eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.04874047264456749,
    "eval_logits/rejected": 0.09153472632169724,
    "eval_logps/chosen": -363.42431640625,
    "eval_logps/rejected": -373.5458984375,
    "eval_loss": 0.036668241024017334,
    "eval_rewards/accuracies": 0.6334999799728394,
    "eval_rewards/chosen": -0.08587750047445297,
    "eval_rewards/margins": 0.04385250434279442,
    "eval_rewards/rejected": -0.1297300010919571,
    "eval_runtime": 545.8695,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.664,
    "eval_steps_per_second": 0.916
}