Stable Diffusion WebUI Forge Docker Image

This Docker image is pre-configured for Stable Diffusion WebUI Forge (not Auto1111 WebUI), offering a streamlined way to run the WebUI with all necessary components bundled. By using this Docker image, you gain access to the following features:

• Pre-configured WebUI Forge setup
• Optional CivitAI model downloader integration
• Optional Cloudflare Tunnel configuration

Simply clone the repository and start the service with docker compose up -d. Note that models are not included and must be downloaded separately using the provided options.

Supported CUDA Versions

Ensure that the NVIDIA driver version on your host machine is compatible with the selected Docker image. Refer to the table below for details:

Docker Image	Required NVIDIA Driver Version (Host Machine)	Forge Configuration
sammrai/sd-forge-docker:12.4.0	550.54.14 or higher	CUDA 12.4 + Pytorch 2.4 (Fastest, but MSVC issues possible, xformers may not work)
sammrai/sd-forge-docker:12.1.0	530.30.02 or higher	CUDA 12.1 + Pytorch 2.3.1 (Recommended)

Key Notes:

1. The NVIDIA driver version on the host machine must meet or exceed the required version for the selected Docker image.
2. The listed driver versions are based on public documentation and do not guarantee full compatibility or official endorsement.

Setup Instructions

1. Clone the Repository

Run the following commands to clone the repository and navigate to the project directory:

git clone https://github.com/sammrai/sd-forge-docker.git
cd sd-forge-docker

2. Start the Docker Containers

Use Docker Compose to start the containers:

docker compose up -d

Note: Models are not included by default. You can either download them manually or use the optional CivitAI integration.

Deployment Options

Choose the deployment method that best suits your needs:

1. Standard GPU Usage

By default, the docker-compose.yml file is configured for NVIDIA GPUs. Ensure your system has:

• The appropriate NVIDIA drivers
• The NVIDIA Container Toolkit installed

2. CPU-Only Usage

If GPU acceleration is not required, modify the docker-compose.yml file to disable GPU support and enable CPU-specific options. Update the ARGS environment variable as follows:

   environment:
-     ARGS: "--listen --enable-insecure-extension-access --port 7680 --api"
+     ARGS: "--listen --enable-insecure-extension-access --port 7680 --api --always-cpu --skip-torch-cuda-test"
-   deploy:
-     resources:
-       reservations:
-         devices:
-           - driver: nvidia
-             count: 1
-             capabilities: [gpu]

3. Using Cloudflare Tunnel

For secure external access to the WebUI, configure a Cloudflare Tunnel. Follow these steps:

1. Set Up Environment Variables:

   Add your TUNNEL_TOKEN to the .env file:

   TUNNEL_TOKEN=yourtoken...
   

2. Replace docker-compose.yml:

   Overwrite the default docker-compose.yml with docker-compose-tunnel.yml:

   cp docker-compose-tunnel.yml docker-compose.yml

3. Start the Containers with the Tunnel Configuration:

   docker compose up -d

4. Complete the Cloudflare Tunnel Setup:

   Refer to the Cloudflare Tunnel Documentation for detailed setup instructions. This includes authenticating with Cloudflare and establishing the tunnel using your TUNNEL_TOKEN.

CivitAI Integration (Optional)

The CivitAI model downloader automates the downloading and placement of models. While optional, it is highly recommended to add your CIVITAI_TOKEN in the .env file, as many models require authentication to download.

1. Add Your CIVITAI Token
   Include your CIVITAI_TOKEN in the .env file to enable seamless integration:

   CIVITAI_TOKEN=yourtoken...

2. Supported Model Types and Example Commands
   The following model types are supported for automatic downloads, along with their corresponding aliases to ensure models are placed in the correct directories:

   • Lora: @lora
   • VAE: @vae
   • Embed: @embed
   • Checkpoint: @checkpoint

   Here are some example commands to download and place models:

   docker compose exec webui civitdl 257749 439889 @checkpoint
   docker compose exec webui civitdl 332646 @embed
   docker compose exec webui civitdl 660673 @vae
   docker compose exec webui civitdl 341353 @lora