Files
llama-3-8b-base-beta-dpo-ul…/eval_results.json
ModelHub XC 4768635cd7 初始化项目,由ModelHub XC社区提供模型
Model: W-61/llama-3-8b-base-beta-dpo-ultrafeedback-4xh200-batch-128-20260424-044124
Source: Original Platform
2026-05-09 15:51:41 +08:00

15 lines
546 B
JSON

{
"epoch": 0.9989528795811519,
"eval_beta_dpo/beta_used": 0.026403456926345825,
"eval_beta_dpo/beta_used_raw": -0.012573433108627796,
"eval_beta_dpo/gap_mean": 33.88580322265625,
"eval_beta_dpo/gap_std": 54.5393180847168,
"eval_beta_dpo/mask_keep_frac": 1.0,
"eval_logits/chosen": -0.8373056650161743,
"eval_logits/rejected": -0.8195577263832092,
"eval_loss": 0.6150403618812561,
"eval_runtime": 81.4443,
"eval_samples": 2000,
"eval_samples_per_second": 24.557,
"eval_steps_per_second": 1.535
}