The training was entirely conducted using the images of Fluffy Tart(SD1.5 checkpoint) ,the purpose is to restore the style of this model.

I have always thought his style was good, but I can't figure out why it can't be trained into a Lora.

No test has been conducted to determine which weight is the most effective. It is recommended to appropriately reduce the weight when using it in combination with other LORA products.