eval_results.json

{
    "epoch": 0.9990049751243781,
    "eval_logits/chosen": -0.9594090580940247,
    "eval_logits/rejected": -0.8731673955917358,
    "eval_logps/chosen": -600.8353271484375,
    "eval_logps/rejected": -547.457763671875,
    "eval_loss": 0.3535589277744293,
    "eval_rewards/accuracies": 0.9395522475242615,
    "eval_rewards/chosen": 0.11425414681434631,
    "eval_rewards/margins": 1.0437376499176025,
    "eval_rewards/rejected": -0.9294834136962891,
    "eval_runtime": 710.4378,
    "eval_samples": 5359,
    "eval_samples_per_second": 7.543,
    "eval_steps_per_second": 0.472
}