Docker image for A1111 Stable Diffusion Web UI, Kohya_ss and ComfyUI

ROCm fork. In progress.

Installs

• Ubuntu 22.04 LTS
• ROCm 6.1
• Python 3.10.12
• Automatic1111 Stable Diffusion Web UI 1.7.0
• Dreambooth extension 1.0.14
• Deforum extension
• ControlNet extension v1.1.424
• After Detailer extension v23.11.1
• Locon extension
• ReActor extension (replaces roop)
• Inpaint Anything extension
• Infinite Image Browsing extension
• CivitAI extension
• CivitAI Browser+ extension
• Kohya_ss v22.4.0
• ComfyUI
• ComfyUI Manager
• Torch 2.4.0
• xformers 0.0.23
• sd_xl_base_1.0.safetensors
• sd_xl_refiner_1.0.safetensors
• sdxl_vae.safetensors
• inswapper_128.onnx
• runpodctl
• croc
• rclone
• Application Manager

Building the Docker image

Since the Stable Diffusion models are pretty large, you will need at least 8GB of system memory (not GPU VRAM) to build this image.

The image CAN be built on a t3a.large AWS EC2 instance which has 2 x vCPU and 8GB of system memory. It CANNOT be built on any instances with less memory, eg. t3a.medium.

# Clone the repo
git clone --recursive https://github.com/samexner/stable-diffusion-ROCker.git

# Download the models
cd stable-diffusion-ROCker
wget https://huggingface.co/runwayml/stable-diffusion-v1-5/resolve/main/v1-5-pruned.safetensors
wget https://huggingface.co/stabilityai/sd-vae-ft-mse-original/resolve/main/vae-ft-mse-840000-ema-pruned.safetensors
wget https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors
wget https://huggingface.co/stabilityai/stable-diffusion-xl-refiner-1.0/resolve/main/sd_xl_refiner_1.0.safetensors
wget https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/resolve/main/sdxl_vae.safetensors

# Build and tag the image
docker build -t username/image-name:1.0.0 .

# Log in to Docker Hub
docker login

# Push the image to Docker Hub
docker push username/image-name:1.0.0

Running Locally

Install AMD ROCm Driver

• Linux

Start the Docker container

docker run -it -v /workspace -it --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --device=/dev/kfd --device=/dev/dri --group-add video --ipc=host --shm-size 8G -p 3000:3001 -p 3010:3011 -p 3020:3021 -p 6006:6066 -p 8888:8888 -p 8000:8000 -e JUPYTER_PASSWORD=Jup1t3R! -e ENABLE_TENSORBOARD=1 psnrainbowtrout/sdrocker:1.0.0

You should substitute the image name and tag with your own.

Acknowledgements

1. RunPod for providing most of the container code.
2. Bernard Maltais (core developer of Kohya_ss) for assisting with optimizing the Docker image.
3. Ashley Kleynhans for the original image configured for NVidia.

Community and Contributing

Pull requests and issues on GitHub are welcome. Bug fixes and new features are encouraged.