Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support for easy hardware acceleration config in values #138

Open
joaopedrocg27 opened this issue Oct 27, 2024 · 7 comments
Open

Support for easy hardware acceleration config in values #138

joaopedrocg27 opened this issue Oct 27, 2024 · 7 comments

Comments

@joaopedrocg27
Copy link

Hi!

I was able to make HW transcoding work by modifying the live yaml but I am not sure how this charts supports it.

image image
@bo0tzz
Copy link
Member

bo0tzz commented Oct 27, 2024

values.yaml:

machine-learning:
  image:
    tag: v1.118.2-openvino

@joaopedrocg27
Copy link
Author

But this way it will not use the global tag declaration right? I have configured renovate to look at the global one

@bo0tzz
Copy link
Member

bo0tzz commented Oct 27, 2024

Correct, there's no way to combine the two. It'd be nice to have hwaccel support in the chart so you can just set a type to use and it handles appending the right tag as well as setting the container resources etc. I'll rename this issue to be a FR for that.

@bo0tzz bo0tzz changed the title How to append "-openvino" to ML image name in order to use HW accel? Support for easy hardware acceleration config in values Oct 27, 2024
@bo0tzz
Copy link
Member

bo0tzz commented Oct 27, 2024

I don't know how any of the hardware accel mappings work in kubernetes, so if people who are currently using it can post their values here for the different methods that'd be a useful reference.

@joaopedrocg27
Copy link
Author

Correct, there's no way to combine the two. It'd be nice to have hwaccel support in the chart so you can just set a type to use and it handles appending the right tag as well as setting the container resources etc. I'll rename this issue to be a FR for that.

That would be perfect! Thanks

@joaopedrocg27
Copy link
Author

I don't know how any of the hardware accel mappings work in kubernetes, so if people who are currently using it can post their values here for the different methods that'd be a useful reference.

Just for some more context, my setup is based on node-feature-discovery with intel operator and gpu plugin. I am able to just give it a request of gpu.intel.com/i915: "1". It can even share the GPU along multiple applications, which you cannot do via mounting the dev as a volume. I can investigate how this works in the background but I think that for this some string manipulation in the tag name is enough!

 server:
    enabled: true
    securityContext: # not sure if it is needed
        privileged: true
    resources:
      requests:
        cpu: 300m
        memory: 300Mi
        gpu.intel.com/i915: "1"
      limits:
        cpu: 3
        memory: 3Gi
        gpu.intel.com/i915: "1"

If I do a lspci in the immich container I can see the GPU

@bo0tzz
Copy link
Member

bo0tzz commented Oct 27, 2024

Adding it via a device plugin is definitely sufficient, no need for any manual dev mount stuff!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants