all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.8344432711601257,
    "eval_logits/rejected": -0.6401407718658447,
    "eval_logps/chosen": -361.0932312011719,
    "eval_logps/rejected": -682.3729858398438,
    "eval_loss": 0.46492835879325867,
    "eval_rewards/accuracies": 0.9166666865348816,
    "eval_rewards/chosen": -0.1737493872642517,
    "eval_rewards/margins": 0.431477814912796,
    "eval_rewards/rejected": -0.6052272319793701,
    "eval_runtime": 12.8431,
    "eval_samples": 96,
    "eval_samples_per_second": 7.475,
    "eval_steps_per_second": 0.234,
    "train_loss": 0.4338115497871681,
    "train_runtime": 2557.4167,
    "train_samples": 8634,
    "train_samples_per_second": 3.376,
    "train_steps_per_second": 0.053
}