zephyr-7b-dpo-lora / all_results.json
LaoRay's picture
End of training
050c558 verified
{
"epoch": 19.84,
"eval_logits/chosen": -1.7818838357925415,
"eval_logits/rejected": -1.7583956718444824,
"eval_logps/chosen": -1971.018798828125,
"eval_logps/rejected": -2233.02197265625,
"eval_loss": 1.56223726272583,
"eval_rewards/accuracies": 0.6499999761581421,
"eval_rewards/chosen": -17.027843475341797,
"eval_rewards/margins": 2.817903757095337,
"eval_rewards/rejected": -19.845748901367188,
"eval_runtime": 67.4863,
"eval_samples": 200,
"eval_samples_per_second": 2.964,
"eval_steps_per_second": 0.37,
"total_flos": 0.0,
"train_loss": 0.13093096087141606,
"train_runtime": 15450.2645,
"train_samples": 1000,
"train_samples_per_second": 1.294,
"train_steps_per_second": 0.08
}