MusPy: A Toolkit for Symbolic Music Generation

TL;DR

MusPy is an open-source Python library that simplifies symbolic music generation by providing tools for dataset management, data processing, and model evaluation, along with analysis of dataset similarities and generalizability.

Contribution

It introduces MusPy, a comprehensive toolkit for symbolic music generation, including dataset analysis and cross-dataset evaluation features.

Findings

Identified domain overlaps among popular music datasets.

Showed some datasets contain more representative cross-genre samples.

Provided a guide for dataset selection in future research.

Abstract

In this paper, we present MusPy, an open source Python library for symbolic music generation. MusPy provides easy-to-use tools for essential components in a music generation system, including dataset management, data I/O, data preprocessing and model evaluation. In order to showcase its potential, we present statistical analysis of the eleven datasets currently supported by MusPy. Moreover, we conduct a cross-dataset generalizability experiment by training an autoregressive model on each dataset and measuring held-out likelihood on the others---a process which is made easier by MusPy's dataset management system. The results provide a map of domain overlap between various commonly used datasets and show that some datasets contain more representative cross-genre samples than others. Along with the dataset analysis, these results might serve as a guide for choosing datasets in future research. Source code and documentation are available at https://github.com/salu133445/muspy .