Frank
4b443c86d8
The qwen_image21 detection check and GGUF node arch list patch are applied by a startup wrapper (comfyui-start.sh) that runs before ComfyUI starts. The patch script and wrapper are baked into the image so they survive container recreations. The quadlet's Exec now points at the wrapper instead of main.py directly.
40 lines
2.2 KiB
Docker
40 lines
2.2 KiB
Docker
# ComfyUI on the gfx1151 (AMD Strix Halo / Radeon 8060S) ROCm torch build.
|
|
#
|
|
# Base image: AMD's gfx1151-tagged vLLM image. It is the only verified build
|
|
# where torch.cuda.is_available() is True on this APU (device capability
|
|
# (11,5) = gfx1151). It already ships the matching ROCm torch family
|
|
# (torch 2.9.1+rocm, torchvision 0.24.0+rocm, torchaudio 2.9.0+rocm), so we
|
|
# must NOT let pip reinstall them: ComfyUI's requirements.txt lists an unpinned
|
|
# `torch`, which would otherwise clobber the gfx1151 kernels with a default build.
|
|
FROM docker.io/rocm/vllm:rocm7.12.0_gfx1151_ubuntu24.04_py3.12_pytorch_2.9.1_vllm_0.16.0
|
|
|
|
# Bake the ComfyUI code into the image (rebuild the image to update the code).
|
|
RUN git clone --depth 1 https://github.com/comfyanonymous/ComfyUI /opt/ComfyUI
|
|
WORKDIR /opt/ComfyUI
|
|
|
|
# Pre-create the user-data dirs. The host mounts its collections over them
|
|
# (see the comfyui service in compose.yaml).
|
|
RUN mkdir -p /opt/ComfyUI/models /opt/ComfyUI/input /opt/ComfyUI/output /opt/ComfyUI/custom_nodes
|
|
|
|
# Install ComfyUI's dependencies, but skip the torch family (already present
|
|
# above as the gfx1151 ROCm builds). torchsde and the rest are kept.
|
|
RUN grep -viE '^(torch|torchvision|torchaudio)([^a-z]|$)' requirements.txt > /tmp/req-notorch.txt \
|
|
&& pip install --no-cache-dir -r /tmp/req-notorch.txt
|
|
|
|
# The city96/ComfyUI-GGUF custom node (loaded from the host's custom_nodes
|
|
# volume) needs the gguf library to decode GGUF-quantized diffusion models.
|
|
# sentencepiece and protobuf are its optional tokenizer deps. Baking them into
|
|
# the image keeps the dependency across container recreations — the node itself
|
|
# lives in the host volume, which persists, but the Python env does not.
|
|
RUN pip install --no-cache-dir "gguf>=0.13.0" sentencepiece protobuf
|
|
|
|
# Bake the qwen_image21 patch script into the image. It is run by the
|
|
# startup wrapper before ComfyUI starts, so the patch survives container
|
|
# recreations (the model_detection.py file is in the image, not a volume).
|
|
COPY apply_qwen21_patch.py /opt/ComfyUI/apply_qwen21_patch.py
|
|
COPY comfyui-start.sh /opt/ComfyUI/comfyui-start.sh
|
|
RUN chmod +x /opt/ComfyUI/comfyui-start.sh
|
|
|
|
CMD ["/opt/ComfyUI/comfyui-start.sh"]
|
|
EXPOSE 8188
|