all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": 1.1020585298538208,
    "eval_logps/chosen": -354.015869140625,
    "eval_logps/rejected": -324.1587219238281,
    "eval_loss": NaN,
    "eval_nll_loss": NaN,
    "eval_rewards/accuracies": 0.6617063283920288,
    "eval_rewards/chosen": 1.3769841194152832,
    "eval_rewards/margins": 0.4590735137462616,
    "eval_rewards/rejected": 0.9184337854385376,
    "eval_runtime": 8.4136,
    "eval_samples_per_second": 118.856,
    "eval_steps_per_second": 7.488
}