eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 98.3665542602539,
    "eval_logits/rejected": 98.35928344726562,
    "eval_logps/chosen": -32.303466796875,
    "eval_logps/rejected": -35.77931594848633,
    "eval_loss": 0.5059725642204285,
    "eval_rewards/accuracies": 0.5220099687576294,
    "eval_rewards/chosen": 0.09779632091522217,
    "eval_rewards/margins": -0.033251672983169556,
    "eval_rewards/rejected": 0.13104799389839172,
    "eval_runtime": 103.819,
    "eval_samples": 343,
    "eval_samples_per_second": 3.304,
    "eval_steps_per_second": 0.414
}