GRAM [NAACL 2022 Main, Oral]

Official PyTorch Implementation of GRAM: Fast Fine-tuning of Pre-trained Language Models for Content-based Collaborative Filtering. Along with Knowledge Tracing and News Recommendation codebase, we provide a simple MNIST example to play around with GRAM.

Below gram folder, you can find respective models for knowledge tracing and news recommendation in knowledge_tracing and news_recommendation repo. To run models, use run_script.py for knowledge tracing and run_pl_nrms.py for news recommendation. Before running the script, we recommend installing the gram package with:

pip install -r requirements.txt
pip install -e gram

Datasets can be found here.

If you find this repo useful, please cite our work! The proceedings bibtex will be available soon.