Scalable BSS Toolkit

SBSS is a research-oriented toolkit for scalable blind source separation, including neural FCA/FastFCA models, dataset recipes, and Lightning integrations.

Key Features

• Reproducible recipes – End-to-end pipelines document every stage (data prep, training, inference, evaluation) for reproducible studies.
• HPC ready – Recipes and Makefiles are tuned for ABCI, TSUBAME, and other clusters, yet remain runnable on a single workstation.
• Highly modular – Lightning tasks, common datasets, models, and utilities are built to swap components and run ablations with minimal friction.

Getting started

For only inference, you can install SBSS by using pip (or uv) as:

pip install git+https://github.com/b-sigpro/sbss

For development, we recommend to use Pixi for installing the dependencies:

git clone github.com:b-sigpro/sbss
pixi install

Full instructions are provided in https://sbss.readthedocs.io/en/latest/user_guide/index.html

Acknowledgement

• Part of this software was developed in a project commissioned by the New Energy and Industrial Technology Development Organization (NEDO).
• Part of this software was developed by using ABCI 3.0 provided by AIST and AIST Solutions.
• Part of this software was developed by using the TSUBAME4.0 supercomputer at Institute of Science Tokyo.