This is an unofficial implementation of MT3: Multi-Task Multitrack Music Transcription in pytorch.
from inference import InferenceHandler
handler = InferenceHandler('./pretrained')
handler.inference('music.mp3')
# training not done yet
@article{gardner2021mt3,
title={MT3: Multi-Task Multitrack Music Transcription},
author={Gardner, Josh and Simon, Ian and Manilow, Ethan and Hawthorne, Curtis and Engel, Jesse},
journal={arXiv preprint arXiv:2111.03017},
year={2021}
}