eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.539411723613739,
    "eval_logits/rejected": 0.5653398036956787,
    "eval_logps/chosen": -192.1762237548828,
    "eval_logps/rejected": -213.7832794189453,
    "eval_loss": 0.5578895807266235,
    "eval_rewards/accuracies": 0.7466216087341309,
    "eval_rewards/chosen": -0.42640554904937744,
    "eval_rewards/margins": 0.47774478793144226,
    "eval_rewards/rejected": -0.9041503667831421,
    "eval_runtime": 231.2679,
    "eval_samples": 2305,
    "eval_samples_per_second": 9.967,
    "eval_steps_per_second": 0.16
}