Model: W-61/llama-3-8b-base-new-dpo-ultrafeedback-4xh200-batch-128-q_t-0.45-s_star-0.45-20260427-221551 Source: Original Platform
21 lines
826 B
JSON
21 lines
826 B
JSON
{
|
|
"epoch": 0.9989528795811519,
|
|
"eval_KL/chosen_KL_mean": -174.78041076660156,
|
|
"eval_KL/mean": -214.0789337158203,
|
|
"eval_KL/rejected_KL_mean": -253.3774871826172,
|
|
"eval_KL/std": 108.87296295166016,
|
|
"eval_fcm_dpo/beta": 0.004879707470536232,
|
|
"eval_logits/chosen": -0.8583829998970032,
|
|
"eval_logits/rejected": -0.8428274393081665,
|
|
"eval_logps/chosen": -462.607177734375,
|
|
"eval_logps/ref_chosen": -287.8268127441406,
|
|
"eval_logps/ref_rejected": -266.9300231933594,
|
|
"eval_logps/rejected": -520.3075561523438,
|
|
"eval_loss": 0.5694514513015747,
|
|
"eval_margin_dpo/margin_mean": 78.59707641601562,
|
|
"eval_margin_dpo/margin_std": 121.47869110107422,
|
|
"eval_runtime": 78.2536,
|
|
"eval_samples": 2000,
|
|
"eval_samples_per_second": 25.558,
|
|
"eval_steps_per_second": 3.195
|
|
} |