zephyr-7b-dpo-qlora / all_results.json
RedaAlami's picture
End of training
04e7632 verified
raw
history blame
768 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": 2.140789270401001,
"eval_logits/rejected": 1.9412086009979248,
"eval_logps/chosen": -503.5469665527344,
"eval_logps/rejected": -1102.4208984375,
"eval_loss": 0.028598472476005554,
"eval_rewards/accuracies": 0.9253731369972229,
"eval_rewards/chosen": -4.707759380340576,
"eval_rewards/margins": 5.957462787628174,
"eval_rewards/rejected": -10.66522216796875,
"eval_runtime": 214.7372,
"eval_samples": 21417,
"eval_samples_per_second": 99.736,
"eval_steps_per_second": 1.56,
"total_flos": 0.0,
"train_loss": 0.19036180805619987,
"train_runtime": 15997.0818,
"train_samples": 406907,
"train_samples_per_second": 25.436,
"train_steps_per_second": 0.199
}