dpo_06230018_policy2_0.01 / eval_results.json
WDong's picture
Upload 17 files
c53888e verified
raw
history blame contribute delete
560 Bytes
{
"epoch": 2.994495412844037,
"eval_logits/chosen": -0.02289891056716442,
"eval_logits/rejected": -0.9417744874954224,
"eval_logps/chosen": -300.2835388183594,
"eval_logps/rejected": -281.21466064453125,
"eval_loss": 0.36351555585861206,
"eval_rewards/accuracies": 0.9777777791023254,
"eval_rewards/chosen": -0.15967033803462982,
"eval_rewards/margins": 0.9239291548728943,
"eval_rewards/rejected": -1.08359956741333,
"eval_runtime": 8.9717,
"eval_samples_per_second": 5.016,
"eval_steps_per_second": 5.016
}