all_results.json

{
    "epoch": 2.9906542056074765,
    "eval_logits/chosen": -2.3522794246673584,
    "eval_logits/rejected": -2.310214042663574,
    "eval_logps/chosen": -268.56451416015625,
    "eval_logps/rejected": -235.67572021484375,
    "eval_loss": 0.5341237187385559,
    "eval_rewards/accuracies": 0.7951388955116272,
    "eval_rewards/chosen": 1.2511848211288452,
    "eval_rewards/margins": 2.614333152770996,
    "eval_rewards/rejected": -1.3631482124328613,
    "eval_runtime": 150.6354,
    "eval_samples_per_second": 15.136,
    "eval_steps_per_second": 0.239,
    "total_flos": 2829829665718272.0,
    "train_loss": 0.33247024814287823,
    "train_runtime": 8773.941,
    "train_samples_per_second": 7.014,
    "train_steps_per_second": 0.027
}