eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 98.4628677368164,
    "eval_logits/rejected": 98.43635559082031,
    "eval_logps/chosen": -32.64057159423828,
    "eval_logps/rejected": -36.409393310546875,
    "eval_loss": 0.686401903629303,
    "eval_rewards/accuracies": 0.5685215592384338,
    "eval_rewards/chosen": -0.07895812392234802,
    "eval_rewards/margins": 0.09818949550390244,
    "eval_rewards/rejected": -0.17714762687683105,
    "eval_runtime": 103.7748,
    "eval_samples": 343,
    "eval_samples_per_second": 3.305,
    "eval_steps_per_second": 0.414
}