qwen_cpo_entropy_0_01 / all_results.json
yakazimir's picture
End of training
058d247 verified
raw
history blame
822 Bytes
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 0.15628807246685028,
"eval_logits/rejected": 0.2810993194580078,
"eval_logps/chosen": -3.3285255432128906,
"eval_logps/rejected": -4.380990505218506,
"eval_loss": 0.5583173632621765,
"eval_rewards/accuracies": 0.7225519418716431,
"eval_rewards/chosen": -3.3285255432128906,
"eval_rewards/margins": 1.0524649620056152,
"eval_rewards/rejected": -4.380990505218506,
"eval_runtime": 50.3171,
"eval_samples": 1345,
"eval_samples_per_second": 26.73,
"eval_sft_loss": 3.470508575439453,
"eval_steps_per_second": 6.698,
"total_flos": 0.0,
"train_loss": 0.5516859670777903,
"train_runtime": 39088.5646,
"train_samples": 59790,
"train_samples_per_second": 4.589,
"train_steps_per_second": 0.143
}