My first architecture LoRA was trained on images from Pinterest (most of them low quality). However, it generated what I needed for my work.
I would really appreciate your advice on LoRA training. Additionally, I'm planning to create an architecture checkpoint for SDXL.