Skip to content

A set of fine tunes for t5-small, specifically on question generation and answering on SQuAD dataset

Notifications You must be signed in to change notification settings

vpgits/t5-mistral-fintetune-cookbook

Repository files navigation

SQuAD Question and Answer Generation

This repository contains Jupyter notebooks for generating questions, answers, and context-based questions and answers using the Stanford Question Answering Dataset (SQuAD). The notebooks demonstrate how to work with the SQuAD dataset and generate natural language questions and answers.

Notebooks

  1. Question Generation: This notebook focuses on generating questions from given text passages. It uses the SQuAD dataset to fine-tune a model for question generation.

  2. Answer Generation: In this notebook, you can generate answers to questions based on a given context. It uses pre-trained models to identify answers within the context.

  3. Context-Based QA: The context-based question and answer generation notebook combines the previous two functionalities. Given a context, it generates questions and answers. This is particularly useful for tasks like question-answering chatbots.

  4. Chain Of Thought QLoRA based QA : Mistral 7B non instruct model finetuned on the SQUAD Dataset with comparable results. Trained till 9000 steps. (Link)[https://huggingface.co/vpgits/Mistral-7B-v0.1-qagen-v0.3]

About

A set of fine tunes for t5-small, specifically on question generation and answering on SQuAD dataset

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published