qwen_cpo_entropy_0_1 / all_results.json
yakazimir's picture
End of training
b37f9ad verified
raw
history blame contribute delete
819 Bytes
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 0.18746721744537354,
"eval_logits/rejected": 0.3000442683696747,
"eval_logps/chosen": -1.7146226167678833,
"eval_logps/rejected": -2.37273907661438,
"eval_loss": 0.7405069470405579,
"eval_rewards/accuracies": 0.6772996783256531,
"eval_rewards/chosen": -1.7146226167678833,
"eval_rewards/margins": 0.6581166386604309,
"eval_rewards/rejected": -2.37273907661438,
"eval_runtime": 47.602,
"eval_samples": 1345,
"eval_samples_per_second": 28.255,
"eval_sft_loss": 1.6847702264785767,
"eval_steps_per_second": 7.08,
"total_flos": 0.0,
"train_loss": 0.7456102068299996,
"train_runtime": 34974.0603,
"train_samples": 59790,
"train_samples_per_second": 5.129,
"train_steps_per_second": 0.16
}