@SidneyLann yes as long as you install it and set the python path up properly.
To answer your other question, we don’t really have any interest in tf-gnn. If you want to be a paying customer or contribute code that’s a different story. Otherwise there’s not a lot of bandwidth for this.
For now our main use cases will be around getting the core right. I believe the TF java people already pointed you at some keras tutorials for this as well. In theory, you could export those models from keras and import them in to dl4j as well. The layers required there are already implemented from the looks of it.