AI Deploy - Tutorial - Deploy Stable Diffusion WebUI

Introduction

Stable Diffusion is a deep learning model released in 2022. It is primarily used to generate detailed images and videos conditioned on text descriptions, though it can also be applied to other tasks such as inpainting, outpainting, and generating image-to-image translations guided by a text prompt. What makes it stand out is that it is open-source, making it more accessible than its predecessors.

Stable Diffusion WebUI, also known as AUTOMATIC1111 (its creator), is the default Web User Interface (WebUI) for many advanced users of Stable Diffusion. This interface lets you choose your Stable Diffusion models (checkpoints), extensions and generation parameters.

Objective

In this tutorial, we will guide you through the process of deploying the AUTOMATIC1111 Stable Diffusion WebUI on AI Deploy. We will also show you how to add new model checkpoints to your application.

Unlike a locally hosted application, using AI Deploy offers extremely fast inference since it is deployed on powerful resources (GPUs). So you can generate the content you want in as little time as possible.

Requirements

To build and deploy your Stable Diffusion app, you need:

• Access to the OVHcloud Control Panel
• An AI Deploy Project created inside a Public Cloud project in your OVHcloud account
• A user for AI Deploy & Object Storage
• The OVHcloud AI CLI installed on your local computer
• Docker installed on your local computer, or access to a Debian Docker Instance, which is available on the Public Cloud

Instructions

You are going to follow different steps to deploy your Stable Diffusion application:

• Step 1: Set up the environment
• Step 2: Stable Diffusion app deployment
• Step 3: Add Stable Diffusion checkpoints

Step 1: Set up the environment

In this section, we will guide you through the process of using a Docker image that will serve as the foundation for deploying the Stable Diffusion model on AI Deploy.

To do this, you will need Docker, either installed directly on your computer, or using a Debian Docker Instance, available on the Public Cloud.

Build a Docker image

The Dockerfile that builds the image to use is already provided. Clone the ai-training-examples GitHub repository:

git clone https://github.com/ovh/ai-training-examples.git

Then run cd ai-training-examples/apps/gradio/stable-diffusion/ to move to the Stable Diffusion Web UI folder. If you run ls to list the existing files, you should see the Dockerfile, which contains all the commands you can call on the command line to run your application (like installing system packages for example). Here is the Dockerfile:

# 🐳 Base image
FROM pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime

# 🚫 Remove interactivity since using the base image will ask for a timezone - This allows to not provide it
ENV DEBIAN_FRONTEND=noninteractive

# 📚 Install missing system packages (git, libgl1, ..., are needed for Stable Diffusion and are not installed in the base image)
RUN apt-get update && \
    apt-get install -y wget git python3 python3-venv libgl1 libglib2.0-0

# 👱 Set the working directory inside the container
WORKDIR /workspace

# 📥 Download the webui.sh script from the specified URL
RUN wget -q https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/webui.sh

# 👮‍♀️ Make the webui.sh script executable
RUN chmod +x webui.sh

# ⌛️ Install the webui.sh file (--exit parameter allows to only install it without without running it)
RUN ./webui.sh -f --exit

# 🔑 Give correct access rights to the OVHcloud user
RUN chown -R 42420:42420 /workspace
ENV HOME=/workspace

As you can see from the comments, this Dockerfile installs several packages, downloads the interface installation script and then runs it. These instructions will be executed step by step during the build process.

To launch this process, make sure you are in the Dockerfile folder (ai-training-examples/apps/gradio/stable-diffusion/). Once you are in it, run one of the following commands to build your application image:

# Build the image using your machine's default architecture
docker build . -t sd_webui:latest

# Build image targeting the linux/amd64 architecture
docker buildx build --platform linux/amd64 -t sd_webui:latest .

• The first command builds the image using your system’s default architecture. This may work if your machine already uses the linux/amd64 architecture, which is required to run containers with our AI products. However, on systems with a different architecture (e.g. ARM64 on Apple Silicon), the resulting image will not be compatible and cannot be deployed.

• The second command explicitly targets the linux/AMD64 architecture to ensure compatibility with our AI services. This requires buildx, which is not installed by default. If you haven’t used buildx before, you can install it by running: docker buildx install

Push the image into a registry

Once your image is built, you will need to tag it and push it to a registry. When working with OVHcloud AI Solutions, several registries can be used (OVHcloud Managed Private Registry, Docker Hub, GitHub packages, ...). In this tutorial, we will use the OVHcloud shared registry, a solution dedicated to AI Solutions.

You can find the address of your shared registry by launching this command:

ovhai registry list

You should see a registry with the SHARED kind. Then, log in on your shared registry with your usual AI Platform user credentials:

docker login -u <user> -p <password> <shared-registry-address>

Tag the compiled image and push it into your shared registry:

docker tag sd_webui:latest <shared-registry-address>/sd_webui:latest
docker push <shared-registry-address>/sd_webui:latest

Step 2: Stable Diffusion app deployment

Once your image has been pushed, it can be used to deploy new AI Solutions.

Run the following command to deploy your Stable Diffusion application by running your customised Docker image:

ovhai app run <shared-registry-address>/sd_webui:latest \
    --name stable-diffusion-webui \
    --gpu 1 \
    --default-http-port 7860 \
    --volume stable_diff_checkpoints@GRA:/workspace/stable-diffusion-webui/models/:rw \
    -- bash -c './webui.sh --xformers --enable-insecure-extension-access --listen --api'

After executing the command, the volume will be created if it doesn't exist, and the Docker image will be pulled. Subsequently, the Web UI will download a default Stable Diffusion checkpoint, making it ready for use.

You can follow the progress of your app's deployment and these steps using the "Info Url" link, or via your app's logs using (--follow parameter allows to stream the logs):

ovhai app logs <app_id> --follow

When your app is ready for use, you will be able to generate your first images using the default checkpoint.

For your information, the --volume parameter allows to use both Swift and S3* compatible Object Storage buckets. However, it's important to note that for S3 compatible usage, a proper configuration is necessary. If S3 compatible Object Storage is not configured yet and you wish to use it, please read the S3 compatible compliance guide.

Step 3: Add Stable Diffusion checkpoints

It is likely that you will be interested in adding new checkpoints to generate images in a specific style, of better quality than those provided by the default checkpoint.

As explained before in the --volume parameter, model files are saved in the stable-diffusion-webui/models/ directory. This directory contains several sub-directories. For example, Stable diffusion checkpoints are stored in the stable-diffusion-webui/models/Stable-diffusion directory, while Variational Autoencoder (VAE) ones are saved in stable-diffusion-webui/models/VAE.

To add a new Stable Diffusion checkpoint, you need to download it and add it to the corresponding folder (stable-diffusion-webui/models/Stable-diffusion).

For example, if you want to add and use the XL version of Stable Diffusion, you will have to:

• Look for the model checkpoint. Many are available on the Stability AI Hugging Face page.
• Download the model checkpoint. This can be done manually or via the command line, running:

curl -L -O https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors

• Add the checkpoint to the bucket you have mounted on your app. Here is the default command to do that:

ovhai bucket object upload <bucket_name>@<datastore_alias> object1 --add-prefix /model_subdirectory/

Following the example given in this tutorial (bucket named stable_diff_checkpoints and created on GRA), we will use:

ovhai bucket object upload stable_diff_checkpoints@GRA sd_xl_base_1.0.safetensors --add-prefix /Stable-diffusion/

Once the checkpoint has been uploaded, you should see it when you list the objects in your volume:

`ovhai bucket object list <bucket_name>@<datastore_alias>`

Following the example given in this tutorial, we will use:

`ovhai bucket object list stable_diff_checkpoints@GRA`

Adding an object to a bucket will not make it immediately available in the app. You will need to manually trigger a data synchronization.

To do this, you will need the UUID of your app. It has been given to you when you deployed it. If you've lost it, you can find it again in the OVHcloud Control Panel, or via the CLI by listing all your apps, using ovhai app list. You will have to find the UUID of the app named stable-diffusion-webui.

You can then start a datasync to synchronize your app with your bucket, which contains your new model checkpoint, by running:

ovhai app pull-data <app_id>

Depending on the size and number of checkpoints added, this will take more or less time. The progress of the data sync can be tracked via the app's information:

ovhai app get <app_id>

In the information displayed, there will be a Data Sync: field containing all the data syncs performed on your app. In your case, you are interested in the most recent one (displayed furthest down). If it's in RUNNING state, it means that it has not finished. You can also see the progress of uploaded data. It will be ready when it reaches DONE status.

Then, on the Web UI, reload the list of usable Stable Diffusion checkpoints at the top left, by clicking on the 🔄 icon. You should see your new model checkpoint in it and ready for use!

Go further

This tutorial has walked you through deploying a Stable Diffusion Web UI for image generation.

If you are interested in using an API, you can use the --api parameter and refer to the AUTOMATIC1111 documentation.

Note that other projects exist as APIs, such as One Diffusion, from BentoML, which is very easy to implement. Here is the One Diffusion command to deploy Stable Diffusion XL:

ovhai app run pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime \
    --name sdxl \
    --default-http-port 3000 \
    --volume standalone:/workspace:rw \
    --probe-path /healthz \
    -- bash -c 'pip install onediffusion && /workspace/.local/bin/onediffusion start stable-diffusion-xl'

If you are interested in image generation concepts, you can learn how image generation networks work and train your own Generative Adversarial Network. Check out this AI Notebooks guide: Create and train an image generation model.

If you need training or technical assistance to implement our solutions, contact your sales representative or click on this link to get a quote and ask our Professional Services experts for a custom analysis of your project.

Feedback

Please send us your questions, feedback and suggestions to improve the service:

• On the OVHcloud Discord server

*: S3 is a trademark of Amazon Technologies, Inc. OVHcloud’s service is not sponsored by, endorsed by, or otherwise affiliated with Amazon Technologies, Inc.