all_results.json

{
    "epoch": 2.67,
    "eval_logits/chosen": 0.8310254812240601,
    "eval_logits/rejected": 0.8885230422019958,
    "eval_logps/chosen": -256.6099853515625,
    "eval_logps/rejected": -233.54428100585938,
    "eval_loss": 0.0004723873280454427,
    "eval_rewards/accuracies": 0.49950000643730164,
    "eval_rewards/chosen": 4.191549669485539e-05,
    "eval_rewards/margins": 0.0001021110438159667,
    "eval_rewards/rejected": -6.019552893121727e-05,
    "eval_runtime": 416.5778,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.801,
    "eval_steps_per_second": 1.2,
    "train_loss": 0.00021988208027323708,
    "train_runtime": 13.6913,
    "train_samples": 61135,
    "train_samples_per_second": 2.922,
    "train_steps_per_second": 0.292
}