eval_results.json

{
    "epoch": 0.9230769230769231,
    "eval_logits/chosen": 13.998739242553711,
    "eval_logits/rejected": 12.849632263183594,
    "eval_logps/chosen": -396.9144287109375,
    "eval_logps/rejected": -395.41522216796875,
    "eval_loss": 0.6937551498413086,
    "eval_rewards/accuracies": 0.5357142686843872,
    "eval_rewards/chosen": -0.00012085953494533896,
    "eval_rewards/margins": 9.88479132502107e-06,
    "eval_rewards/rejected": -0.00013074450544081628,
    "eval_runtime": 89.0222,
    "eval_samples": 1000,
    "eval_samples_per_second": 11.233,
    "eval_steps_per_second": 0.708
}