Usage of stop_gradient generates error (Keras)

Question

I am trying to figure out how stop_gradient works in Keras.

Python 3.7, Tensorflow 1.14.0, Keras 2.2.4

I have coded a simple NN example. But it generates the f