all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": 1.1602182388305664,
    "eval_logps/chosen": -357.71429443359375,
    "eval_logps/rejected": -325.9364929199219,
    "eval_loss": NaN,
    "eval_nll_loss": NaN,
    "eval_rewards/accuracies": 0.6051587462425232,
    "eval_rewards/chosen": 1.00458824634552,
    "eval_rewards/margins": 0.2564484179019928,
    "eval_rewards/rejected": 0.7491319179534912,
    "eval_runtime": 8.4131,
    "eval_samples_per_second": 118.863,
    "eval_steps_per_second": 7.488
}