It depends. I use pytorch and mostly it works. But sometimes I see a cool library and download it and try to run it and I get a C level error where it says "unsupported hardware" -- in which case I need to run that code on linux.

I think it should be fine since your laptop should just be a client for doing deep learning, and not the server. So whenever you have problems you can just test on a linux machine.

I've personally never written code myself that throws the unsupported hardware error. So it must be some specialty accelerated code that only works with intel or whatever. But yeah this hasn't been an issue for writing code, only for trying to use other people's code (and even then it's pretty rare, usually only happens when you clone from like nvidia or something).