What CLIP Machine Learning Model can I use for Immich?
from Ek-Hou-Van-Braai@piefed.social to selfhosted@lemmy.world on 05 Jun 13:21
https://piefed.social/post/837473

I’m currently running my Immich server on a mini PC with proxmox

It’s got 3x N97 CPU cores available to it and 7gb of ram It’s using the default ViT-B-32__openai model, I was wondering if I can use a more powerful model, but I’m not sure which one or if I should enable hardware acceleration etc.

This is my yaml file

  immich-machine-learning:  
    container_name: immich_machine_learning  
    # For hardware acceleration, add one of -[armnn, cuda, rocm, openvino, rknn] to the image tag.  
    # Example tag: ${IMMICH_VERSION:-release}-cuda  
    image: ghcr.io/immich-app/immich-machine-learning:${IMMICH_VERSION:-release}  
    # extends: # uncomment this section for hardware acceleration - see https://immich.app/docs/features/ml-hardware-acceleration  
    #   file: hwaccel.ml.yml  
    #   service: cpu # set to one of [armnn, cuda, rocm, openvino, openvino-wsl, rknn] for accelerated inference - use the `-wsl` version for WSL2 where applicable  
    volumes:  
      - immich-model-cache:/cache  
    env_file:  
      - stack.env  
    restart: always  
    healthcheck:  
      disable: false

I looked at the docs but it’s a bit confusing so that’s why I’m here.

#docker #immich #machine learning #selfhost #selfhosted

threaded - newest

just_another_person@lemmy.world on 05 Jun 13:41 next collapse

According to this paste, you’re not even using inference at all, or rather, it’s using CPU.

Change the release tag and “cpu” to openvino and see if that performs any better.

Ek-Hou-Van-Braai@piefed.social on 06 Jun 07:39 collapse

thanks I managed to enable openvino and pick a stronger model and it's working quite well, I also bumped the ram to 10gb

non_burglar@lemmy.world on 05 Jun 13:59 next collapse

OpenVino is about your only option here. It is not super efficient and will increase system load during those jobs.

MangoPenguin@piefed.social on 05 Jun 14:22 collapse

They have a list here of the models with performance and RAM usage data: https://immich.app/docs/features/searching/

You kind of just have to pick one, try it, and see if it crashes from low memory.

Also enable OpenVINO HWaccel, because it will be extremely slow otherwise.