We used LLMD-FULL from Lakh MIDI Dataset v0.1. The compressed file name is lmd_full.tar.gz. extract the file under data/.
>wget **link**
>tar -zxvf **filename.tar.gz**
We have filtered midi files to contain lyrics, and split the latter into four sections. You can download our dataset called LyricalLakh.zip
Make sure to make a python virtual environment
python -m venv .venv
Activate venv. this should make your terminal start with (.venv)
source .venv/bin/activate
Windows?
> cd .venv/Scripts/
> activate.bat
> cd ../../
Now install dependencies in that venv
> pip install -r requirements.txt
Shortcut:
> python -m venv .venv && source .venv/bin/activate && pip install -r requirements.txt
Make sure before running scripts, that you are in the path
EnsembleEmbedders/
not in
EnsembleEmbedders/src/
prettymidi ==
tensorflow == 2.16.1
joblib == 1.4.0