Should we move to pytorch?

[mathomp4]

As seen in #28, tensorflow is not yet Python 3.13 ready.

That said, a lot of the internet seems to like PyTorch more than Tensorflow. So, should we move GEOSpyD to use it?

I call on @katherbreen, @dbarahon, @amdasilva, and @JulesKouatchou as people more knowledgeable than I to see.

It would be nice for us to support a Python 3.13 GEOSpyD stack

[JulesKouatchou]

I am in favor of adding PyTorch (if possible) to GEOSpyD. I recently wrote introductory materials on PyTorch:

https://github.com/astg606/astg_pymaterials/tree/main/pytorch

The examples in the three notebooks are for CPUs only. I will soon make small modification to make them work on both CPUs and GPUs.

PyTorch appears to be more popular than TensorFlow. People like its ability to change the computational graph at run time. It is good to have both options available.

[mathomp4]

@JulesKouatchou Okay. I can look at maybe adding it and then not installing Tensorflow if Python 3.13 (for now)