Run the following from the repo root to download the Clean MIDI subset of The Lakh MIDI Dataset
cd data
./getData.sh
To add the environment with conda, use the following commands:
conda create -n midiocrity python=3.9
conda activate midiocrity
pip install -r requirements.txt
To train the model, navigate to the midiocrity folder, and run:
python train.py
We are using a modified version of the architecture referenced in the Off the Beaten Track paper.
