all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -3.02286958694458,
    "eval_logits/rejected": -3.0082170963287354,
    "eval_logps/chosen": -781.7958984375,
    "eval_logps/rejected": -817.7778930664062,
    "eval_loss": 0.7211850881576538,
    "eval_rewards/accuracies": 0.57421875,
    "eval_rewards/chosen": -2.3879425525665283,
    "eval_rewards/margins": 0.11641353368759155,
    "eval_rewards/rejected": -2.5043559074401855,
    "eval_runtime": 133.2747,
    "eval_samples": 1000,
    "eval_samples_per_second": 7.503,
    "eval_steps_per_second": 0.24,
    "train_loss": 0.5059080181213526,
    "train_runtime": 5445.7158,
    "train_samples": 19996,
    "train_samples_per_second": 3.672,
    "train_steps_per_second": 0.057
}