Model: jackf857/llama-3-8b-base-new-dpo-hh-harmless-4xh200-batch-64-q_t-0.5-s_star-0.6 Source: Original Platform
23 lines
837 B
JSON
23 lines
837 B
JSON
{
|
|
"epoch": 0.999244142101285,
|
|
"eval_fcm_dpo/beta": 0.017207348719239235,
|
|
"eval_logits/chosen": 0.6921316981315613,
|
|
"eval_logits/rejected": 0.6450904607772827,
|
|
"eval_logps/chosen": -139.775146484375,
|
|
"eval_logps/ref_chosen": -74.85946655273438,
|
|
"eval_logps/ref_rejected": -79.54898834228516,
|
|
"eval_logps/rejected": -178.9794158935547,
|
|
"eval_loss": 0.5342118144035339,
|
|
"eval_margin_dpo/margin_mean": 34.5147590637207,
|
|
"eval_margin_dpo/margin_std": 53.65407180786133,
|
|
"eval_runtime": 37.9398,
|
|
"eval_samples": 2303,
|
|
"eval_samples_per_second": 60.701,
|
|
"eval_steps_per_second": 1.898,
|
|
"total_flos": 0.0,
|
|
"train_loss": 1.0886367675756723,
|
|
"train_runtime": 1755.0349,
|
|
"train_samples": 42336,
|
|
"train_samples_per_second": 24.123,
|
|
"train_steps_per_second": 0.377
|
|
} |