all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.4954090118408203,
    "eval_logits/rejected": -2.477094888687134,
    "eval_logps/chosen": -81.00131225585938,
    "eval_logps/rejected": -90.83526611328125,
    "eval_loss": 0.7656615972518921,
    "eval_rewards/accuracies": 0.3194444477558136,
    "eval_rewards/chosen": -1.753847360610962,
    "eval_rewards/margins": 0.8612387776374817,
    "eval_rewards/rejected": -2.615086078643799,
    "eval_runtime": 113.6621,
    "eval_samples": 2000,
    "eval_samples_per_second": 17.596,
    "eval_steps_per_second": 0.554,
    "total_flos": 0.0,
    "train_loss": 0.4762688593731986,
    "train_runtime": 3179.8752,
    "train_samples": 6113,
    "train_samples_per_second": 5.767,
    "train_steps_per_second": 0.091
}