NNSVS on Docker on WSL 2

Installation

First, you have to install WSL 2 (Windows Subsystem for Linux 2) and set up CUDA Toolkit and Docker on it. See CUDA on WSL User Guide.

Running NNSVS

Use the Docker Compose script to build NNSVS image from Dockerfile.

docker-compose build

Download kiritan_singing.zip from 研究者向け東北きりたん歌唱データベース　ログインページ and unzip to ./kiritan_singing/.

Start up JupyterLab by running:

docker-compose up

Open http://localhost:8888/ to access the notebook.

References

• CUDA on WSL User Guide
• 待ってました CUDA on WSL 2
• r9y9/nnsvs: Neural network-based singing voice synthesis library for research
• Google Colaboratory で NNSVS で遊ぶ mini-HOWTO
• oatsu-gh/setup-nnsvs-on-wsl: WSL 上で NNSVS が動く環境を作るためのシェルスクリプトとか