zephyr-7b-dpo-qlora / all_results.json
RedaAlami's picture
End of training
7d0f0dc verified
raw
history blame contribute delete
774 Bytes
{
"epoch": 1.0,
"eval_logits/chosen": 2.2557647228240967,
"eval_logits/rejected": 2.0469555854797363,
"eval_logps/chosen": -496.4129333496094,
"eval_logps/rejected": -1080.7012939453125,
"eval_loss": 0.029871011152863503,
"eval_rewards/accuracies": 0.9305970072746277,
"eval_rewards/chosen": -4.636161804199219,
"eval_rewards/margins": 5.8116912841796875,
"eval_rewards/rejected": -10.44785213470459,
"eval_runtime": 215.9125,
"eval_samples": 21417,
"eval_samples_per_second": 99.193,
"eval_steps_per_second": 1.552,
"total_flos": 0.0,
"train_loss": 0.18858218038370866,
"train_runtime": 16104.4918,
"train_samples": 406907,
"train_samples_per_second": 25.267,
"train_steps_per_second": 0.197
}