midiocrity

Populate the ./data folder

Run the following from the repo root to download the Clean MIDI subset of The Lakh MIDI Dataset

cd data
./getData.sh

Environment Setup

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

Training the model

To train the model, navigate to the midiocrity folder, and run:

python train.py

Architecture

We are using a modified version of the architecture referenced in the Off the Beaten Track paper.