all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": 0.6308508515357971,
    "eval_logits/rejected": 0.7334519028663635,
    "eval_logps/chosen": -285.9486999511719,
    "eval_logps/rejected": -269.1014404296875,
    "eval_loss": 0.6537346243858337,
    "eval_rewards/accuracies": 0.6579999923706055,
    "eval_rewards/chosen": -0.2570453882217407,
    "eval_rewards/margins": 0.11960798501968384,
    "eval_rewards/rejected": -0.37665337324142456,
    "eval_runtime": 325.7825,
    "eval_samples": 2000,
    "eval_samples_per_second": 6.139,
    "eval_steps_per_second": 0.384,
    "train_loss": 0.6680307045922589,
    "train_runtime": 18174.3674,
    "train_samples": 30567,
    "train_samples_per_second": 3.364,
    "train_steps_per_second": 0.052
}