qwen_orpo_entropy_0_01 / all_results.json
yakazimir's picture
End of training
58f3217 verified
{
"epoch": 2.999297541394882,
"eval_logits/chosen": 0.05207434669137001,
"eval_logits/rejected": 0.16449397802352905,
"eval_logps/chosen": -3.185479164123535,
"eval_logps/rejected": -4.173889636993408,
"eval_loss": 0.558941662311554,
"eval_rewards/accuracies": 0.7225519418716431,
"eval_rewards/chosen": -3.185479164123535,
"eval_rewards/margins": 0.9884104132652283,
"eval_rewards/rejected": -4.173889636993408,
"eval_runtime": 50.5035,
"eval_samples": 1345,
"eval_samples_per_second": 26.632,
"eval_sft_loss": 3.3163414001464844,
"eval_steps_per_second": 6.673,
"total_flos": 0.0,
"train_loss": 0.5562219639746825,
"train_runtime": 39292.0664,
"train_samples": 59790,
"train_samples_per_second": 4.565,
"train_steps_per_second": 0.143
}