frechet-music-distance 0.1.0

A library for computing Frechet Music Distance.

Frechet Music Distance

Table of Contents

• Introduction
• Features
• Installation
• Usage
• Citation
• Acknowledgements
• License

Introduction

A library for calculating Frechet Music Distance (FMD). This is an official implementation of the paper Frechet Music Distance: A Metric For Generative Symbolic Music Evaluation.

Features

• Calculating FMD and FMD-Inf scores between two datasets for evaluation
• Caching extracted features and distribution parameters to speedup subsequent computations
• Support for various symbolic music representations (MIDI and ABC)
• Support for various embedding models (CLaMP 2)(TODO: CLaMP 1)
• Support for various methods of estimating embedding distribution parameters (TODO)
• Computation of per-song FMD to find outliers in the dataset (TODO)

Installation

The library can be installed from from PyPi using pip:

pip install frechet-music-distance

or directly from source by cloning the repository and installing it locally:

git clone https://github.com/jryban/frechet-music-distance.git
cd frechet-music-distance
pip install -e .

The library was tested on Linux and MacOS, but it should Work on Windows as well.

Usage

The library currently supports MIDI and ABC symbolic music representaions.

Note: When using ABC Notation please ensure that each song is located in a separate file

Command Line

fmd [-h] [--model {clamp2,clamp}] [--reference_ext REFERENCE_EXT] [--test_ext TEST_EXT] [--inf]
                              [--steps STEPS] [--min_n MIN_N] [--clear-cache]
                              <reference_dataset> <test_dataset>

Positional arguments:

• reference_dataset: Path to reference dataset
• test_dataset: Path to test dataset

Options:

• --model {clamp2,clamp}, -m {clamp2,clamp} Embedding model name
• --reference_ext, -r REFERENCE_EXT Music file extension in referene dataset (e.g. .midi). The program will automatically ifer this if not provided
• --test_ext, -t TEST_EXT Music file extension in test dataset (e.g. .midi). The program will automatically ifer this if not provided.
• --inf Use FMD-Inf extrapolation
• --steps, -s STEPS Number of steps when calculating FMD-Inf
• --min_n, -n MIN_N Mininum sample size when calculating FMD-Inf (Must be smaller than the size of test dataset)

Cleanup

Additionaly the pre-computed cache can be cleared by executing:

fmd --clear-cache

Python API

Standard FMD score

from frechet_music_distance import FrechetMusicDistance

metric = FrechetMusicDistance()
score = metric.score(
    reference_dataset="<reference_dataset>",
    test_dataset="<test_dataset>"
)

FMD-Inf score

from frechet_music_distance import FrechetMusicDistance

metric = FrechetMusicDistance()
result = metric.score_inf(
    reference_dataset="<reference_dataset>",
    test_dataset="<test_dataset>",
    steps=<num_steps> # default=25
    min_n=<minumum_sample_size> # default=500
)

result.score   # To get the FMD-Inf score
result.r2      # To get the R^2 of FMD-Inf linear regression
result.slope   # To get the slope of the regression
result.points  # To get the point estimates used in FMD-Inf regression

Cleanup

Additionaly the pre-computed cache can be cleared like so:

from frechet_music_distance import FrechetMusicDistance

metric = FrechetMusicDistance()
metric.clear_cache()

Supported Models

Model	Name in library	Description	Creator
CLaMP2	clamp2	CLaMP 2: Multimodal Music Information Retrieval Across 101 Languages Using Large Language Models	sanderwood

Citation

If you use Frecheet Music Distance in your research, please cite the following paper:

@article{retkowski2024frechet,
  title={Frechet Music Distance: A Metric For Generative Symbolic Music Evaluation},
  author={Retkowski, Jan and St{\k{e}}pniak, Jakub and Modrzejewski, Mateusz},
  journal={arXiv preprint arXiv:2412.07948},
  year={2024}
}

Acknowledgements

This library uses code from the following repositories for handling the embedding models:

• CLaMP 2: sanderwood/clamp2

License

This project is licensed under the MIT License. See the LICENSE file for details.

---