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
