MusiComb

MusiComb is a program born as an extension of Hyun et al., 2022 (paper | code) and aimed at tackling the combinatorial music generation task.

Setup

1. Clone the repo and cd into its directory:

   $ git clone https://github.com/alitavanaali/MusiComb.git
   $ cd midicomb
   

2. Create a virtual environment and install the required packages:

   $ python -m venv .venv
   $ source .venv/bin/activate
   $ pip install -r requirements.txt
   

3. Unzip the dataset:

   $ tar -xvf dataset/commu_midi.tar -C dataset/
   $ unzip dataset/groove_drum.zip -d dataset/
   

   You should get the following directory structure:

   dataset
   ├── commu_meta.csv
   ├── groove_drum
   │   ├── drummer1
   │   ├── drummer2
   │   ├── ...
   │   ├── drummer10
   ├── commu_midi
   │   ├── train
   │   │   └── raw
   │   │       └── midifiles(.mid)
   │   └── val
   │       └── raw
   │           └── midifiles(.mid)
   ├── commu_midi.tar
   └── README.md
   

4. [OPTIONAL] If you want the samples to be generated (see the following section), download the model weights from here and move them into the ckpt directory:

   ckpt
   ├── checkpoint_best.pt
   └── README.md
   

   Note: the weights are provided by Hyun et al. together with their implementation.

Run

Run app.py with its required arguments, e.g.

$ python app.py

Then you can open http://127.0.0.1:5000 in your browser to open the user interface of the application.

Once the program successfully terminates you will see screen below:

and then, you will find an out directory with the following structure:

out
└── <date>_<time>
    ├── metadata.yaml
    └── tune.mid
    └── tune_notmerged_sounds.mid

where metadata.yaml contains the arguments of the corresponding run and tune.mid is the generated MIDI file and 'tune_notmerged_sounds.mid' is separated tracks version.