Model: jackf857/llama-3-8b-base-cpo-ultrafeedback-4xH200-batch-128-rerun Source: Original Platform
17 lines
627 B
JSON
17 lines
627 B
JSON
{
|
|
"epoch": 0.9989528795811519,
|
|
"eval_logits/chosen": -0.7155267000198364,
|
|
"eval_logits/rejected": -0.7125465869903564,
|
|
"eval_logps/chosen": -272.6193542480469,
|
|
"eval_logps/rejected": -267.1548767089844,
|
|
"eval_loss": 2.030885934829712,
|
|
"eval_nll_loss": 0.9492784738540649,
|
|
"eval_rewards/accuracies": 0.5180000066757202,
|
|
"eval_rewards/chosen": -2.7261929512023926,
|
|
"eval_rewards/margins": -0.05464465916156769,
|
|
"eval_rewards/rejected": -2.67154860496521,
|
|
"eval_runtime": 40.4909,
|
|
"eval_samples": 2000,
|
|
"eval_samples_per_second": 49.394,
|
|
"eval_steps_per_second": 3.087
|
|
} |