Welcome to fastabx documentation¶
fastabx is a Python package for efficient computation of ABX discriminability.
The ABX discriminability is measures how well categories of interest are separated in the representation space by determining whether tokens from the same category are closer to each other than to those from a different category. While ABX has been mostly used to evaluate speech representations, it is a generic framework that can be applied to other domains of representation learning.
This package provides a simple interface that can be adapted to any ABX conditions, and to any input modality.
Install¶
Install the package with pip:
$ pip install fastabx
fastabx requires Python 3.12 or later, and depends on PyTorch 2.10.0 or later, NumPy, Polars, tqdm, and torchdtw.
fastabx is available on Linux x86-64 (with glibc 2.34 or later [1]), macOS arm64, and Windows x86-64.
Citation¶
A preprint is available on arXiv: https://arxiv.org/abs/2505.02692
If you use fastabx in your work, please cite it:
@misc{fastabx,
title={fastabx: A library for efficient computation of ABX discriminability},
author={Maxime Poli and Emmanuel Chemla and Emmanuel Dupoux},
year={2025},
eprint={2505.02692},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2505.02692},
}