RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS on AMD GPU

[vKireina]

Context: I was originally using version 1.6.0 of Automatic1111 just fine. Decided should update since there has been many. Updated to latest version and now get this error. (Yes I followed the AMD Install directions).

I was also previously using version 3.10.11 of Python from the Microsoft Store which worked fine before. I did already try installing 3.10.6 and made no difference.

Specs:
GPU - AMD Radeon RX 6900 XT
CPU - AMD Ryzen 9 5900 12-Core Processor - 12 Cores
RAM - 32 GB
VRAM - 16368 MB - GDDR6 2000 MHz
Memory Type - GDDR6
Driver Version - 23.40.31.05-240417a-402338C-AMD-Software-Adrenalin-Edition
AMD Windows Driver Version - 31.0.24031.5001
AMD Audio Driver Version - 10.0.1.30
Windows Edition - Windows 11 Core (64 bit)
Windows Version - 23H2

I have already tried deleting everything and starting from scratch, I have tried adding arguments, I have tried downgrading back to 1.6.0, I have even tried the SD Forge webui which supposedly doesnt matter what GPU you have. All result in same error.

I followed these following video guides exactly and did a fresh install from scratch for each one I attempted in this order:

1. https://youtu.be/eO88i8o-BoY?si=rEZ1EzVuDrMT1e2G
2. https://youtu.be/8xR0vms0e0U?si=XiLYcUQZbEj8kctc
3. https://youtu.be/TJ98hAIN5io?si=IzJRhSjmvz59PZxG
4. https://youtu.be/mHvnHA04sbc?si=SsProDPebDvIbfDX
5. https://youtu.be/PdkBmDoUIy8?si=I7Tk3w8JmliqzzQp

Im seriously at a lost of what to do now. No matter what I do it just doesnt work despite following all these videos and guides meant specifically for AMD users. Any help would be appreciated.

[hqnicolas]

Hello @vKireina Today I have an REPO that solves AMD ROCm problems with docker containers
https://github.com/hqnicolas/StableDiffusionROCm
it was stable at 1.9.3 (latest)
if you like this automation REPO please let a Star on it ⭐

[gudo09]

is just for linux, right?

[hqnicolas]

You can use docker on Windows and MacOs, but the commands will need to be translated.
Also Debian Runs Faster for ROCm cards