This code was built on top of the implementation of Dr. Tristan Behrens of the "MMM: Exploring Conditional Multi-Track Music Generation with the Transformer" paper. The changes to the original code are just a few adaptations so the repo could work with the LMD dataset.
This repository allows the user to represent the MIDI files from the LMD clean dataset as text tokens, as explained in the paper "MMM: Exploring Conditional Multi-Track Music Generation with the Transformer" paper. It can be used to create the datasets using the MMMTrack version from the paper.
To be defined
Released under the Apache-2.0 License.