eval_results.json

{
    "epoch": 0.64,
    "eval_logits/chosen": -0.4409240782260895,
    "eval_logits/rejected": -0.3783179819583893,
    "eval_logps/chosen": -1.6689454317092896,
    "eval_logps/rejected": -1.7649023532867432,
    "eval_loss": 5.835046768188477,
    "eval_rewards/accuracies": 0.5245413780212402,
    "eval_rewards/chosen": -16.689455032348633,
    "eval_rewards/margins": 0.9595676064491272,
    "eval_rewards/rejected": -17.649023056030273,
    "eval_runtime": 294.9751,
    "eval_samples": 5595,
    "eval_samples_per_second": 18.968,
    "eval_steps_per_second": 4.743
}