update: for people who are waiting on windows, it is unlikely they will support older versions, and the probability of the rest on the list at windows support listed being supported is slim, because they are gonna drop rocm in 2-3 years when they release the 8000 series.
note: even more reason to use linux over windows if intent is mostly AI: https://rocm.docs.amd.com/projects/install-on-windows/en/latest/reference/component-support.html
doesnt have hipify or AI pytorch support on their windows ROCM
to start:
zypper ar -f -p 100 obs://home:simonizor:zow home_simonizor_zow
zypper ref home_simonizor_zow
zypper in zow
zow install git python3 python-pip wget make perl-URI-Encode
^ all dependencies next do:
https://rocm.docs.amd.com/projects/install-on-linux/en/latest/how-to/native-install/sle.html
sudo zypper ref
sudo zypper install rocm-hip-libraries
now go open YaST and go to software> scroll down to user group and management> select that > groups tab > set filter top right > system groups edit the video and render groups to add yourself to them
Edit ~/.profile with your editor of choice.
sudo nano ~/.profile
OR
sudo nano /etc/environment
the difference is environment applies on boot whereas the profile applies to specified user. note to remove the "export" word if you put in environment.
Paste the following line at the bottom of the file, then press ctrl-x and save the file before exiting.
export HSA_OVERRIDE_GFX_VERSION=10.3.0
(change the version number to 11.0.0 for 7000 series)
Now make sure to restart your computer before continuing.
after reboot
do this in either home folder or you can do:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
cd stable-diffusion-webui
before hand to skip some pain in having to reinstall the pytorch
now activate the venv using
python3 -m venv venv
. venv/bin/activate
you will now see "(venv)" beside the front of console
we want to seperate these two commands so that it will install in necessary order for it to install "WHL"
pip3 install wheel
THEN
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.7
or
pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm6.0
Let's verify PyTorch was installed correctly with GPU support, so lets first enter the Python console.
python3
Now enter the following two lines of code. If it returns True then everything was installed correctly.
import torch
torch.cuda.is_available()
Then enter exit() to exit the Python console.
If you see a warning mentioning “Cannot locate TCMalloc (improves CPU memory usage)”, then you can improve performance by installing the following package before running the application again:
zow install gperftools
after that finishes run webui.sh remember first gen takes a minute
Note: if you get this error:
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "/home/ellia/stable-diffusion-webui/venv/lib/python3.11/site-packages/cv2/qt/plugins" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.
Available platform plugins are: xcb, eglfs, linuxfb, minimal, minimalegl, offscreen, vnc, wayland-egl, wayland, wayland-xcomposite-egl, wayland-xcomposite-glx.
/usr/bin/xdg-open: line 634: 12402 Aborted (core dumped) kde-open${KDE_SESSION_VERSION} "$1"
you need to do 1 thing
1: go to stable-diffusion-webui/venv/lib/python3.11/site-packages/cv2/qt/plugins/platforms/ and delete the xcb file there and re open webui.sh
If you would like to help bring xformers to AMD please refer to this issue and participate testing, make sure to include your specs https://github.com/facebookresearch/xformers/pull/978