Thanks for creating the repo. It allowed me to understand the basics of GCNs, a good starting point. I wanted to learn by building it from scratch, so I have ported the repo to work with TF 2.2.
I have modified your following scripts to work with TF 2.2
layers/graph.py
karate_supervised.py
karate_unsupervised.py
Used poetry for virtualenv creation, as the README says, you wished to add support for alternatives. So might be worth considering.
Updated README with the required steps to setup locally
Feel free to test out the implementation & see if they work as expected. If you find any issues, kindly add a comment.
Thanks for creating the repo. It allowed me to understand the basics of GCNs, a good starting point. I wanted to learn by building it from scratch, so I have ported the repo to work with TF 2.2.
Feel free to test out the implementation & see if they work as expected. If you find any issues, kindly add a comment.