all_results.json

{
    "epoch": 2.9993562508046865,
    "eval_logits/chosen": NaN,
    "eval_logits/rejected": -6.718367099761963,
    "eval_logps/chosen": -708.9367065429688,
    "eval_logps/rejected": -595.2987670898438,
    "eval_loss": 0.6400840878486633,
    "eval_rewards/accuracies": 0.593999981880188,
    "eval_rewards/chosen": -0.11233817785978317,
    "eval_rewards/margins": 0.16944476962089539,
    "eval_rewards/rejected": -0.28178295493125916,
    "eval_runtime": 30.4441,
    "eval_samples_per_second": 32.847,
    "eval_steps_per_second": 4.106
}