eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.22763729095459,
    "eval_logits/rejected": -2.222792387008667,
    "eval_logps/chosen": -34.090328216552734,
    "eval_logps/rejected": -37.59687423706055,
    "eval_loss": 0.6925650835037231,
    "eval_rewards/accuracies": 0.5211793780326843,
    "eval_rewards/chosen": -0.02788795717060566,
    "eval_rewards/margins": 0.01224011555314064,
    "eval_rewards/rejected": -0.04012807086110115,
    "eval_runtime": 145.6508,
    "eval_samples": 343,
    "eval_samples_per_second": 2.355,
    "eval_steps_per_second": 0.295
}