all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2437870502471924,
    "eval_logits/rejected": -2.2389047145843506,
    "eval_logps/chosen": -34.03290939331055,
    "eval_logps/rejected": -37.49677658081055,
    "eval_loss": 0.5752949118614197,
    "eval_rewards/accuracies": 0.45764121413230896,
    "eval_rewards/chosen": 0.001152246491983533,
    "eval_rewards/margins": -0.012740416452288628,
    "eval_rewards/rejected": 0.013892664574086666,
    "eval_runtime": 145.7802,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.5317366457604743,
    "train_runtime": 3254.062,
    "train_samples": 3079,
    "train_samples_per_second": 0.946,
    "train_steps_per_second": 0.118
}