eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.8007283806800842,
    "eval_logits/rejected": -0.7758755683898926,
    "eval_logps/chosen": -1081.9208984375,
    "eval_logps/rejected": -1242.8665771484375,
    "eval_loss": 0.6785902976989746,
    "eval_rewards/accuracies": 0.6244725584983826,
    "eval_rewards/chosen": -0.003962064627557993,
    "eval_rewards/margins": 0.032243408262729645,
    "eval_rewards/rejected": -0.0362054705619812,
    "eval_runtime": 850.5886,
    "eval_samples": 7584,
    "eval_samples_per_second": 8.916,
    "eval_steps_per_second": 0.279
}