all_results.json

{
    "epoch": 2.67,
    "eval_logits/chosen": -0.009816838428378105,
    "eval_logits/rejected": 0.08653248846530914,
    "eval_logps/chosen": -306.496337890625,
    "eval_logps/rejected": -278.6934509277344,
    "eval_loss": 0.001210334594361484,
    "eval_rewards/accuracies": 0.4779999852180481,
    "eval_rewards/chosen": -0.0006128409877419472,
    "eval_rewards/margins": -0.00028469687094911933,
    "eval_rewards/rejected": -0.00032814397127367556,
    "eval_runtime": 418.4862,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.779,
    "eval_steps_per_second": 1.195,
    "train_loss": 0.00033688666007947177,
    "train_runtime": 14.5188,
    "train_samples": 61135,
    "train_samples_per_second": 2.755,
    "train_steps_per_second": 0.276
}