Stable-Diffusion-Webui-Docker

• Dockerized Stable-Diffusion-Webui based on AUTOMATIC1111/stable-diffusion-webui
• Includes prompt auto-completion javascript credited by shounksu
• Image is based on nvidia/cuda:11.7.1-devel-ubuntu22.04 image
• Prebuilt images are available on Docker Hub:
  • kestr3l/stable-diffusion-webui

0 Prequisites

• Linux environment (Including WSL2)
• 'Appropriate' GPU
• docker-ce
• docker-compose
• NVIDIA Container Toolkit
• Basic knowledges of Linux commands

1 stable-diffusion-webui

1.1 Building stable-diffusion-webui Docker Image

DOCKER_BUILDKIT=1 \
SD_WEBUI_VERSION=v1.6.0 && \
 docker build --no-cache \
--build-arg BASEIMAGE=nvidia/cuda \
--build-arg BASETAG=11.7.1-cudnn8-devel-ubuntu22.04 \
--build-arg SD_WEBUI_VERSION=${SD_WEBUI_VERSION} \
-t kestr3l/stable-diffusion-webui:${SD_WEBUI_VERSION} \
-f Dockerfile .

1.2 Running docker-stable-diffusion-webui Container

• Simply execute setup.sh.
• On the first run, it will create data directory for the WebUI on Documents directory and exit.
  • Run it again after then.
• If you want to run this on background, run by nohup ./setup.sh &

1.3 Debugging stable-diffusion-webui Docker Image

• Set DEBUG_MODE=1 on run.env
• This will make container do 'nothing'. So that user can try anything manually
• You can enter a container by docker exec -it --user user stable-diffusion-webui bash
• On default, sudo is available for the user.

2 Future Plans

• CI/CD for automated image update
• Image for training environment

3 References

1. AUTOMATIC1111/stable-diffusion-webui
2. facebookreseasrch/xformers
3. WebUI 태그 자동완성
4. 리눅스 xformers 빌드 방법 (GPU 불필요) (작성 중)
5. Is it possible to map a user inside the docker container to an outside user?