openeuler/comfyui repository overview

⁠Quick reference

• The official ComfyUI Ascend docker images

• Maintained by: openEuler CloudNative SIG⁠

• Where to get help: openEuler CloudNative SIG⁠, openEuler⁠

⁠ComfyUI | openEuler

Current ComfyUI docker images are built on the openEuler⁠. This repository is free to use and exempted from per-user rate limits.

ComfyUI⁠ is an open-source node-based interface for diffusion models. It provides a modular graph editor to assemble image, video and audio generation workflows, and exposes the same workflows through an HTTP API on port 8188.

This image ships ComfyUI with torch_npu⁠ so that diffusion models run seamlessly on the Ascend NPU: ComfyUI detects torch_npu automatically and selects the npu device, no extra launch arguments are required. The software stack is validated on a single Ascend 910B4 card with CANN 9.1.0, torch==2.9.0 and torch_npu==2.9.0.post8 on Python 3.12, following the Ascend quick start guard⁠.

Read more about Ascend at hiascend.com⁠ and explore the ComfyUI documentation at docs.comfy.org⁠.

⁠Supported tags and respective Dockerfile links

The tag of each ComfyUI docker image consists of the ComfyUI version and the version of the software stack. The details are as follows

⁠Usage

⁠Quick start 1: supported devices

• Atlas A2 Training series (Atlas 800T A2, Atlas 900 A2 PoD, Atlas 200T A2 Box16, Atlas 300T A2)

• Atlas 800I A2 Inference series (Atlas 800I A2)

⁠Quick start 2: setup environment using container

# Update DEVICE according to your device (/dev/davinci[0-7])
export DEVICE=/dev/davinci0
# Update the comfyui image
export IMAGE=openeuler/comfyui:0.44.0-torch_npu2.9.0-cann9.1.0-python3.12-oe2403sp4
docker run --rm \
--name comfyui \
--device $DEVICE \
--device /dev/davinci_manager \
--device /dev/devmm_svm \
--device /dev/hisi_hdc \
-v /usr/local/dcmi:/usr/local/dcmi \
-v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
-v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
-v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
-v /etc/ascend_install.info:/etc/ascend_install.info \
-v /root/.cache:/root/.cache \
-p 8188:8188 \
-d $IMAGE

The interactive web UI is now available at http://<host>:8188.

⁠Quick start 3: text-to-image on Ascend NPU

Validate the environment inside the container:

docker exec comfyui python3 -c "import torch, torch_npu; print('torch=', torch.__version__); print('torch_npu=', torch_npu.__version__); print('is_available:', torch.npu.is_available()); print('count:', torch.npu.device_count())"

Download the Stable Diffusion 1.5 checkpoint from ModelScope, link it into models/checkpoints/ and run the upstream API example (default text-to-image workflow: SD 1.5, 512x512, 20 steps euler):

docker exec -it comfyui bash

python3 -c "from modelscope import snapshot_download; print(snapshot_download('AI-ModelScope/stable-diffusion-v1-5', revision='master', allow_patterns=['v1-5-pruned-emaonly.safetensors']))"
ln -sf /root/.cache/modelscope/hub/models/AI-ModelScope/stable-diffusion-v1-5/v1-5-pruned-emaonly.safetensors models/checkpoints/
python3 script_examples/basic_api_example.py
ls output/ComfyUI_*.png

The generated image is written to output/ComfyUI_00001_.png. Confirm that ComfyUI picked the NPU device:

python3 -c "import comfy.model_management as mm; print('comfy device:', mm.get_torch_device())"

For faster model download in China, use the ModelScope channel (already installed in this image).

⁠Question and answering

If you have any questions or want to use some special features, please submit an issue or a pull request on openeuler-docker-images⁠.