ComfyUI-Docker

Why we included this project

Getting ComfyUI running on your own GPU usually means wrestling with Python environments and driver versions. This project packages the node-based image-generation interface into Docker images that cover the common hardware cases: CUDA builds for NVIDIA cards, ROCm for AMD, an XPU tag for Intel, and a CPU-only fallback, so you can grab the one that matches the machine in front of you. The repository also spells out the CUDA compatibility limits imposed by the PyTorch toolchain, which saves real time when you are deciding which tag to pull. The slim images ship ComfyUI plus the Manager plugin and keep the dependencies needed for later custom-node installs, while the megapak bundles add development kits and dozens of preinstalled nodes. If you want a reproducible, containerized setup instead of a hand-built Python environment, this is a solid place to start.