all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 0.41779446601867676,
    "eval_logits/rejected": 1.454390525817871,
    "eval_logps/chosen": -324.46234130859375,
    "eval_logps/rejected": -342.5584716796875,
    "eval_loss": 0.11311393231153488,
    "eval_rewards/accuracies": 0.7543103694915771,
    "eval_rewards/chosen": -0.393722802400589,
    "eval_rewards/margins": 0.566637396812439,
    "eval_rewards/rejected": -0.9603601098060608,
    "eval_runtime": 93.576,
    "eval_samples": 1818,
    "eval_samples_per_second": 19.428,
    "eval_steps_per_second": 0.31,
    "total_flos": 0.0,
    "train_loss": 0.13197535372108493,
    "train_runtime": 11523.0815,
    "train_samples": 55758,
    "train_samples_per_second": 4.839,
    "train_steps_per_second": 0.038
}