all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": 1.2637648582458496,
    "eval_logps/chosen": -362.6666564941406,
    "eval_logps/rejected": -329.5079345703125,
    "eval_loss": NaN,
    "eval_nll_loss": NaN,
    "eval_rewards/accuracies": 0.5257936716079712,
    "eval_rewards/chosen": 0.504169762134552,
    "eval_rewards/margins": 0.1219865083694458,
    "eval_rewards/rejected": 0.3821769654750824,
    "eval_runtime": 8.3925,
    "eval_samples_per_second": 119.153,
    "eval_steps_per_second": 7.507
}