Stable Diffusion WebUI in Docker
This builds a Docker image containing the Stable Diffusion WebUI by AUTOMATIC111 from https://github.com/AUTOMATIC1111/stable-diffusion-webui
Build the image (and optionally push it to your registry)
Set the version and image name to use:
export VERSION="v1.x.y"
export IMAGE_NAME=registry.example.com/stable-diffusion-webuiBuild the image and also add the latest tag:
docker build --build-arg VERSION=${VERSION} -t ${IMAGE_NAME}:${VERSION} .
docker tag ${IMAGE_NAME}:${VERSION} ${IMAGE_NAME}:latestOptionally, push both tags to your registry:
docker push ${IMAGE_NAME}:${VERSION}
docker push ${IMAGE_NAME}:latestBuilding an OCI image
nerdctl build --progress=plain --build-arg="VERSION=${VERSION}" --output="type=oci,name=${IMAGE_NAME}:${VERSION},push=true" -t ${IMAGE_NAME}:${VERSION} .Run the container
Run the image without retaining any data or outputs on exit:
docker run --gpus all --rm --name stable-diffusion -p 7860:7860 ${IMAGE_NAME}:latestRun the image with persisted models, downloaded data and outputs:
docker run --tty --gpus all --rm --name stable-diffusion-webui -p 7860:7860 \
-v /storage/stable-diffusion/data/:/app/data \
-v /storage/stable-diffusion/repositories:/app/repositories \
-v /storage/stable-diffusion/cache:/root/.cache \
-v /storage/stable-diffusion/outputs:/app/outputs \
${IMAGE_NAME}:latestTroubleshooting
- There is no output to the console when starting the image until you generate an image.
This can be fixed by adding the
--ttyargument afterdocker run