To reproduce the paper results, run the following in the terminal

It is recommended you have 4 GPUs for speed and performance
You will need to place your API code in run5 file
You will need to place you huggingface key in src/models.py
The code may need to be modified further depending on your precise setup.

pip install -r requirements.txt
mkdir datasets
cd datasets
mkdir 20newsgroups
mkdir agnews
mkdir amazon_reviews
mkdir dbpedia
mkdir imdb
mkdir trec

#### Download these dataset zip files to datasets/{dataset_name}
https://www.kaggle.com/datasets/ducanger/imdb-dataset
https://www.kaggle.com/datasets/crawford/20-newsgroups
https://www.kaggle.com/datasets/amananandrai/ag-news-classification-dataset
https://www.kaggle.com/datasets/abdallahwagih/amazon-reviews
https://www.kaggle.com/datasets/danofer/dbpedia-classes
https://www.kaggle.com/datasets/thedevastator/the-trec-question-classification-dataset-a-longi

python reproduce_paper.py
