eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 98.40702056884766,
    "eval_logits/rejected": 98.40592193603516,
    "eval_logps/chosen": -32.23640441894531,
    "eval_logps/rejected": -35.8593864440918,
    "eval_loss": 0.4850274920463562,
    "eval_rewards/accuracies": 0.5398671627044678,
    "eval_rewards/chosen": 0.1654178649187088,
    "eval_rewards/margins": 0.07970692217350006,
    "eval_rewards/rejected": 0.08571097254753113,
    "eval_runtime": 103.9265,
    "eval_samples": 343,
    "eval_samples_per_second": 3.3,
    "eval_steps_per_second": 0.414
}