eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.5471837520599365,
    "eval_logits/rejected": -2.4984941482543945,
    "eval_logps/chosen": -255.5736846923828,
    "eval_logps/rejected": -296.26116943359375,
    "eval_loss": 0.5782328248023987,
    "eval_rewards/accuracies": 0.6926282048225403,
    "eval_rewards/chosen": -0.21196967363357544,
    "eval_rewards/margins": 0.4882549047470093,
    "eval_rewards/rejected": -0.7002245783805847,
    "eval_runtime": 862.1588,
    "eval_samples": 12451,
    "eval_samples_per_second": 14.442,
    "eval_steps_per_second": 0.452
}