rlced-conifer-zephyr-7b-dpo-2e / eval_results.json
NicholasCorrado's picture
End of training
cb19d38 verified
raw
history blame contribute delete
580 Bytes
{
"epoch": 1.996886351842242,
"eval_logits/chosen": -0.16769856214523315,
"eval_logits/rejected": 3.019533395767212,
"eval_logps/chosen": -1311.4659423828125,
"eval_logps/rejected": -2617.66845703125,
"eval_loss": 0.15781185030937195,
"eval_rewards/accuracies": 0.9215686321258545,
"eval_rewards/chosen": -8.88057804107666,
"eval_rewards/margins": 12.836706161499023,
"eval_rewards/rejected": -21.71728515625,
"eval_runtime": 297.083,
"eval_samples": 6491,
"eval_samples_per_second": 21.849,
"eval_steps_per_second": 0.343
}