all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.228447198867798,
    "eval_logits/rejected": -2.223632335662842,
    "eval_logps/chosen": -33.74469757080078,
    "eval_logps/rejected": -37.27058410644531,
    "eval_loss": 0.49989354610443115,
    "eval_rewards/accuracies": 0.5220099687576294,
    "eval_rewards/chosen": 0.002898545004427433,
    "eval_rewards/margins": 0.0004381774051580578,
    "eval_rewards/rejected": 0.002460367511957884,
    "eval_runtime": 145.5095,
    "eval_samples": 343,
    "eval_samples_per_second": 2.357,
    "eval_steps_per_second": 0.296,
    "train_loss": 0.39746343532165923,
    "train_runtime": 10799.7737,
    "train_samples": 3079,
    "train_samples_per_second": 1.14,
    "train_steps_per_second": 0.143
}