Asteroid: the PyTorch-based audio source separation toolkit for researchers

This paper describes Asteroid, the PyTorch-based audio source separation for researchers. Inspired by the most successful neural source systems, it provides all neural building blocks required to build a system. To improve reproducibility, Kaldi-style recipes on common audio separation datasets are also provided. This paper describes the software of Asteroid and its most important features. By showing results obtained with Asteroid's recipes, we show that our are at least on par with most results reported in reference. The toolkit is publicly available at://github.com/mpariente/asteroid .