all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.0955071449279785,
    "eval_logits/rejected": -2.0908851623535156,
    "eval_logps/chosen": -33.783447265625,
    "eval_logps/rejected": -37.54340744018555,
    "eval_loss": 0.49944302439689636,
    "eval_rewards/accuracies": 0.6038206219673157,
    "eval_rewards/chosen": 0.0025110626593232155,
    "eval_rewards/margins": 0.0027789229061454535,
    "eval_rewards/rejected": -0.00026786039234139025,
    "eval_runtime": 145.762,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.39746343532165923,
    "train_runtime": 10799.7737,
    "train_samples": 3079,
    "train_samples_per_second": 1.14,
    "train_steps_per_second": 0.143
}