Finch GPU support

[ktoulgaridis]

What is the problem you're trying to solve?. I quickly went through the README and couldn't find an entry regarding GPU usage with Finch. The project looks exciting overall, however, in my case, GPU availability is more than a nice-to-have, and I suppose most professionals working with ML would have a similar opinion.

Describe the feature you'd like It would be amazing to have finch run to support nvidia-container-toolkit and have a flag --gpus allowing the user to provide one or more NVIDIA devices to the running container.

Additional context

[pendo324]

Currently, GPU passthrough is not supported in QEMU on MacOS. It is possible that this could be added to Finch / Lima / QEMU at a later date, but there are no plans right now.

Since ML containers are definitely a legitimate usecase, I'm going to mark this issue as an enhancement and update the title slightly.

Thanks for the question!