YanWenKun/ComfyUI-Docker

Provides multi-variant Docker images optimized for different GPU architectures (NVIDIA CUDA 12.6–13.0, AMD ROCm, Intel XPU) and use cases, from minimal "slim" distributions with ComfyUI-Manager to "megapak" bundles pre-loaded with dozens of custom nodes and development tools. Uses volume mounts to separate persistent model caches (Hugging Face, PyTorch), user workflows, and input/output directories, enabling seamless model management across container restarts while maintaining CUDA toolchain compatibility with PyTorch's build constraints.

1,413 stars. Actively maintained with 106 commits in the last 30 days.