eval_results.json

{
    "epoch": 2.955223880597015,
    "eval_logits/chosen": -2.3310205936431885,
    "eval_logits/rejected": -2.3245856761932373,
    "eval_logps/chosen": -280.4070129394531,
    "eval_logps/rejected": -236.11550903320312,
    "eval_loss": 0.5229008197784424,
    "eval_rewards/accuracies": 0.8166666626930237,
    "eval_rewards/chosen": 2.1573832035064697,
    "eval_rewards/margins": 2.773956775665283,
    "eval_rewards/rejected": -0.616573691368103,
    "eval_runtime": 61.6988,
    "eval_samples_per_second": 15.397,
    "eval_steps_per_second": 0.243
}