cancel
Showing results for 
Search instead for 
Did you mean: 

AI Discussions

sasoriza
Journeyman III

Looking for an up-to-date guide on installing Automatic1111 with directml and support for AMD GPU

I have tried various guides so far, including the Updated How To (which is outdated) here (https://community.amd.com/t5/ai/updated-how-to-running-optimized-automatic1111-stable-diffusion/ba-p...) and the official instructions on https://github.com/lshqqytiger/stable-diffusion-webui-directml but keep running into problems. Even with various extra steps of installing requirements manually, I can never get it to run without having to add --skip-torch-cuda-test which kinda defeats the whole process of these AMD GPU workarounds.
Has anybody gotten the current stable-diffusion-webui-directml release running with actual GPU support and decent image generation speed? What is the recommended way? Is installing it on windows even recommended or would running it on a virtual machine via wsl2 make more sense or would I run into similar problems when trying to use my GPU for image generation?
Thank you in advance!

2 Replies

I am having a hard time figuring out, which steps to follow from which guide and what repos to use. No matter what I try, eventually I will run into an error.

Using this guide https://community.amd.com/t5/ai/how-to-running-optimized-automatic1111-stable-diffusion-webui-on/ba-... to optimize the models results in an error and it seems both of the other guides rely on these optimized models. Also, all three guides are outdated and the current stable-diffusion-webui-directml release seems to suffer from a bug so while I appreciate the help, sadly I was not able to get it to work.

Donased
Journeyman III

I know I'm a bit late to the party but in case anyone sees this. I finally got it to run. Go here and scroll down to Vexillens comment. It shows how to get the Torch stuff to install. 
https://github.com/lshqqytiger/stable-diffusion-webui-directml/issues/351

0 Likes