Setup ComfyUI on Arch with AMD Support
Follow these steps to install ComfyUI on an Arch Linux system with AMD GPU support using ROCm (Radeon Open Compute).
Step-by-Step Guide
1. Install ROCm Base Components
Install essential ROCm libraries for GPU acceleration:
sudo pacman -S rocm-core rocm-toolchain rocm-hip-runtime rocm-hip-sdk rocblas rocfft rocthrust rccl hipfort magma-hip onnxruntime-rocm
Note: Ensure all packages are available in the official Arch repositories. If missing, consider using AUR (Arch User Repository) for some components.
Python Setup
1. Create a Folder
mkdir -p ~/Python/comfy
2. Enter the Folder
cd ~/Python/comfy
3. Install Python and pip
Install Python 3 and pip:
sudo pacman -S python python-pip
Note: If using a newer Arch system, python refers to Python 3 by default.
4. Create a Virtual Environment
python -m venv ./venv
5. Activate the Virtual Environment
source ./venv/bin/activate
ComfyUI Setup
1. Install Comfy CLI
pip install comfy-cli
2. Install ComfyUI
Run the installation command:
comfy install
Note: Ensure you're in the ~/Python/comfy directory when running this command.
PyTorch for ROCm
- Visit the official PyTorch ROCm page to get the correct bash command: https://pytorch.org/blog/pytorch-for-amd-rocm-platform-now-available-as-python-package/
- Example command for ROCm 4.x (update version as needed):
pip install torch -f https://download.pytorch.org/whl/rocm4.1.0/torch_stable.html pip install ninja pip install 'git+https://github.com/pytorch/vision.git@v0.9.0'
Note: Replace rocm4.1.0 with the latest version available from PyTorch's ROCm release page.
Final Checks
- Ensure all packages are correctly installed in your virtual environment.
- Verify ROCm and PyTorch compatibility by checking:
torch --version rocminfo
Key Notes
- ROCm Compatibility: Confirm that your AMD GPU is supported by ROCm (e.g., RDNA2 or newer).
- PyTorch Version: Use the latest stable version compatible with your ROCm release.
- Virtual Environment: Always use a virtual environment to avoid conflicts.