all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.3485426902770996,
    "eval_logits/rejected": -2.32399582862854,
    "eval_logps/chosen": -207.17623901367188,
    "eval_logps/rejected": -242.1480255126953,
    "eval_loss": 0.45268043875694275,
    "eval_rewards/accuracies": 0.8165322542190552,
    "eval_rewards/chosen": -0.6856781244277954,
    "eval_rewards/margins": 3.7083182334899902,
    "eval_rewards/rejected": -4.393996715545654,
    "eval_runtime": 246.3688,
    "eval_samples_per_second": 15.919,
    "eval_steps_per_second": 0.252,
    "total_flos": 4881795388538880.0,
    "train_loss": 0.21833302825689316,
    "train_runtime": 16364.2441,
    "train_samples_per_second": 6.47,
    "train_steps_per_second": 0.025
}