Chainer is a minor DL library so many reseachers can't run my code. So I prepared some notebooks. You can reproduce my results with only a browser and no need to install chainer or other packages.
You need two notebooks.
- train.ipynb (for training)
- generate.ipynb (for generating)
Download each file or clone this repository.
Upload these notebooks on Google Drive. The dataset(CMU ARCTIC) will be downloaded into the instance of Colaboratory so it doesn't need your strage of Google Drive for dataset. But traind model's parameters, losses and etc. are saved into your Google Drive so check your starage.
Now you can train on Colaboratory. Open train.ipynb via Colaboratory and run cells. You need authentications to mount Google Drive. Please follow the indications.
You can resume training from a snapshot. Please check comments in the notebook.
Open generate.ipynb via Colaboratory and run cells. You can use trained models in your Google Drive.