zephyr-7b-dpo-full / all_results.json
Beanpow's picture
End of training
349f587 verified
raw
history blame contribute delete
784 Bytes
{
"epoch": 0.9994767137624281,
"eval_logits/chosen": -1.0034536123275757,
"eval_logits/rejected": -0.6818602085113525,
"eval_logps/chosen": -1738.2310791015625,
"eval_logps/rejected": -2073.0146484375,
"eval_loss": 1.3376935720443726,
"eval_rewards/accuracies": 0.6388888955116272,
"eval_rewards/chosen": -14.562629699707031,
"eval_rewards/margins": 3.5654470920562744,
"eval_rewards/rejected": -18.12807846069336,
"eval_runtime": 177.2147,
"eval_samples": 2000,
"eval_samples_per_second": 11.286,
"eval_steps_per_second": 0.356,
"total_flos": 0.0,
"train_loss": 2.1165736393154604,
"train_runtime": 18133.1885,
"train_samples": 61134,
"train_samples_per_second": 3.371,
"train_steps_per_second": 0.053
}