Is it possible for PixelCNN to tell us what it generates?

Question

I coded PixelCNN with the help of Keras official website. Also, I read the paper. I can use PixelCNN, similar to a decoder or generator (to generate samples). My question is, "is it possible to train PixelCNN to tell us what is predicted?". For example, consider we train an autoencoder on a dataset of animals. The PixelCNN samples from latent space and can predict the encoded picture. Now, I want to know, is there a way for PixelCNN to tell us what (e.g. which animal) it generated after prediction? I thought about probability methods like softmax, but I am not sure about them.