In this chapter, we'll be preparing some data for training. Note that this will be covered in more detail in Chapter 7, Training Magenta Models. Preparing data and training models are two different activities that are done in tandem—first, we prepare the data, then train the models, and finally go back to preparing the data to improve our model's performance.

First, we'll start by looking at symbolic representations other than MIDI, such as MusicXML and ABCNotation, since Magenta also handles them, even if the datasets we'll be working with in this chapter will be in MIDI only. Then, we'll provide an overview of existing datasets, including datasets from the Magenta team that were used to train some models we've already covered. This overview is by no means exhaustive but can serve as a starting point when it comes...