GitHub - neospe/autofit2: Automated end-to-end data preprocessing, model training, and evaluation pipeline

▤ Summary

This project introduces an automated end-to-end pipeline for data preprocessing, model training, evaluation, and deployment. It targets few-shot text classification and supports more than 50 languages through SetFit and SBERT embeddings. The workflow runs from a single JSON config and generates deployable model artifacts, model cards, and CO emission tracking. It also includes reproducibility features such as resumable execution and configurable loaders for training and benchmark targets.

▦ Linked Companies