eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.8107545375823975,
    "eval_logits/rejected": -1.547254204750061,
    "eval_logps/chosen": -121.36690521240234,
    "eval_logps/rejected": -144.17279052734375,
    "eval_loss": 0.6354531049728394,
    "eval_rewards/accuracies": 0.7326388955116272,
    "eval_rewards/chosen": 0.1376209855079651,
    "eval_rewards/margins": 0.13830368220806122,
    "eval_rewards/rejected": -0.0006826933822594583,
    "eval_runtime": 108.9795,
    "eval_samples": 1144,
    "eval_samples_per_second": 10.497,
    "eval_steps_per_second": 0.33
}