eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.8309964537620544,
    "eval_logits/rejected": -0.9202789068222046,
    "eval_logps/chosen": -859.0581665039062,
    "eval_logps/rejected": -1444.6895751953125,
    "eval_loss": 0.4459438920021057,
    "eval_rewards/accuracies": 0.8068181872367859,
    "eval_rewards/chosen": -0.4565673768520355,
    "eval_rewards/margins": 0.9446592926979065,
    "eval_rewards/rejected": -1.4012266397476196,
    "eval_runtime": 80.5227,
    "eval_samples": 676,
    "eval_samples_per_second": 8.395,
    "eval_steps_per_second": 0.273
}