Files
qwen3-8b-base-new-dpo-hh-ha…/eval_results.json

17 lines
641 B
JSON
Raw Normal View History

{
"epoch": 0.999244142101285,
"eval_fcm_dpo/beta": 0.00711857108399272,
"eval_logits/chosen": 1.5523120164871216,
"eval_logits/rejected": 1.4120718240737915,
"eval_logps/chosen": -173.13360595703125,
"eval_logps/ref_chosen": -86.90177917480469,
"eval_logps/ref_rejected": -96.69639587402344,
"eval_logps/rejected": -234.51638793945312,
"eval_loss": 0.5718919038772583,
"eval_margin_dpo/margin_mean": 51.58818817138672,
"eval_margin_dpo/margin_std": 86.1289291381836,
"eval_runtime": 42.2472,
"eval_samples": 2303,
"eval_samples_per_second": 54.513,
"eval_steps_per_second": 1.704
}