Files
llama-3-8b-base-margin-dpo-…/eval_results.json
ModelHub XC 444c4324b6 初始化项目,由ModelHub XC社区提供模型
Model: W-61/llama-3-8b-base-margin-dpo-hh-helpful-8xh200
Source: Original Platform
2026-04-24 11:44:07 +08:00

16 lines
584 B
JSON

{
"epoch": 1.0,
"eval_logits/chosen": -0.5745480060577393,
"eval_logits/rejected": -0.535261869430542,
"eval_logps/chosen": -119.71295928955078,
"eval_logps/ref_chosen": -97.0617446899414,
"eval_logps/ref_rejected": -80.18183135986328,
"eval_logps/rejected": -113.98292541503906,
"eval_loss": 0.4582546055316925,
"eval_margin_dpo/margin_mean": 11.149892807006836,
"eval_margin_dpo/margin_std": 15.04123592376709,
"eval_runtime": 20.2691,
"eval_samples": 2339,
"eval_samples_per_second": 115.397,
"eval_steps_per_second": 0.937
}