Model: jackf857/llama-3-8b-base-new-dpo-hh-helpful-4xh200-batch-64-q_t-0.5-s_star-0.85 Source: Original Platform
17 lines
631 B
JSON
17 lines
631 B
JSON
{
|
|
"epoch": 1.0,
|
|
"eval_fcm_dpo/beta": 0.0035767327062785625,
|
|
"eval_logits/chosen": 0.3511611819267273,
|
|
"eval_logits/rejected": 0.37540391087532043,
|
|
"eval_logps/chosen": -578.7294311523438,
|
|
"eval_logps/ref_chosen": -79.05104064941406,
|
|
"eval_logps/ref_rejected": -86.79793548583984,
|
|
"eval_logps/rejected": -737.1654663085938,
|
|
"eval_loss": 0.5333206057548523,
|
|
"eval_margin_dpo/margin_mean": 150.68910217285156,
|
|
"eval_margin_dpo/margin_std": 227.09837341308594,
|
|
"eval_runtime": 39.2308,
|
|
"eval_samples": 2339,
|
|
"eval_samples_per_second": 59.622,
|
|
"eval_steps_per_second": 1.886
|
|
} |