“No gradients provided for any variable: ” when using TF2.0 Keras and using tf.GradientTape

Question

\' In my code, I need to take the derivative of the network output with respect to its two inputs. Then use these two new variables to calculate a loss, then calculate the gradi