Kicks in with weights at 0.6 and up to around 1.1

Lora doesn't need any keyword.

This is my first attempt and has been trained without texts, on a dataset of around 800 images