Model: jackf857/qwen3-8b-base-orpo-ultrafeedback-4xh200-batch-128 Source: Original Platform
19 lines
730 B
JSON
19 lines
730 B
JSON
{
|
|
"epoch": 0.9989528795811519,
|
|
"eval_log_odds_chosen": 0.30259791016578674,
|
|
"eval_log_odds_ratio": -0.6631017923355103,
|
|
"eval_logits/chosen": 2.102327585220337,
|
|
"eval_logits/rejected": 2.1291346549987793,
|
|
"eval_logps/chosen": -0.8522344827651978,
|
|
"eval_logps/rejected": -1.0493180751800537,
|
|
"eval_loss": 1.078126072883606,
|
|
"eval_nll_loss": 1.1097657680511475,
|
|
"eval_rewards/accuracies": 0.6039999723434448,
|
|
"eval_rewards/chosen": -0.008522345684468746,
|
|
"eval_rewards/margins": 0.0019708359614014626,
|
|
"eval_rewards/rejected": -0.010493181645870209,
|
|
"eval_runtime": 45.1175,
|
|
"eval_samples": 2000,
|
|
"eval_samples_per_second": 44.329,
|
|
"eval_steps_per_second": 5.541
|
|
} |