please add requirements regarding Apple M1/M2 chip

[rlzijdeman]

Hi, I've tried to follow the installation procedure, but both following the guidelines for using and not-using the GPU. I think the issue at hand is that I have Apple M1/M2 chip and docker complains about running Loghi (despite me having the right docker version).

Would it be possible to run Loghi and a Mac M1/M2? Would I need to follow the docker compilation instructions?

[dolfs]

See my separate issue filed, where I attempted to create docker images for M1/M2/M3... it fails

[rvankoert]

Hi Richard/Dolf,

We're considering this now and will try to get a laptop so we can investigate this further.

[dolfs]

There are several possible stages to this:

1. Make the docker images build/work on an Apple Silicon machine
2. Make the code work natively on A.S.
3. Take advantage of Apple recently released AI frameworks (sort of a CUDA replacement)

Realistically 1. should be doable, 2. and 3 are more complicated. A.S. CPUs are quite fast so CPU only operations may not be too bad. Having said that, taking advantage of the GPU, or even better, the Neural Engine should make this scream.

Keep in mind that using docker on MacOS effectively runs the code in a Linux host, so accessing A.S. Native chip features might be difficult.