Show HN: Local pretrained classifiers, GPU not needed

Category: ai-ml

Tags: machine-learning, cpu-inference, text-classification, multilingual, logistic-regression, pretrained-models

Score: 6.8/10 (Innovation: 6, Technical: 6, Documentation: 8, Utility: 7)

Jeffy provides 68 pretrained text classifiers (including 51 multilingual intent classifiers) that run on CPU using logistic regression on top of frozen sentence embeddings. It also allows users to train custom classifiers with a simple CLI or Python API, and includes an HTTP server for easy deployment. The project fills a niche for lightweight, offline, GPU-free classification with strong documentation and benchmark results.

Target audience: backend developers, data scientists, ML engineers, and anyone needing lightweight local text classification without GPU

Repository: https://github.com/nicobrenner/jeffy · Python · MIT · 106 stars

View on Hacker News