GRAPHSAGE "model.fit()" reaching exception in StellarGraph "Inductive node classification and representation learning using GraphSAGE" demo

[cloudmadeofcandy]

Describe the bug

I was using the GraphSAGE model in my own code today and realized that whenever I fit the model, instead of using the main pipeline, StellarGraph met an exception, then fired the error handlers.

The graph shows that GraphSAGE indeed learns from the data inductively, however the training time shot up to around 6 minutes per 10 epochs.

This is reproducible via running the stable (1.2.1) demo "https://colab.research.google.com/github/stellargraph/stellargraph/blob/master/demos/node-classification/graphsage-inductive-node-classification.ipynb".

To Reproduce

Steps to reproduce the behavior:

Run the following demo:

https://colab.research.google.com/github/stellargraph/stellargraph/blob/master/demos/node-classification/graphsage-inductive-node-classification.ipynb

Observed behavior

The model.fit() function reduce to error_handler, then uses Keras default fit function

Expected behavior

StellarGraph should be using the adapted fit function

Environment

Operating system: Ubuntu, Google Colab

Python version: 3.8.10

Package versions: stellargraph==1.2.1, tensorflow==2.11.0, keras==2.11.0