eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.4666408896446228,
    "eval_logits/rejected": -0.32667988538742065,
    "eval_logps/chosen": -426.2104187011719,
    "eval_logps/rejected": -941.0735473632812,
    "eval_loss": 0.6054050922393799,
    "eval_rewards/accuracies": 0.7867646813392639,
    "eval_rewards/chosen": 0.03273159638047218,
    "eval_rewards/margins": 0.20916865766048431,
    "eval_rewards/rejected": -0.17643706500530243,
    "eval_runtime": 62.3876,
    "eval_samples": 520,
    "eval_samples_per_second": 8.335,
    "eval_steps_per_second": 0.272
}