Hi all, I’ve been working on a tutorial for how to use xbatcher with Keras/Tensorflow for a few weeks.
The basic idea is to have a (mostly) self-contained example of how to use xbatcher in a realistic scenario. This particular example closely follows a published research paper, but is designed to allow for user experimentation with the training/target variables and CNN architecture, which was a significant part of the work that led to the publication.
There’s also a bit of guidance on the software engineering side to show how to manage a complex research situation with clarity.
I have another notebook along similar lines that uses a graph neural network with Pytorch that I hope to include soon as well.
As it stands, it’s still in a draft state, but I’m curious to know what people think about this idea.