eval_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.107677698135376,
    "eval_logits/rejected": -2.095085382461548,
    "eval_logps/chosen": -144.84811401367188,
    "eval_logps/rejected": -168.68983459472656,
    "eval_loss": 0.6382296681404114,
    "eval_rewards/accuracies": 0.634061336517334,
    "eval_rewards/chosen": -0.8613619208335876,
    "eval_rewards/margins": 0.1937350332736969,
    "eval_rewards/rejected": -1.0550971031188965,
    "eval_runtime": 383.6957,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.217,
    "eval_steps_per_second": 1.402
}