Notes on installing on an M1 Mac

[jebyrnes]

I wrestled with install for a few days. Here's what finally did it for me (with some ideas about how it might work for others)

1. Install miniconda3 from https://docs.conda.io/en/latest/miniconda.html
   using the proper installer for your chip architecture. Note, TagLab wants
   python3 to be in the 3.10 family, NOT 3.11 at the time of writing this
   (2023-04-09).

2. Make sure you have installed homebrew from https://brew.sh/

3. Using homebrew, install qt, qt5, gdal, and cmake. I suppose you could
   skip the above installation of miniconda if you can install a version of
   python3 that is in the 3.10 family, but currently homebrew is installing
   in the 3.11 family by default.

4. Check what version of python is your default by opening a new terminal
   window and typing python3 --version and make sure if it is running the
   3.10.* version. You will likely have multiple versions on your computer.
   type where python3 to see which ones. Try each, and see which gives you
   the right version. You can alias this to python3 in ~/bin or just remember
   to type out the full path from this point forward. Alternately, you can
   likely muck about in your .zshrc to get the path you want set to be your
   primary python3 version. Your call.

5. Once you've got everything set so that you're running the right version
   of python3 and using the right architecture, install pyqt5 with the
   following

pip3 install pyqt5 --config-settings --confirm-license= --verbose

as otherwise it will make your installation of TagLab hang.

6. While you're at it, get the right version of tensorflow and torch
   installed for the M1/arm64 architecture

conda install pytorch torchvision torchaudio -c pytorch-nightly

pip3 install --pre torch torchvision torchaudio --index-url 
https://download.pytorch.org/whl/nightly/cpu

conda install -c apple tensorflow-deps

pip3 install tensorflow-macos

pip3 install tensorflow-metal

You can read more about this on
https://towardsdatascience.com/installing-pytorch-on-apple-m1-chip-with-gpu-acceleration-3351dc44d67c
and
https://betterprogramming.pub/installing-tensorflow-on-apple-m1-with-new-metal-plugin-6d3cb9cb00ca
or just googling "install torch m1 osx"

7. In the TagLab directory, you should be OK from terminal now to run

python3 install.sh

and if it installs, you should be able to run

python3 TagLab.py