qwen_orpo_entropy / all_results.json
yakazimir's picture
End of training
f64a019 verified
raw
history blame contribute delete
779 Bytes
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 0.23467637598514557,
"eval_logits/rejected": 0.3310812711715698,
"eval_logps/chosen": -5.230490684509277,
"eval_logps/rejected": -6.346008777618408,
"eval_loss": 0.525705099105835,
"eval_rewards/accuracies": 0.7284866571426392,
"eval_rewards/chosen": -5.230490684509277,
"eval_rewards/margins": 1.1155189275741577,
"eval_rewards/rejected": -6.346008777618408,
"eval_runtime": 40.4557,
"eval_samples": 1345,
"eval_samples_per_second": 33.246,
"eval_steps_per_second": 8.33,
"total_flos": 0.0,
"train_loss": 0.5174550147332267,
"train_runtime": 29971.1802,
"train_samples": 59790,
"train_samples_per_second": 5.985,
"train_steps_per_second": 0.187
}