Understanding TensorFlow code

In this section, we'll take a quick look at the TensorFlow code to understand a bit more how the sampling, interpolating, and humanizing code works. This will also make references to the first section of this chapter, Continuous latent space in VAEs, so that we make sense of both the theory and the hands-on practice we've had.

But first, let's do an overview of the model's initialization code. For this section, we'll take the cat-drums_2bar_small configuration as an example and the same model initialization code we've been using for this chapter, meaning batch_size of 8.

Building the VAE graph

We'll start by looking at the TrainedModel constructor in the models...