Fine-tuning T5 for Question Answering

This repository contains a T5 (Text-to-Text Transfer Transformer) model fine-tuned on SQuAD v1.1 for extractive/span-based question answering.

Introduction

Setup

Note: requirements.txt is likely incorrect——torch 2 may be incompatible with sentence-transformers and you may need to downgrade.

To load in the trained model for use, see inference.py.

For points of contact, please directly contact the authors of this repository.

Data Acknowledgements

Stanford Question Answering Dataset (SQuAD) v1.1: Acquired on HuggingFace, owned by Stanford NLP Group, License: Creative Commons (cc-by-4.0) https://huggingface.co/datasets/squad