eval_results.json

{
    "epoch": 0.995053003533569,
    "eval_logits/chosen": 1.8515108823776245,
    "eval_logits/rejected": 1.5509034395217896,
    "eval_logps/chosen": -1.8485636711120605,
    "eval_logps/rejected": -4.492039680480957,
    "eval_loss": 0.8060641288757324,
    "eval_rewards/accuracies": 0.9873417615890503,
    "eval_rewards/chosen": -0.1848563700914383,
    "eval_rewards/margins": 0.264347642660141,
    "eval_rewards/rejected": -0.4492039680480957,
    "eval_runtime": 111.5971,
    "eval_samples_per_second": 11.273,
    "eval_steps_per_second": 1.416
}