eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.5171043872833252,
    "eval_logits/rejected": -1.2001606225967407,
    "eval_logps/chosen": -377.12396240234375,
    "eval_logps/rejected": -598.4039916992188,
    "eval_loss": 0.03705598786473274,
    "eval_rewards/accuracies": 0.9856902360916138,
    "eval_rewards/chosen": 4.279876708984375,
    "eval_rewards/margins": 8.16864013671875,
    "eval_rewards/rejected": -3.888763189315796,
    "eval_runtime": 560.6128,
    "eval_samples": 9500,
    "eval_samples_per_second": 16.946,
    "eval_steps_per_second": 0.53
}