,
|
|
"loss": -0.1713,
|
|
"num_tokens": 3935.0,
|
|
"reward": 0.8996135592460632,
|
|
"reward_std": 0.049514204263687134,
|
|
"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": 0.27461355924606323,
|
|
"rewards/question_recreation_reward_func/std": 0.044617120176553726,
|
|
"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": 0.625,
|
|
"rewards/xmlcount_reward_func/std": 0.14433756470680237,
|
|
"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": 0.0,
|
|
"completions/max_length": 29.0,
|
|
"completions/max_terminated_length": 29.0,
|
|
"completions/mean_length": 27.25,
|
|
"completions/mean_terminated_length": 27.25,
|
|
"completions/min_length": 24.0,
|
|
"completions/min_terminated_length": 24.0,
|
|
"epoch": 0.6666666666666666,
|
|
"grad_norm": 23.830913543701172,
|
|
"kl": 0.04720709100365639,
|
|
"learning_rate": 8.945702546981968e-07,
|
|
"loss": -0.0311,
|
|
"num_tokens": 4556.0,
|
|
"reward": 0.5360786318778992,
|
|
"reward_std": 0.008525591343641281,
|
|
"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": 0.03607860952615738,
|
|
"rewards/question_recreation_reward_func/std": 0.02430449239909649,
|
|
"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": 0.5,
|
|
"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": 0.0,
|
|
"completions/max_length": 56.0,
|
|
"completions/max_terminated_length": 56.0,
|
|
"completions/mean_length": 40.5,
|
|
"completions/mean_terminated_length": 40.5,
|
|
"completions/min_length": 30.0,
|
|
"completions/min_terminated_length": 30.0,
|
|
"epoch": 0.7777777777777778,
|
|
"grad_norm": 39.75885772705078,
|
|
"kl": 0.056546732783317566,
|
|
"learning_rate": 8.386407858128706e-07,
|
|
"loss": -0.07,
|
|
"num_tokens": 5230.0,
|
|
"reward": 0.6005383729934692,
|
|
"reward_std": 0.08559251576662064,
|
|
"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": 0.03853834420442581,
|
|
"rewards/question_recreation_reward_func/std": 0.007710414472967386,
|
|
"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": 0.562000036239624,
|
|
"rewards/xmlcount_reward_func/std": 0.12400001287460327,
|
|
"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": 95.0,
|
|
"completions/max_terminated_length": 95.0,
|
|
"completions/mean_length": 61.25,
|
|
"completions/mean_terminated_length": 61.25,
|
|
"completions/min_length": 30.0,
|
|
"completions/min_terminated_length": 30.0,
|
|
"epoch": 0.8888888888888888,
|
|
"grad_norm": 25.805145263671875,
|
|
"kl": 0.07301068678498268,
|
|
"learning_rate": 7.734740790612136e-07,
|
|
"loss": -0.3117,
|
|
"num_tokens": 5987.0,
|
|
"reward": 0.7229201197624207,
|
|
"reward_std": 0.0429324246942997,
|
|
"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": 0.22292014956474304,
|
|
"rewards/question_recreation_reward_func/std": 0.05922666937112808,
|
|
"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": 0.5,
|
|
"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": 47.0,
|
|
"completions/max_terminated_length": 47.0,
|
|
"completions/mean_length": 35.5,
|
|
"completions/mean_terminated_length": 35.5,
|
|
"completions/min_length": 24.0,
|
|
"completions/min_terminated_length": 24.0,
|
|
"epoch": 1.0,
|
|
"grad_norm": 30.85164451599121,
|
|
"kl": 0.02306750975549221,
|
|
"learning_rate": 7.008477123264847e-07,
|
|
"loss": 0.0617,
|
|
"num_tokens": 6634.0,
|
|
"reward": 0.5571061372756958,
|
|
"reward_std": 0.00527472048997879,
|
|
"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": 0.0571061372756958,
|
|
"rewards/question_recreation_reward_func/std": 0.011652503162622452,
|
|
"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": 0.5,
|
|
"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": 0.0,
|
|
"completions/max_length": 99.0,
|
|
"completions/max_terminated_length": 99.0,
|
|
"completions/mean_length": 59.5,
|
|
"completions/mean_terminated_length": 59.5,
|
|
"completions/min_length": 31.0,
|
|
"completions/min_terminated_length": 31.0,
|
|
"epoch": 1.1111111111111112,
|
|
"grad_norm": 18.589075088500977,
|
|
"kl": 0.057917265221476555,
|
|
"learning_rate": 6.227427435703995e-07,
|
|
"loss": -0.2448,
|
|
"num_tokens": 7384.0,
|
|
"reward": 0.7360702753067017,
|
|
"reward_std": 0.04403897002339363,
|
|
"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": 0.23607021570205688,
|
|
"rewards/question_recreation_reward_func/std": 0.07128877192735672,
|
|
"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": 0.5,
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
|
"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": 0.0,
|
|
"completions/max_length": 31.0,
|
|
"completions/max_terminated_length": 31.0,
|
|
"completions/mean_length": 28.0,
|
|
"completions/mean_terminated_length": 28.0,
|
|
"completions/min_length": 25.0,
|
|
"completions/min_terminated_length": 25.0,
|
|
"epoch": 1.2222222222222223,
|
|
"grad_norm": 34.077091217041016,
|
|
"kl": 0.13308210298419,
|
|
"learning_rate": 5.412896727361662e-07,
|
|
"loss": 0.0104,
|
|
"num_tokens": 8008.0,
|
|
"reward": 0.53204345703125,
|
|
"reward_std": 0.0063663022592663765,
|
|
"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": 0.032043468207120895,
|
|
"rewards/question_recreation_reward_func/std": 0.016084665432572365,
|
|
"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": 0.5,
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
|
"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": 69.0,
|
|
"completions/max_terminated_length": 69.0,
|
|
"completions/mean_length": 39.5,
|
|
"completions/mean_terminated_length": 39.5,
|
|
"completions/min_length": 22.0,
|
|
"completions/min_terminated_length": 22.0,
|
|
"epoch": 1.3333333333333333,
|
|
"grad_norm": 19.56139373779297,
|
|
"kl": 0.09075122326612473,
|
|
"learning_rate": 4.5871032726383385e-07,
|
|
"loss": -0.2792,
|
|
"num_tokens": 8678.0,
|
|
"reward": 0.5326830744743347,
|
|
"reward_std": 0.011520188301801682,
|
|
"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": 0.032683055847883224,
|
|
"rewards/question_recreation_reward_func/std": 0.009901088662445545,
|
|
"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": 0.5,
|
|
"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": 0.0,
|
|
"completions/max_length": 89.0,
|
|
"completions/max_terminated_length": 89.0,
|
|
"completions/mean_length": 79.5,
|
|
"completions/mean_terminated_length": 79.5,
|
|
"completions/min_length": 70.0,
|
|
"completions/min_terminated_length": 70.0,
|
|
"epoch": 1.4444444444444444,
|
|
"grad_norm": 26.145700454711914,
|
|
"kl": 0.027578551787883043,
|
|
"learning_rate": 3.772572564296004e-07,
|
|
"loss": -0.0213,
|
|
"num_tokens": 9508.0,
|
|
"reward": 0.8646507263183594,
|
|
"reward_std": 0.08279794454574585,
|
|
"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": 0.1771506667137146,
|
|
"rewards/question_recreation_reward_func/std": 0.056235793977975845,
|
|
"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": 0.6875,
|
|
"rewards/xmlcount_reward_func/std": 0.125,
|
|
"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": 127.0,
|
|
"completions/max_terminated_length": 127.0,
|
|
"completions/mean_length": 92.75,
|
|
"completions/mean_terminated_length": 92.75,
|
|
"completions/min_length": 68.0,
|
|
"completions/min_terminated_length": 68.0,
|
|
"epoch": 1.5555555555555556,
|
|
"grad_norm": 19.497051239013672,
|
|
"kl": 0.03215572563931346,
|
|
"learning_rate": 2.9915228767351535e-07,
|
|
"loss": -0.084,
|
|
"num_tokens": 10391.0,
|
|
"reward": 0.6083483099937439,
|
|
"reward_std": 0.10050921142101288,
|
|
"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": 0.04584832862019539,
|
|
"rewards/question_recreation_reward_func/std": 0.010448978282511234,
|
|
"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": 0.5625,
|
|
"rewards/xmlcount_reward_func/std": 0.125,
|
|
"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": 0.25,
|
|
"completions/max_length": 448.0,
|
|
"completions/max_terminated_length": 47.0,
|
|
"completions/mean_length": 138.75,
|
|
"completions/mean_terminated_length": 35.66666793823242,
|
|
"completions/min_length": 30.0,
|
|
"completions/min_terminated_length": 30.0,
|
|
"epoch": 1.6666666666666665,
|
|
"grad_norm": 14.958171844482422,
|
|
"kl": 0.036621191538870335,
|
|
"learning_rate": 2.2652592093878665e-07,
|
|
"loss": 0.2878,
|
|
"num_tokens": 11458.0,
|
|
"reward": 0.45188108086586,
|
|
"reward_std": 0.15637479722499847,
|
|
"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": 0.04563111066818237,
|
|
"rewards/question_recreation_reward_func/std": 0.020204132422804832,
|
|
"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": 0.40625,
|
|
"rewards/xmlcount_reward_func/std": 0.1875,
|
|
"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": 55.0,
|
|
"completions/max_terminated_length": 55.0,
|
|
"completions/mean_length": 37.25,
|
|
"completions/mean_terminated_length": 37.25,
|
|
"completions/min_length": 29.0,
|
|
"completions/min_terminated_length": 29.0,
|
|
"epoch": 1.7777777777777777,
|
|
"grad_norm": 27.896305084228516,
|
|
"kl": 0.17185895144939423,
|
|
"learning_rate": 1.6135921418712955e-07,
|
|
"loss": -0.064,
|
|
"num_tokens": 12119.0,
|
|
"reward": 0.4707985520362854,
|
|
"reward_std": 0.27111726999282837,
|
|
"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": 0.03329858183860779,
|
|
"rewards/question_recreation_reward_func/std": 0.012222301214933395,
|
|
"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": 0.4375,
|
|
"rewards/xmlcount_reward_func/std": 0.23935678601264954,
|
|
"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": 0.0,
|
|
"completions/max_length": 96.0,
|
|
"completions/max_terminated_length": 96.0,
|
|
"completions/mean_length": 68.5,
|
|
"completions/mean_terminated_length": 68.5,
|
|
"completions/min_length": 27.0,
|
|
"completions/min_terminated_length": 27.0,
|
|
"epoch": 1.8888888888888888,
|
|
"grad_norm": 17.811830520629883,
|
|
"kl": 0.056419532746076584,
|
|
"learning_rate": 1.0542974530180327e-07,
|
|
"loss": -0.2264,
|
|
"num_tokens": 12905.0,
|
|
"reward": 1.2158468961715698,
|
|
"reward_std": 0.8011028170585632,
|
|
"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": 0.5,
|
|
"rewards/final_correctness_reward_func/std": 1.0,
|
|
"rewards/question_recreation_reward_func/mean": 0.02884695678949356,
|
|
"rewards/question_recreation_reward_func/std": 0.008567274548113346,
|
|
"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": 0.687000036239624,
|
|
"rewards/xmlcount_reward_func/std": 0.12467022985219955,
|
|
"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": 59.0,
|
|
"completions/max_terminated_length": 59.0,
|
|
"completions/mean_length": 56.5,
|
|
"completions/mean_terminated_length": 56.5,
|
|
"completions/min_length": 54.0,
|
|
"completions/min_terminated_length": 54.0,
|
|
"epoch": 2.0,
|
|
"grad_norm": 24.79212188720703,
|
|
"kl": 0.0544373020529747,
|
|
"learning_rate": 6.026312439675551e-08,
|
|
"loss": 0.0433,
|
|
"num_tokens": 13653.0,
|
|
"reward": 0.6995986700057983,
|
|
"reward_std": 0.012008922174572945,
|
|
"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": 0.07459867000579834,
|
|
"rewards/question_recreation_reward_func/std": 0.05076611042022705,
|
|
"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": 0.625,
|
|
"rewards/xmlcount_reward_func/std": 0.14433756470680237,
|
|
"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": 0.0,
|
|
"completions/max_length": 79.0,
|
|
"completions/max_terminated_length": 79.0,
|
|
"completions/mean_length": 65.5,
|
|
"completions/mean_terminated_length": 65.5,
|
|
"completions/min_length": 55.0,
|
|
"completions/min_terminated_length": 55.0,
|
|
"epoch": 2.111111111111111,
|
|
"grad_norm": 23.983592987060547,
|
|
"kl": 0.030249727424234152,
|
|
"learning_rate": 2.7091379149682682e-08,
|
|
"loss": 0.07,
|
|
"num_tokens": 14427.0,
|
|
"reward": 0.9142850637435913,
|
|
"reward_std": 0.027192028239369392,
|
|
"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": 0.2892850637435913,
|
|
"rewards/question_recreation_reward_func/std": 0.027173683047294617,
|
|
"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": 0.625,
|
|
"rewards/xmlcount_reward_func/std": 0.14433756470680237,
|
|
"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": 85.0,
|
|
"completions/max_terminated_length": 85.0,
|
|
"completions/mean_length": 56.25,
|
|
"completions/mean_terminated_length": 56.25,
|
|
"completions/min_length": 33.0,
|
|
"completions/min_terminated_length": 33.0,
|
|
"epoch": 2.2222222222222223,
|
|
"grad_norm": 19.803834915161133,
|
|
"kl": 0.08400401845574379,
|
|
"learning_rate": 6.819348298638839e-09,
|
|
"loss": -0.051,
|
|
"num_tokens": 15164.0,
|
|
"reward": 0.762055516242981,
|
|
"reward_std": 0.058255478739738464,
|
|
"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": 0.26205551624298096,
|
|
"rewards/question_recreation_reward_func/std": 0.15907590091228485,
|
|
"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": 0.5,
|
|
"rewards/xmlcount_reward_func/std": 0.0,
|
|
"step": 20
|
|
},
|
|
{
|
|
"epoch": 2.2222222222222223,
|
|
"step": 20,
|
|
"total_flos": 0.0,
|
|
"train_loss": -0.04282761747017503,
|
|
"train_runtime": 6256.5978,
|
|
"train_samples_per_second": 0.013,
|
|
"train_steps_per_second": 0.003
|
|
}
|
|
],
|
|
"logging_steps": 1,
|
|
"max_steps": 20,
|
|
"num_input_tokens_seen": 15164,
|
|
"num_train_epochs": 3,
|
|
"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
|
|
}
|