Tensorflow: How to handle preprocessing.Normalization returning NaN

Question

I am building a keras model and trying to use the tf.keras.layers.experimental.preprocessing.Normalization from https://www.tensorflow.org/api_docs/python/tf/keras/layers/ex