eval_results.json

{
    "eval_logits/chosen": -0.5001821517944336,
    "eval_logits/rejected": -0.5001798868179321,
    "eval_logps/chosen": -268.5600891113281,
    "eval_logps/rejected": -269.17041015625,
    "eval_loss": -0.029902135953307152,
    "eval_rewards/chosen": -0.8525950312614441,
    "eval_rewards/rejected": -1.1856228113174438,
    "eval_runtime": 4761.786,
    "eval_samples": 20000,
    "eval_samples_per_second": 4.2,
    "eval_steps_per_second": 2.1
}