Artifex
A Python library for running and fine-tuning small, task-specific language models on CPU, generating synthetic training data instead of requiring a labeled dataset.
Artifex is a Python library for using pre-trained, task-specific small language models on CPU and for fine-tuning them on CPU without any training data of your own, based only on instructions describing the task. It generates synthetic training data on the fly from those instructions and uses it to fine-tune the model, and it also tracks model performance locally with built-in evaluation and monitoring.
It ships pre-trained models for tasks such as guardrail moderation, intent classification, reranking, sentiment analysis, emotion detection, named entity recognition, text anonymization and spam detection, in English and, for some tasks, German, Spanish or Italian. The default models are around 100 million parameters and 500MB, small enough to run without a GPU or an external API.