TensorFlow Data Type Mismatch in Style Cost Calculation

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

In Neural Style Transfer (NST), while computing the denominator of the style cost function, I encountered a data type mismatch error when using TensorFlow operations (tf.square()) but not when using standard Python operations (**2).

Observations:

:white_check_mark: This works fine (Python Operations):

denominator = (4 * (n_H * n_W)**2 * (n_C)**2)

:cross_mark: This gives a TypeError (TensorFlow Operations):

denominator = (4 * tf.square(n_C) * tf.square(n_C * n_H))

So, my questions are:

  • Why does the Python **2 operator work fine, but TensorFlow’s tf.square() raises a dtype mismatch error?
  • How does TensorFlow handle data types differently compared to standard Python?
2

Differences:

  1. You’ve changed the order of the operands.
  2. n_H, n_W, and n_C are scalars, not tensors.