eval_results.json

{
    "epoch": 0.9984827495217362,
    "eval_logits/chosen": -1.4033699035644531,
    "eval_logits/rejected": -1.3698593378067017,
    "eval_logps/chosen": -627.2847900390625,
    "eval_logps/rejected": -837.4208374023438,
    "eval_loss": 0.3820858299732208,
    "eval_rewards/accuracies": 0.8663967847824097,
    "eval_rewards/chosen": -3.5029525756835938,
    "eval_rewards/margins": 2.0145938396453857,
    "eval_rewards/rejected": -5.517546653747559,
    "eval_runtime": 316.7033,
    "eval_samples": 1976,
    "eval_samples_per_second": 6.239,
    "eval_steps_per_second": 1.56
}