OperatorNotAllowedInGraphError: using a `tf.Tensor` as a Python `bool` is not allowed in Graph execution. Use Eager execution or decorate this function with @tf.function

Course Q&A Deep Learning Specialization Convolutional Neural Networks
5

You can see the bug in the code that @priyamthakkar8 posted above. Notice that the first call to ReLU above that one is different. The syntax is pretty confusing here and they don’t do a good job of explaining it, but you are doing everything in two steps here:

  1. You call a TFL function to “instantiate” it with a given set of parameters.
  2. Then you “invoke” that function with an actual tensor as input and it outputs a tensor.

That’s why you see two sets of parens in the earlier correct expression:

The ReLU() is the “instantiation” step. That gives you back a function and then you invoke it with the input Z1.

Here is a great Discourse post from Ai_Curious which explains the Keras Sequential and Functional APIs. Way more useful than the minimal explanations in the course materials.

8 Likes