all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.649322271347046,
    "eval_logits/rejected": -2.6439428329467773,
    "eval_logps/chosen": -197.75340270996094,
    "eval_logps/rejected": -183.8756866455078,
    "eval_loss": 0.9304828643798828,
    "eval_rewards/accuracies": 0.582426905632019,
    "eval_rewards/chosen": -0.21480964124202728,
    "eval_rewards/margins": 0.08063079416751862,
    "eval_rewards/rejected": -0.2954404354095459,
    "eval_runtime": 443.5148,
    "eval_samples": 11765,
    "eval_samples_per_second": 26.527,
    "eval_steps_per_second": 3.317,
    "train_loss": 0.9410081987063832,
    "train_runtime": 92137.4772,
    "train_samples": 219411,
    "train_samples_per_second": 7.144,
    "train_steps_per_second": 0.112
}