miditok 1.4.3

A convenient MIDI tokenizer for Deep Learning networks, with multiple encoding strategies

MidiTok

Python package to tokenize MIDI music files, presented at the ISMIR 2021 LBD.

MidiTok converts MIDI music files into sequences of tokens, ready to be fed to sequential deep learning models like Transformers. MidiTok features most known MIDI tokenizations (e.g. REMI, Compound Word...), and is built around the idea that they all share common parameters and methods. It contains methods allowing to properly pre-process any MIDI file, and also supports Byte Pair Encoding (BPE).

Documentation: miditok.readthedocs.com

Install

pip install miditok

MidiTok uses MIDIToolkit, which itself uses Mido to read and write MIDI files.

Usage example

The most basic and useful methods are summurized here.

from miditok import REMI
from miditok.utils import get_midi_programs
from miditoolkit import MidiFile
from pathlib import Path

# Creates the tokenizer and loads a MIDI
tokenizer = REMI()  # using the default parameters, read the documentation to customize your tokenizer
midi = MidiFile('path/to/your_midi.mid')

# Converts MIDI to tokens, and back to a MIDI
tokens = tokenizer(midi)  # automatically detects MIDIs and tokens before converting
converted_back_midi = tokenizer(tokens, get_midi_programs(midi))  # PyTorch / Tensorflow / Numpy tensors supported

# Converts MIDI files to tokens saved as JSON files
midi_paths = list(Path('path', 'to', 'dataset').glob('**/*.mid'))
data_augmentation_offsets = [2, 2, 1]  # data augmentation on 2 pitch octaves, 2 velocity and 1 duration values
tokenizer.tokenize_midi_dataset(midi_paths, Path('path', 'to', 'tokens_noBPE'),
                                data_augment_offsets=data_augmentation_offsets)

# Constructs the vocabulary with BPE
tokenizer.learn_bpe(tokens_path=Path('path', 'to', 'tokens_noBPE'), vocab_size=500,
                    out_dir=Path('path', 'to', 'tokens_BPE'), files_lim=300)

# Converts the tokenized musics into tokens with BPE
tokenizer.apply_bpe_to_dataset(Path('path', 'to', 'tokens_noBPE'), Path('path', 'to', 'tokens_BPE'))

Tokenizations

• REMI
• MIDI-Like
• TSD
• Structured
• CPWord
• Octuple
• MuMIDI

Limitations

Tokenizations using Bar tokens (REMI, Compound Word and MuMIDI) only considers a 4/x time signature for now. This means that each bar is considered covering 4 beats.

Contributions

Contributions are gratefully welcomed, feel free to open an issue or send a PR if you want to add a tokenization or speed up the code. Just make sure to pass the tests.

Todo

• Time Signature
• Control Change messages
• Speeding up MIDI read / load (Rust / C++ binding)
• Speeding up BPE (Rust / C++ binding)
• Data augmentation on duration values at the MIDI level

Citation

If you use MidiTok for your research, a citation in your manuscript would be gladly appreciated. ❤️

MidiTok paper

@inproceedings{miditok2021,
    title={{MidiTok}: A Python package for {MIDI} file tokenization},
    author={Fradet, Nathan and Briot, Jean-Pierre and Chhel, Fabien and El Fallah Seghrouchni, Amal and Gutowski, Nicolas},
    booktitle={Extended Abstracts for the Late-Breaking Demo Session of the 22nd International Society for Music Information Retrieval Conference},
    year={2021},
    url={https://archives.ismir.net/ismir2021/latebreaking/000005.pdf},
}

The BibTeX citations of all tokenizations can be found in the documentation

Acknowledgments

We acknowledge Aubay, the LIP6, LERIA and ESEO for the financing and support of this project. Special thanks to all the contributors.