all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": 1.106212854385376,
    "eval_logps/chosen": -365.8730163574219,
    "eval_logps/rejected": -334.4444580078125,
    "eval_loss": NaN,
    "eval_nll_loss": NaN,
    "eval_rewards/accuracies": 0.608134925365448,
    "eval_rewards/chosen": 0.1799234002828598,
    "eval_rewards/margins": 0.2850082516670227,
    "eval_rewards/rejected": -0.1048341765999794,
    "eval_runtime": 8.3996,
    "eval_samples_per_second": 119.053,
    "eval_steps_per_second": 7.5
}