A docker image to automatically setup everything you needed to run SD Next in a container
- Ubuntu 22.04 LTS
- Support CUDA from 11.8 to 12.5
- Support ROCm from 5.5 to 6.1
- Python 3.10
- PyTorch 2.3.1
- SD Next (94f6f0d)
- Jupyter Lab Installed
- Original + Diffusers Backend supported
- Vast majority of model types supported
- Plug'n'Play (packages pre-installed)
- Easy management(start, git version, download from links) with pre-built jupyter notebook
You can deploy this image in Runpod with this template
The template use cu121 version, so please remember to use the pod that support CUDA 12.1
Nvidia GPU: You can deploy this image in Vast.ai with this template
The template use cu121 version, so please remember to use the instance that support CUDA 12.1
AMD GPU: You can deploy this image in Vast.ai with this template
The template use rocm5.7 version, so please remember to use the instance that support ROCm 5.7
Run the following command in your terminal For Nvidia CUDA:
docker run -d \
--gpus all \
-v /workspace \
-p PORT:3000 \
-p PORT:3001 \
-e JUPYTER_LAB_PASSWORD="" \
yoinky3000/sd-next-docker:x.x.x-cuxxx
For AMD ROCm
docker run -d \
--gpus all \
-v /workspace \
-p PORT:3000 \
-p PORT:3001 \
-e JUPYTER_LAB_PASSWORD="" \
yoinky3000/sd-next-docker:x.x.x-rocmx.x
Note
:latest
points to :latest-cuda
points to latest tag for CUDA 12.1
:latest-rocm
points to latest tag for ROCm 5.7
You will need to replace PORT
with the port number you want the apps to expose to,
For the details of the ports, please scroll down to the PORTS section
You can change the value of JUPYTER_LAB_PASSWORD if you need it
Important
Each version of the image will be built for each CUDA and ROCm version listed here specifically
To check which CUDA version of the image is suitable, open the terminal and use nvidia-smi
to check
the CUDA version your system has installed, you should see CUDA Version: XX.Y
in the output (below is an example)
now you can replace cuxxx with cuXXY
To check which ROCm version of the image is suitable, open the terminal and use apt show rocm-libs
to check
the ROCm version your system has installed, you should see Version: X.Y.Z...
in the output (below is an example)
now you can replace rocmx.x with rocmX.Y
Internal Port | Apps |
---|---|
3000 | Jupyter LAB |
3001 | SD Next |
Variable | Description | Default |
---|---|---|
JUPYTER_LAB_PASSWORD | This will be used as the password for jupyter lab | None (no password is required to access the lab) |
Contributions are welcome! Create an issue if you have any problem with this image, or create pull request if you want to add new features in this repo
The authors of this project are not responsible for any content generated using this interface.