qwen_cCPO_entropy_0_01 / all_results.json
yakazimir's picture
End of training
60f572b verified
raw
history blame contribute delete
824 Bytes
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 0.13457134366035461,
"eval_logits/rejected": 0.2371775060892105,
"eval_logps/chosen": -1.6158760786056519,
"eval_logps/rejected": -2.194223403930664,
"eval_loss": 0.48958608508110046,
"eval_rewards/accuracies": 0.671364963054657,
"eval_rewards/chosen": -1.6158760786056519,
"eval_rewards/margins": 0.5783473253250122,
"eval_rewards/rejected": -2.194223403930664,
"eval_runtime": 42.892,
"eval_samples": 1345,
"eval_samples_per_second": 31.358,
"eval_sft_loss": 1.6913238763809204,
"eval_steps_per_second": 7.857,
"total_flos": 0.0,
"train_loss": 0.49661777635968474,
"train_runtime": 31536.5493,
"train_samples": 59790,
"train_samples_per_second": 5.688,
"train_steps_per_second": 0.178
}