eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 97.9656753540039,
    "eval_logits/rejected": 97.93604278564453,
    "eval_logps/chosen": -33.003868103027344,
    "eval_logps/rejected": -36.95783615112305,
    "eval_loss": 0.6761856079101562,
    "eval_rewards/accuracies": 0.5888704061508179,
    "eval_rewards/chosen": -0.05606912076473236,
    "eval_rewards/margins": 0.04306147247552872,
    "eval_rewards/rejected": -0.09913058578968048,
    "eval_runtime": 103.6924,
    "eval_samples": 343,
    "eval_samples_per_second": 3.308,
    "eval_steps_per_second": 0.415
}