Installing Tensorflow to make sure it properly detects the GPU on M1 machines it's cumbersome.
Apple provides a solution that needs conda. If you don't want to install miniconda and prefer a more lightweight solution you can follow those steps. Instead we present here a solution that uses Python native virtual environments.
- Install
hdf5
withbrew install hdf5
- Create a virtual enviroment
python3 -m venv /tf-gpu-m1