eval_results.json

{
    "epoch": 0.9995638901003053,
    "eval_logits/chosen": -2.598531484603882,
    "eval_logits/rejected": -2.6629889011383057,
    "eval_logps/chosen": -549.3650512695312,
    "eval_logps/rejected": -645.505126953125,
    "eval_loss": 0.47948434948921204,
    "eval_rewards/accuracies": 0.7259036302566528,
    "eval_rewards/chosen": -1.6860378980636597,
    "eval_rewards/margins": 1.1272025108337402,
    "eval_rewards/rejected": -2.8132402896881104,
    "eval_runtime": 114.4643,
    "eval_samples": 2643,
    "eval_samples_per_second": 23.09,
    "eval_steps_per_second": 0.725
}