Help Needed: Setting Up SDXL LoRA Training on Google Cloud (non-coder, AI interior design project)

Hi everyone,

I’m working on a special AI interior design app based on my own design styles. I’ve already prepared everything — my image-caption datasets (800+ images per style + text files), training scripts (chat gpt) and I’d like to train on Google Cloud Vertex AI Workbench.

But I’m not a coder, and after many hours of trying, I just can’t seem to get LoRA SDXL training to work. I’ve tried different scripts and setups, but something always breaks.

If anyone here has 10–15 minutes to spare, I would be deeply grateful for a bit of help — maybe even a quick Zoom or screen share — just to guide me through setting up the environment properly so I can begin training.

I’m very passionate about this and happy to share more details about the project. Ideally, I’d love to train using the Diffusers + PEFT setup or anything that gets good results.

Thank you so much in advance to anyone kind enough to help.

Warm regards,
Ivana

1 Like

Diffusers+PEFT itself is an environment designed for coders…

For SDXL LoRA training, it’s best to try KohyaSS first. Diffusers are used internally. There seem to be several other good scripts also available. (One of tutorials)

In any case, you won’t really understand it until you get it working properly. Try it out on Colab Free or something similar.

For tips on training style LoRA, online resources are a good reference.

1 Like