all_results.json

{
    "epoch": 2.983957219251337,
    "eval_logits/chosen": -2.2057065963745117,
    "eval_logits/rejected": -2.149101972579956,
    "eval_logps/chosen": -248.10618591308594,
    "eval_logps/rejected": -244.79750061035156,
    "eval_loss": 0.5014801621437073,
    "eval_rewards/accuracies": 0.8273809552192688,
    "eval_rewards/chosen": 0.29323506355285645,
    "eval_rewards/margins": 2.8863983154296875,
    "eval_rewards/rejected": -2.593163251876831,
    "eval_runtime": 179.2613,
    "eval_samples_per_second": 14.839,
    "eval_steps_per_second": 0.234,
    "total_flos": 3289753017384960.0,
    "train_loss": 0.3353286673945765,
    "train_runtime": 10412.0978,
    "train_samples_per_second": 6.896,
    "train_steps_per_second": 0.027
}