Running Stable Diffusion

Technical Topics Data Science, ML & AI
, , , ,
1

Hi, all. I’ve gotten Stable Diffusion running on a CPU using anaconda-project, which is provided by default when you install the Anaconda Distribution, to automate conda environment setup and launching of the Jupyter Notebook. Here’s my project

This project will work on Mac, Linux, and Windows. For me it takes between 10-15 minutes to generate one image on my M1 Pro MacBook. I’m curious to know what hardware you have and how long it takes.

3 Likes
2

I was able to save a few extra minutes (i.e. < 10mins) it using torch MPS backend on Apple M1 chip :slight_smile:

For anyone interested, It would just need a small update to the installed package, as well as targeting the mps device for the model.
Namely replacing the pipe object with:
pipe=StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5", low_cpu_mem_usage=False).to('mps')