Drop tensorflow-macos requirement

[rmorshea]

As of Tensorflow 2.16.1 tensorflow-macos has been deprecated.

Releases of tensorflow>=2.16.1 cannot be properly installed with tensorflow-macos.

The solution is for m-loop to drop the tensorflow-macos requirement and require tensorflow>=2.16.1.

[zakv]

Hmm, the default Anaconda channel still only has 2.12.0 as it latest release. Then installing M-LOOP (with pip) would install newer tensorflow from PyPI. However, last I checked, installing tensorflow in a conda environment using pip would complete successfully, but then importing M-LOOP would cause errors with huge/incomprehensible tracebacks. I'll check to see if that's still an issue.

Unfortunately M-LOOP doesn't have CI set up. I can test things out on my linux box to see whether it works there, but could you try building the environment and running pytest on your Mac as well?

Also, could you also update the minimum required tensorflow version listed in the "Dependencies" section of install.rst?

[zakv]

Oh and it looks to me like requirements.txt isn't read in by setup.py so I'm not sure if it has any effect, but we should update the version there as well for consistency

[rmorshea]

I set up an environment, installed with pip, and ran the test suite without issue on my Mac.

[zakv]

Awesome, I ran the following on my Ubuntu 20.04 box with these changes and the

mamba create --name mloop-dev python=3.11 pip tensorflow
mamba activate mloop-dev
pip install -e .
pytest

So seems like installing tensorflow with pip in a conda environment is no longer an issue.

@rmorshea could you also update the minimum required tensorflow version listed in the "Dependencies" section of install.rst? And then I figure you'll want a new M-LOOP release with these changes? (edit: nevermind that first part, I just pushed a commit to edit install.rst)