Why do I get a “ValueError: Shape mismatch” when using CategoricalCrossentropy loss function?

Question

I created a simple TensorFlow model (no convolution layers) using the MNIST dataset. I initially used SparseCategoricalCrossentropy loss function and it worked fin