Image Currently Out of Order

imagegenius/immich-cuda-node

immich-cuda-node is a remote machine leaning worker for Immich

Supported Architectures

We use Docker manifest for cross-platform compatibility. More details can be found on Docker's website.

To obtain the appropriate image for your architecture, simply pull ghcr.io/imagegenius/immich-cuda-node:latest. Alternatively, you can also obtain specific architecture images by using tags.

This image supports the following architectures:

Architecture	Available	Tag
x86-64	✅	amd64-<version tag>
arm64	❌
armhf	❌

Application Setup

To start, you will need to modify your Immich instance by disabling its built-in machine learning service.

On your main Immich container, set the necessary environment variables:

• Set CUDA_ACCELERATION to false. This will disable CUDA acceleration.
• Set IMMICH_MACHINE_LEARNING_URL to http://<ip>:3003.

Replace <ip> with the IP address of your immich-cuda-node container/host.

Enabling GPU Acceleration with this container

To start this image with GPU support, you need to modify your Docker settings.

If you're using the Docker CLI, add the --gpus=all flag to your command.

docker run --gpus=all ...

If you're using Docker Compose, add the following under your service:

    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              capabilities: [gpu]

This will reserve all available GPUs for your container, using the NVIDIA driver.

Usage

Example snippets to start creating a container:

Docker Compose

---
version: "2.1"
services:
  immich-cuda-node:
    image: ghcr.io/imagegenius/immich-cuda-node:latest
    container_name: immich-cuda-node
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Etc/UTC
    volumes:
      - path_to_appdata:/config
    ports:
      - 3003:3003
    restart: unless-stopped

Docker CLI (Click here for more info)

docker run -d \
  --name=immich-cuda-node \
  -e PUID=1000 \
  -e PGID=1000 \
  -e TZ=Etc/UTC \
  -p 3003:3003 \
  -v path_to_appdata:/config \
  --restart unless-stopped \
  ghcr.io/imagegenius/immich-cuda-node:latest

Variables

To configure the container, pass variables at runtime using the format <external>:<internal>. For instance, -p 8080:80 exposes port 80 inside the container, making it accessible outside the container via the host's IP on port 8080.

Variable	Description
-p 3003	WebUI Port
-e PUID=1000	UID for permissions - see below for explanation
-e PGID=1000	GID for permissions - see below for explanation
-e TZ=Etc/UTC	Specify a timezone to use, see this list.
-v /config	Package cache, this folder could potentially use +5GB!

Umask for running applications

All of our images allow overriding the default umask setting for services started within the containers using the optional -e UMASK=022 option. Note that umask works differently than chmod and subtracts permissions based on its value, not adding. For more information, please refer to the Wikipedia article on umask here.

User / Group Identifiers

To avoid permissions issues when using volumes (-v flags) between the host OS and the container, you can specify the user (PUID) and group (PGID). Make sure that the volume directories on the host are owned by the same user you specify, and the issues will disappear.

Example: PUID=1000 and PGID=1000. To find your PUID and PGID, run id user.

  $ id username
    uid=1000(dockeruser) gid=1000(dockergroup) groups=1000(dockergroup)

Updating the Container

Most of our images are static, versioned, and require an image update and container recreation to update the app. We do not recommend or support updating apps inside the container. Check the Application Setup section for recommendations for the specific image.

Instructions for updating containers:

Via Docker Compose

• Update all images: docker-compose pull
  • or update a single image: docker-compose pull immich-cuda-node
• Let compose update all containers as necessary: docker-compose up -d
  • or update a single container: docker-compose up -d immich-cuda-node
• You can also remove the old dangling images: docker image prune

Via Docker Run

• Update the image: docker pull ghcr.io/imagegenius/immich-cuda-node:latest
• Stop the running container: docker stop immich-cuda-node
• Delete the container: docker rm immich-cuda-node
• Recreate a new container with the same docker run parameters as instructed above (if mapped correctly to a host folder, your /config folder and settings will be preserved)
• You can also remove the old dangling images: docker image prune

Versions

• 25.05.23: - Initial Release.