eval_results.json

{
    "epoch": 0.9984301412872841,
    "eval_logits/chosen": -1.234063982963562,
    "eval_logits/rejected": -1.2614085674285889,
    "eval_logps/chosen": -752.220947265625,
    "eval_logps/rejected": -849.2994384765625,
    "eval_loss": 0.51555335521698,
    "eval_rewards/accuracies": 0.765625,
    "eval_rewards/chosen": -4.394815921783447,
    "eval_rewards/margins": 1.0356953144073486,
    "eval_rewards/rejected": -5.430510997772217,
    "eval_runtime": 33.965,
    "eval_samples": 2000,
    "eval_samples_per_second": 58.884,
    "eval_steps_per_second": 0.942
}