all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.220597267150879,
    "eval_logits/rejected": -2.2157645225524902,
    "eval_logps/chosen": -34.07185745239258,
    "eval_logps/rejected": -37.57231903076172,
    "eval_loss": 0.6930569410324097,
    "eval_rewards/accuracies": 0.5161960124969482,
    "eval_rewards/chosen": -0.00037304984289221466,
    "eval_rewards/margins": 0.00018396865925751626,
    "eval_rewards/rejected": -0.00055701844394207,
    "eval_runtime": 145.5452,
    "eval_samples": 343,
    "eval_samples_per_second": 2.357,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.5396727961379212,
    "train_runtime": 10793.7948,
    "train_samples": 3079,
    "train_samples_per_second": 1.141,
    "train_steps_per_second": 0.143
}