The finetuning configuration is not provided in paper, could you plz provide some more details here? Is the finetuning use QLoRA provided by peft?