eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.8041099309921265,
    "eval_logits/rejected": -1.6730901002883911,
    "eval_logps/chosen": -373.1398620605469,
    "eval_logps/rejected": -354.2070617675781,
    "eval_loss": 0.6285176873207092,
    "eval_rewards/accuracies": 0.6805555820465088,
    "eval_rewards/chosen": -0.30501797795295715,
    "eval_rewards/margins": 0.23024281859397888,
    "eval_rewards/rejected": -0.535260796546936,
    "eval_runtime": 337.0045,
    "eval_samples": 2000,
    "eval_samples_per_second": 5.935,
    "eval_steps_per_second": 0.187
}