tinyllava/training_recipe/lora_recipe.py使用lora时，并没有真正使用lora

[LuShuaie]

tinyllava/training_recipe/lora_recipe.py中

if model.peft_config is None:
log("Adding LoRA adapters...")
model = get_peft_model(model, lora_config)

这个会直接跳过吧，应该把if model.peft_config is None:进行注销