search

lobe / lobe-python

Python toolset for working with Lobe models
MIT License
185 stars 38 forks source link

any support for apple silicon? #26

Open celaraze opened 3 years ago

celaraze commented 3 years ago

Now we can install tensorflow on apple silicon by miniforge and apple/tensorflow-macos But I dont know if there's any support for it.

Thanks.

mbeissinger commented 3 years ago

For this library, you can install tensorflow for the mac on your own (as long as it follows the tensorflow~=2.5.0 API and is callable in python with import tensorflow), and then pip install lobe should work with the tensorflow backend. I believe ONNX should also work on M1 chips, but I don't have a machine to test with.

celaraze commented 3 years ago

Thanks for replying. I will try it these days, and will give some feedbacks to you~ Have a nice day.

celaraze commented 3 years ago

I found a way. M1 cannot directly install lobe [TF], but it can be implemented by installing parallel virtual machine on M1. The direct installation of lobe [TF] on M1 will prompt the unsupported architecture (armv7 is OK), but in Debian in the virtual machine, I successfully installed and ran lobe [TF]. Debian is also arm64, which is amazing!

celaraze commented 3 years ago

I found a way. M1 cannot directly install lobe [TF], but it can be implemented by installing parallel virtual machine on M1. The direct installation of lobe [TF] on M1 will prompt the unsupported architecture (armv7 is OK), but in Debian in the virtual machine, I successfully installed and ran lobe [TF]. Debian is also arm64, which is amazing!

update, before run pip install "lobe[tf]" , Your must run conda install tensorflow first, or won't install lobe[tf] correctlly.