CS 670 Project - Finetuning Language Models

Milestone-3 notebook: https://github.com/aye-thuzar/CS670Project/blob/milestone-3/CS670_milestone_3_AyeThuzar.ipynb

Hugging Face App:

Landing Page for the App:

App Demonstration Video:

Summary

milestone1: https://github.com/aye-thuzar/CS670Project/blob/main/README_milestone_1.md

milestone2: https://github.com/aye-thuzar/CS670Project/blob/main/README_milestone-2.md

Dataset: https://github.com/suzgunmirac/hupd

Data Preprocessing

I used the load_dataset function to load all the patent applications that were filed to the USPTO in January 2016. We specify the date ranges of the training and validation sets as January 1-21, 2016 and January 22-31, 2016, respectively. This is a smaller dataset.

There are two datasets: train and validation. Here are the steps I did:

Label-to-index mapping for the decision status field
map the 'abstract' and 'claims' sections
format them
use DataLoader with batch_size = 16

milestone3:

milestone3 notebook:

milestone4:

Please see Milestone4Documentation.md:

Here is the landing page for my app:

References: