update Dockerfile according to comments and move to separated folder

Co-authored-by: Lukas <[email protected]>
Signed-off-by: Peter Pan <[email protected]>

.dockerignore

@@ -1,2 +1,9 @@
-models
-models/
+.venv
+.che__/
+*.log
+*.git
+*.gitignore
+docker
+models/*
+embeddings/*
+extensions/*

README.md

@@ -106,14 +106,43 @@ Alternatively, use online services (like Google Colab):
 
 ### Running with Docker
 
+#### a) Run with Docker Compose
 DISCLAIMER: This currently only works with NVIDIA GPUs
 
-You need to have [Docker](https://www.docker.com/) installed on your system. Then clone this repository and execute `docker compose up` in the root of the repository. The first time you execute this command will take a long time as all the dependencies are installed. Subsequent runs of the command should start up the webui pretty much instantly. To stop the webui press CTRL+C and wait a few seconds.
+You need to have [Docker](https://www.docker.com/) installed on your system. Then clone this repository and execute `docker compose -f docker/compose.yml up` in the root path of the repository. The first time you execute this command will take a long time as all the dependencies are installed. Subsequent runs of the command should start up the webui pretty much instantly. To stop the webui press CTRL+C and wait a few seconds.
 
 Models are provided to the Docker container using a bind mount. This means that if you add a new model to the models directory it should be available in the webui after a checkpoint refresh without needing to rebuild or restart the container.
 
 The server will be accessible at [localhost:7860](localhost:7860)
 
+#### b) Run with docker CLI
+```
+git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui
+cd stable-diffusion-webui
+export TAG=$(git describe --abbrev=0)        # the image tag, here example uses the latest git tag
+export IMG=AUTOMATIC1111/stable-diffusion-webui:${TAG}
+docker build -t ${IMG} -f docker/Dockerfile .
+docker run --gpus all -d -p 7860:7860 -v $(pwd)/models/:/webui/models  -v $(pwd)/embeddings:/webui/models  $IMG
+# Those `-v` mean mounting your local pre-downloaded model weights, embeddings, extensions( you can also do this for your local textual_inversion_templates, localizations..etc ) to the container, in the same manner.
+```
+
+#### c) Run on Kubernetes
+
+Prerequisite: 
+
+ - You already have a Kubernetes Cluster in place and kube.conf in your machine.
+ - build the docker images as above step (b), and load it to your K8S cluster.
+ - Modify the `YOUR-IMAGE-NAME` and `YOUR-LOCAL-PATH` in `docker/k8s-sd-webui.yaml`
+
+```
+kubectl apply -f docker/k8s-sd-webui.yaml      # Create k8s workload and nodeport service
+kubectl get po -l app=stable-diffusion-webui   # List the container
+#kubectl wait --for=condition=available endpoints/stable-diffusion-webui-service # wait for pod ready, you can CTRL+C to skip it
+kubectl get svc stable-diffusion-webui-service # To show the access NodePort port and access it thru K8S NodePort
+```
+
+To debug, you can check logs from `kubectl logs -f deploy/stable-diffusion-webui`
+
 ### Installation on Windows 10/11 with NVidia-GPUs using release package
 1. Download `sd.webui.zip` from [v1.0.0-pre](https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre) and extract its contents.
 2. Run `update.bat`.

Dockerfile → docker/Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.10-bookworm
+FROM python:3.10
 
 WORKDIR /webui
 
@@ -14,6 +14,6 @@ RUN groupadd --system --gid 1000 webui && \
     chown -R webui:webui .
 USER 1000:1000
 
-RUN ./webui.sh --prepare-environment-only --skip-torch-cuda-test
+RUN ./webui.sh --exit --skip-torch-cuda-test
 
-CMD [ "./webui.sh", "--skip-prepare-environment" ]
+CMD [ "./webui.sh", "--skip-prepare-environment", "--listen" ]

docker/compose.yml

@@ -0,0 +1,28 @@
+services:
+  webui:
+    build: .
+    volumes:
+      - type: bind
+        source: ./models
+        target: /webui/models
+      - type: bind
+        source: ./outputs
+        target: /webui/outputs
+      - type: bind
+        source: ./extensions
+        target: /webui/extensions
+      - type: bind
+        source: ./embeddings
+        target: /webui/embeddings
+      - type: bind
+        source: ./configs
+        target: /webui/configs
+    ports:
+      - 7860:7860
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]

docker/k8s-sd-webui.yaml

@@ -0,0 +1,64 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: stable-diffusion-webui
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: stable-diffusion-webui
+  template:
+    metadata:
+      labels:
+        app: stable-diffusion-webui
+    spec:
+      containers:
+        - name: stable-diffusion-webui
+          image: $(YOUR-IMAGE-NAME) # the image name specified when doing `docker build`
+          ports:
+            - containerPort: 7860
+          volumeMounts:
+            - mountPath: /webui/models
+              name: models-volume
+            - mountPath: /webui/outputs
+              name: image-outputs
+            - mountPath: /webui/extensions
+              name: extensions-volume
+            - mountPath: /webui/embeddings
+              name: embeddings-volume
+          resources:
+            limits:
+              nvidia.com/gpu: 1  # Adjust according to your needs
+          readinessProbe:
+            httpGet:
+              path: /
+              port: 7860
+            initialDelaySeconds: 120
+            periodSeconds: 30
+      volumes:
+        - name: models-volume
+          hostPath:
+            path: $(YOUR-LOCAL-PATH)/models  # absolute path of pre-download model on the host machine
+        - name: image-outputs
+          hostPath:
+            path: $(YOUR-LOCAL-PATH)/outputs
+        - name: extensions-volume
+          hostPath:
+            path: $(YOUR-LOCAL-PATH)/extensions # absolute path of extensions
+        - name: embeddings-volume
+          hostPath:
+            path: $(YOUR-LOCAL-PATH)/embeddings # absolute path of pre-download embeddings
+
+
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: stable-diffusion-webui-service
+spec:
+  type: NodePort  # You can change this to LoadBalancer if needed
+  ports:
+    - port: 7860
+      targetPort: 7860
+  selector:
+    app: stable-diffusion-webui

launch.py

@@ -34,18 +34,14 @@ def main():
 
     launch_utils.startup_timer.record("initial startup")
 
-    if args.prepare_environment_only:
-        print("Setting up requirements wihout starting server as --setup-only flag was passed")
-
     with launch_utils.startup_timer.subcategory("prepare environment"):
         if not args.skip_prepare_environment:
             prepare_environment()
 
     if args.test_server:
         configure_for_tests()
 
-    if not args.prepare_environment_only:
-        start()
+    start()
 
 
 if __name__ == "__main__":

modules/cmd_args.py

@@ -126,4 +126,3 @@
 parser.add_argument("--unix-filenames-sanitization", action='store_true', help="allow any symbols except '/' in filenames. May conflict with your browser and file system")
 parser.add_argument("--filenames-max-length", type=int, default=128, help='maximal length of filenames of saved images. If you override it, it can conflict with your file system')
 parser.add_argument("--no-prompt-history", action='store_true', help="disable read prompt from last generation feature; settings this argument will not create '--data_path/params.txt' file")
-parser.add_argument("--prepare-environment-only", action='store_true', help="launch.py argument: only prepare environment without launching webui run with --skip-torch-cuda-test")

webui-user.sh

@@ -10,7 +10,7 @@
 #clone_dir="stable-diffusion-webui"
 
 # Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
-export COMMANDLINE_ARGS="--listen"
+#export COMMANDLINE_ARGS=""
 
 # python3 executable
 #python_cmd="python3"