eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.15863379836082458,
    "eval_logits/rejected": -0.3815014362335205,
    "eval_logps/chosen": -523.9338989257812,
    "eval_logps/rejected": -821.0360717773438,
    "eval_loss": 0.6073186993598938,
    "eval_rewards/accuracies": 0.8214285969734192,
    "eval_rewards/chosen": 0.07047966867685318,
    "eval_rewards/margins": 0.24993744492530823,
    "eval_rewards/rejected": -0.17945778369903564,
    "eval_runtime": 22.8385,
    "eval_samples": 208,
    "eval_samples_per_second": 9.107,
    "eval_steps_per_second": 0.307
}