eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.4874803125858307,
    "eval_logits/rejected": -0.5277639627456665,
    "eval_logps/chosen": -418.41363525390625,
    "eval_logps/rejected": -439.39300537109375,
    "eval_loss": 0.5796785354614258,
    "eval_rewards/accuracies": 0.716269850730896,
    "eval_rewards/chosen": -0.7179543972015381,
    "eval_rewards/margins": 0.5342229008674622,
    "eval_rewards/rejected": -1.2521772384643555,
    "eval_runtime": 305.3815,
    "eval_samples": 2000,
    "eval_samples_per_second": 6.549,
    "eval_steps_per_second": 0.206
}