Integrate with mlx

[r1cc4rd0m4zz4]

I think it might be a good idea to incorporate txtai with mlx.

https://github.com/ml-explore/mlx

I believe the experience on an Apple Silicon may be smoother than on other models of LLM with a transformer.

[davidmezzetti]

Thank you for the suggestion. Does this library offer any sort of emulation or way to run on non apple devices? I'm primarily a Linux developer, so just thinking of a way I could develop a feature like this without an Apple machine.

[r1cc4rd0m4zz4]

I see your point, but this framework is designed for those who use Apple silicon, such as those who use frameworks to take advantage of Nvidia GPUs, because they offer speed advantages.

[davidmezzetti]

Are you aware of any Apple silicon emulators of virtual machines that I could spin up from a Linux environment to develop/test this?

[r1cc4rd0m4zz4]

This could be good news!

https://github.blog/changelog/2024-01-30-github-actions-introducing-the-new-m1-macos-runner-available-to-open-source/

[davidmezzetti]

Thank you for sharing. I'll take a look!