Model: W-61/llama3-8b-base-new-method-s_star0.6-20260426-230653 Source: Original Platform
23 lines
839 B
JSON
23 lines
839 B
JSON
{
|
|
"epoch": 0.9989528795811519,
|
|
"eval_fcm_dpo/beta": 0.009538780897855759,
|
|
"eval_logits/chosen": -0.8344770073890686,
|
|
"eval_logits/rejected": -0.8241644501686096,
|
|
"eval_logps/chosen": -390.4581604003906,
|
|
"eval_logps/ref_chosen": -287.8268127441406,
|
|
"eval_logps/ref_rejected": -266.9300231933594,
|
|
"eval_logps/rejected": -424.9966735839844,
|
|
"eval_loss": 0.5423681139945984,
|
|
"eval_margin_dpo/margin_mean": 55.43527603149414,
|
|
"eval_margin_dpo/margin_std": 80.3389663696289,
|
|
"eval_runtime": 78.5759,
|
|
"eval_samples": 2000,
|
|
"eval_samples_per_second": 25.453,
|
|
"eval_steps_per_second": 3.182,
|
|
"total_flos": 0.0,
|
|
"train_loss": 4.544049396954742,
|
|
"train_runtime": 6685.0038,
|
|
"train_samples": 61135,
|
|
"train_samples_per_second": 9.145,
|
|
"train_steps_per_second": 0.071
|
|
} |