all_results.json

{
    "epoch": 2.994495412844037,
    "eval_logits/chosen": -0.02289891056716442,
    "eval_logits/rejected": -0.9417744874954224,
    "eval_logps/chosen": -300.2835388183594,
    "eval_logps/rejected": -281.21466064453125,
    "eval_loss": 0.36351555585861206,
    "eval_rewards/accuracies": 0.9777777791023254,
    "eval_rewards/chosen": -0.15967033803462982,
    "eval_rewards/margins": 0.9239291548728943,
    "eval_rewards/rejected": -1.08359956741333,
    "eval_runtime": 8.9717,
    "eval_samples_per_second": 5.016,
    "eval_steps_per_second": 5.016,
    "total_flos": 7.837376281021809e+17,
    "train_loss": 0.514469311225648,
    "train_runtime": 8073.8639,
    "train_samples_per_second": 1.619,
    "train_steps_per_second": 0.051
}