Using a few Touhou music MIDIs, this project trains simple Generative Adversarial Networks in order to generate music.
In order to install the required python packages, use:
pip install requirements.txt
In order to prepare the music dataset, use the command:
cd data
python midi.py
All the parameters can be changed in the config.py file. Afterwards simply run the following command (from the main repository):
python train.py
This will create a checkpoint folder in checkpoints/[experiment name]/. Warning: the code was written for GPU training, if you don't have a GPU or too little VRAM, you will have to remove all the .cuda() from the code.
If you want to listen to created music (MIDI format) from the training that just finished, run:
python test.py
If you want to generate music using a previous training, you can specify the .pth checkpoint path:
python test.py --checkpoint [your G.pth checkpoint]