all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.226956844329834,
    "eval_logits/rejected": -2.2221333980560303,
    "eval_logps/chosen": -34.03876876831055,
    "eval_logps/rejected": -37.55620193481445,
    "eval_loss": 0.6902801394462585,
    "eval_rewards/accuracies": 0.5519103407859802,
    "eval_rewards/chosen": -0.002528026234358549,
    "eval_rewards/margins": 0.021220751106739044,
    "eval_rewards/rejected": -0.02374877780675888,
    "eval_runtime": 145.6815,
    "eval_samples": 343,
    "eval_samples_per_second": 2.354,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.6364745784115482,
    "train_runtime": 3251.6479,
    "train_samples": 3079,
    "train_samples_per_second": 0.947,
    "train_steps_per_second": 0.118
}