When LLMs are Unfit Use FastFit: Fast and Effective Text Classification with Many Classes

TL;DR

FastFit is a new method and Python package that enables rapid and accurate few-shot text classification for many classes, outperforming existing methods in speed and accuracy, especially in multiclass scenarios.

Contribution

FastFit introduces a novel combination of batch contrastive learning and token-level similarity for efficient multiclass classification, with a publicly available implementation.

Findings

3-20x faster training than existing methods

Significantly improved accuracy in multiclass classification

Effective on both English and multilingual datasets

Abstract

We present FastFit, a method, and a Python package design to provide fast and accurate few-shot classification, especially for scenarios with many semantically similar classes. FastFit utilizes a novel approach integrating batch contrastive learning and token-level similarity score. Compared to existing few-shot learning packages, such as SetFit, Transformers, or few-shot prompting of large language models via API calls, FastFit significantly improves multiclass classification performance in speed and accuracy across FewMany, our newly curated English benchmark, and Multilingual datasets. FastFit demonstrates a 3-20x improvement in training speed, completing training in just a few seconds. The FastFit package is now available on GitHub and PyPi, presenting a user-friendly solution for NLP practitioners.