eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.4944536685943604,
    "eval_logits/rejected": -2.4963433742523193,
    "eval_logps/chosen": -269.25555419921875,
    "eval_logps/rejected": -253.21238708496094,
    "eval_loss": 0.6399702429771423,
    "eval_rewards/accuracies": 0.6370000243186951,
    "eval_rewards/chosen": 0.030110126361250877,
    "eval_rewards/margins": 0.05743245780467987,
    "eval_rewards/rejected": -0.027322327718138695,
    "eval_runtime": 803.6977,
    "eval_samples": 2000,
    "eval_samples_per_second": 2.488,
    "eval_steps_per_second": 0.311
}