Model: jackf857/llama-3-8b-base-new-dpo-hh-helpful-s_star0.4-4xh200-batch-64-20260421-214335-rerun Source: Original Platform
20 lines
770 B
JSON
20 lines
770 B
JSON
{
|
|
"epoch": 1.0,
|
|
"eval_fcm_dpo/beta": 0.001050772494636476,
|
|
"eval_fcm_dpo/delta": 0.029544111341238022,
|
|
"eval_fcm_dpo/margin": 286.21771240234375,
|
|
"eval_fcm_dpo/q_t": 0.4338943064212799,
|
|
"eval_logits/chosen": -1.0337938070297241,
|
|
"eval_logits/rejected": -1.0381489992141724,
|
|
"eval_logps/chosen": -853.6936645507812,
|
|
"eval_logps/ref_chosen": -79.05104064941406,
|
|
"eval_logps/ref_rejected": -86.79793548583984,
|
|
"eval_logps/rejected": -1147.6583251953125,
|
|
"eval_loss": 0.6071858406066895,
|
|
"eval_margin_dpo/margin_mean": 286.21771240234375,
|
|
"eval_margin_dpo/margin_std": 631.6002807617188,
|
|
"eval_runtime": 39.5883,
|
|
"eval_samples": 2339,
|
|
"eval_samples_per_second": 59.083,
|
|
"eval_steps_per_second": 1.869
|
|
} |