all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.4388577938079834,
    "eval_logits/rejected": -1.448164463043213,
    "eval_logps/chosen": -1.648034930229187,
    "eval_logps/rejected": -2.2085607051849365,
    "eval_loss": 1.3243666887283325,
    "eval_rewards/accuracies": 0.8353658318519592,
    "eval_rewards/chosen": -16.480350494384766,
    "eval_rewards/margins": 5.605257034301758,
    "eval_rewards/rejected": -22.08560562133789,
    "eval_runtime": 87.6562,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.371,
    "eval_steps_per_second": 1.403,
    "total_flos": 0.0,
    "train_loss": 1.8032665589636858,
    "train_runtime": 11474.0462,
    "train_samples": 59876,
    "train_samples_per_second": 5.218,
    "train_steps_per_second": 0.041
}