all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.226093053817749,
    "eval_logits/rejected": -2.221263885498047,
    "eval_logps/chosen": -34.05622100830078,
    "eval_logps/rejected": -37.58139419555664,
    "eval_loss": 2495.922119140625,
    "eval_rewards/accuracies": 0.529900312423706,
    "eval_rewards/chosen": -0.00021668746194336563,
    "eval_rewards/margins": 0.00043102249037474394,
    "eval_rewards/rejected": -0.0006477099377661943,
    "eval_runtime": 145.39,
    "eval_samples": 343,
    "eval_samples_per_second": 2.359,
    "eval_steps_per_second": 0.296,
    "train_loss": 1756.4536297686689,
    "train_runtime": 10797.265,
    "train_samples": 3079,
    "train_samples_per_second": 1.141,
    "train_steps_per_second": 0.143
}