allow to add --gpus all and add Local AI community container with GPU support

[luzfcb]

Questions:

1 - How to force the local-ai community container to use a nvidia GPU-enabled tag like: v2.9.0-cublas-cuda12-ffmpeg instead of the regular v2.0.0-ffmpeg-core tag?

2 - For the Nvidia GPU to be able to be used by a container based on the image-based quay.io/go-skynet/local-ai:v2.9.0-cublas-cuda12-ffmpeg it's required to use the --gpus all on the docker run command

How to configure the Nextcloud AIO to include the --gpus all argument on the docker run command of Nextcloud AIO Local-Ai community container?

https://github.com/mudler/LocalAI/blob/master/docs/content/docs/features/GPU-acceleration.md
https://docs.docker.com/config/containers/resource_constraints/#gpu

Note: v2.9.0-cublas-cuda12-ffmpeg tag has more than 16GB in size

[szaimen]

Hi, in order to make this work, first a capability for AIO would need to be added to add all gpus to a container. See #1659 as inspiration.

When that is done, one (you?) could add an additional community container, e.g. name it local-ai-gpu or something that itself is built from v2.9.0-cublas-cuda12-ffmpeg. See #3495 and https://github.com/szaimen/aio-local-ai as inspiration

[szaimen]

Nvidia gpu passthrough is now being worked on in #5132

[szaimen]

This is now released with v10.2.0 Beta. Testing and feedback is welcome! See https://github.com/nextcloud/all-in-one#how-to-switch-the-channel