Model: jackf857/qwen3-8b-base-new-dpo-hh-helpful-4xh200-batch-64-q_t-0.45-s_star-0.6 Source: Original Platform
23 lines
829 B
JSON
23 lines
829 B
JSON
{
|
|
"epoch": 1.0,
|
|
"eval_fcm_dpo/beta": 0.0065701245330274105,
|
|
"eval_logits/chosen": -1.905737042427063,
|
|
"eval_logits/rejected": -1.5243432521820068,
|
|
"eval_logps/chosen": -265.6040954589844,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -333.0381164550781,
|
|
"eval_loss": 0.5309035181999207,
|
|
"eval_margin_dpo/margin_mean": 73.85983276367188,
|
|
"eval_margin_dpo/margin_std": 100.37892150878906,
|
|
"eval_runtime": 43.7708,
|
|
"eval_samples": 2339,
|
|
"eval_samples_per_second": 53.437,
|
|
"eval_steps_per_second": 1.691,
|
|
"total_flos": 0.0,
|
|
"train_loss": 1.0153061533114283,
|
|
"train_runtime": 2239.7507,
|
|
"train_samples": 43598,
|
|
"train_samples_per_second": 19.466,
|
|
"train_steps_per_second": 0.304
|
|
} |