The pytorch implementation for Conditional RhythmGAN+TransformerKernal

This project is the pytorch implemtation for Conditional Generative Adversarial Network(CGAN) to generate the rhythm. Please cite the original work from Nao Tokui for more detail: https://cclab.sfc.keio.ac.jp/projects/rhythmcan/

The Keras implemntation:https://github.com/naotokui/CreativeGAN-Rhythm

Conditional Generative Adversarial Nets: https://arxiv.org/abs/1411.1784

Besides the LSTM kernal, we are inspired by Attention Is All You Need and provide Transformer kernal via transformer encoder block.

There are two options for training, either Google Colab or your local machine.

• Train RhythmGAN using LSTM Kernal

• Train RhythmGAN using Transformer Kernal

Download the dataset

wget https://www.dropbox.com/s/a8mwk8rdv08cu2l/data.zip

unzip data.zip

Configure the environment

pip install -r requirements.txt

sudo apt-get install fluidsynth

Run training script

python main.py --kernal=LSTM

python main.py --kernal=Transformer

Please refer Play_music.ipynb and Save_music.ipynb

OR, Use the interactive interface: https://github.com/naotokui/M4L-CreativeGAN-Rhythm/