SD models are much smaller so you can probably already run it with the github code.

Image models work well at least: https://huggingface.co/docs/diffusers/optimization/mps

It is even possible to run them in the browser: https://stablediffusionweb.com/