Linux: AMD Radeon 9070XT : ComfyUI is starting and is trying to detect an nVidia GPU when I have installed all the AMD dependancies/setup. #10680
Description
Custom Node Testing
- I have tried disabling custom nodes and the issue persists (see how to disable custom nodes if you need help)
Expected Behavior
for the backend to start so that the front end can start so that I can use the program
Actual Behavior
The backend system is showing errors that imply the system is not finding an nVidia GPU (Which is correct there is only an AMG 9070XT on the PC), however, the system does not load the AMD GPU systems - even though I have downloaded/installed all the AMD specific systems listed on the 'how to install' on the ComyUI GIT page.
Steps to Reproduce
starting the app with : python main.py
Debug Logs
base ❯ cd ComfyUI (base)
python main.py
cd: The directory “ComfyUI” does not exist
Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "/home/mark/Apps/ComfyUI/ComfyUI/main.py", line 149, in <module>
import execution
File "/home/mark/Apps/ComfyUI/ComfyUI/execution.py", line 15, in <module>
import comfy.model_management
File "/home/mark/Apps/ComfyUI/ComfyUI/comfy/model_management.py", line 238, in <module>
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
~~~~~~~~~~~~~~~~^^
File "/home/mark/Apps/ComfyUI/ComfyUI/comfy/model_management.py", line 188, in get_torch_device
return torch.device(torch.cuda.current_device())
~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/home/mark/.local/lib/python3.13/site-packages/torch/cuda/__init__.py", line 1069, in current_device
_lazy_init()
~~~~~~~~~~^^
File "/home/mark/.local/lib/python3.13/site-packages/torch/cuda/__init__.py", line 410, in _lazy_init
torch._C._cuda_init()
~~~~~~~~~~~~~~~~~~~^^
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
[W1108 06:38:06.179817331 AllocatorConfig.cpp:28] Warning: PYTORCH_CUDA_ALLOC_CONF is deprecated, use PYTORCH_ALLOC_CONF instead (function operator())Other
before running ComyUI i have tried setting the following:
export ROCM_PATH=/opt/rocm
export HCC_AMDGPU_TARGET=gfx1201
export HSA_OVERRIDE_GFX_VERSION=10.0.0
But the result was the same.
Side note: I am running CachyOS, which is an optimised rolling release of Arch Linux, all the CUDA/Pytorch etc libraries are fully up to date.