[Linux] RX 6750 XT not detected

[josiah-roberts]

The docs suggest that AMD users can simply run the start script and things should work. However, my RX 6750 XT doesn't work with this - is ROCM not supported?

################################################################
Install script for stable-diffusion + Web UI
Tested on Debian 11 (Bullseye)
################################################################

################################################################
Running on trinity user
################################################################

################################################################
Repo already cloned, using it as install directory
################################################################

################################################################
Create and activate python venv
################################################################

################################################################
Launching launch.py...
################################################################
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
Commit hash: 0cc0ee1bcb4c24a8c9715f66cede06601bfc00c8
Traceback (most recent call last):
  File "/home/trinity/stable-diffusion-webui/launch.py", line 360, in <module>
    prepare_environment()
  File "/home/trinity/stable-diffusion-webui/launch.py", line 272, in prepare_environment
    run_python("import torch; assert torch.cuda.is_available(), 'Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check'")
  File "/home/trinity/stable-diffusion-webui/launch.py", line 129, in run_python
    return run(f'"{python}" -c "{code}"', desc, errdesc)
  File "/home/trinity/stable-diffusion-webui/launch.py", line 105, in run
    raise RuntimeError(message)
RuntimeError: Error running command.
Command: "/home/trinity/stable-diffusion-webui/venv/bin/python3" -c "import torch; assert torch.cuda.is_available(), 'Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check'"
Error code: 1
stdout: <empty>
stderr: /home/trinity/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/cuda/__init__.py:88: UserWarning: HIP initialization: Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice (Triggered internally at ../c10/hip/HIPFunctions.cpp:110.)
  return torch._C._cuda_getDeviceCount() > 0
Traceback (most recent call last):
  File "<string>", line 1, in <module>
AssertionError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

[fractal-fumbler]

gfx1031 isn't supported by rocm (yet?)

atm, there are 2 options

1. use HSA_OVERRIDE_GFX_VERSION=10.3.0 as env variable
2. compile pytorch locally for gfx1031

[rieve-nailo]

I'm guessing that you're like me and are normally a Windows user. While it is right, you do run the script and it goes just fine, it makes some assumptions that it doesn't tell you about. Mainly that you've installed AMD device drivers and ROCm drivers or something along those lines and restarted your system. At least that was the pain in the but that worked for me, and I had to go to AMD to get the process started as the drivers weren't in the regular repos that are included in the OS I'm using.

[elen07zz]

i have a rx 6700, but im using ubuntu and i'm using rocm. I think it's your better option.
as they told you before you have to use
HSA_OVERRIDE_GFX_VERSION=10.3.0

[Soulreaver90]

I followed the instructions and did not work at all. Unfortunately the instructions for AMD are vague and incomplete. They should include steps on how to install the ROCM drivers or point to the AMD page. Rebooting after installing the drivers is a must! Then you must include the HSA override or else it will default to the CPU. I’ve perfected the setup on my machine running Ubuntu 22.04 and might do a quick right up.

Edit: The script also did not correctly install the PyTorch+rocm drivers as stated, no idea why so it’s another must do if you wanna get it up and running.

[elen07zz]

Video tutorial
https://www.youtube.com/watch?v=6cTjF1OESGE

Just install rocm drivers

sudo apt-get update
wget https://repo.radeon.com/amdgpu-install/5.3/ubuntu/jammy/amdgpu-install_5.3.50300-1_all.deb
sudo apt-get install ./amdgpu-install_5.3.50300-1_all.deb
sudo amdgpu-install --usecase=rocm,hip,mllib --no-dkms
sudo usermod -a -G video,render $LOGNAME
sudo reboot

Then download and install python 3.10 Only if you didn't have it before.
and..

Next install automatic1111

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
cd stable-diffusion-webui
python -m venv venv
source venv/bin/activate
python -m pip install --upgrade pip wheel

TORCH_COMMAND='pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/rocm5.1.1' python launch.py 

It takes a while to download everything, Sometimes it seems that nothing is downloading because there is no percentage. Just leave it until it shows a message.

After installation you only need to run it with:

source venv/bin/activate
export HSA_OVERRIDE_GFX_VERSION=10.3.0
python3 launch.py --opt-split-attention-v1 --upcast-sampling --opt-sub-quad-attention --medvram