all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.0034948738757520914,
    "eval_logits/rejected": -0.007565807551145554,
    "eval_logps/chosen": -2.20627498626709,
    "eval_logps/rejected": -60.60332489013672,
    "eval_loss": 0.15350937843322754,
    "eval_rewards/accuracies": 0.9610000848770142,
    "eval_rewards/chosen": 17.28226661682129,
    "eval_rewards/margins": 5.981877326965332,
    "eval_rewards/rejected": 11.30038833618164,
    "eval_runtime": 361.5652,
    "eval_samples_per_second": 2.766,
    "eval_steps_per_second": 0.346,
    "total_flos": 1.4338459346927616e+18,
    "train_loss": 0.23597783709896936,
    "train_runtime": 13850.7044,
    "train_samples_per_second": 0.65,
    "train_steps_per_second": 0.081
}