A modular framework for recommender systems

gh repo clone Jyonn/Legommenders
cd Legommenders
pip install -r requirements.txt  # Note: Legommenders is not compatible to the latest version of transformers yet if you want to finetune LLaMA-based models.

• LSTUR model is now re-added to the Legommenders package, which was not compatible from Jan. 2024.
• LLMs can be used for item encoder.

• Legommenders partially supports the flatten sequential recommendation model.
• New models are added, including: MaskNet, GDCN, etc.

• We clean the code and convert names of the item-side parameters.

• The first recommender system package, Legommenders, with a modular-design is released!
• Legommenders involves a set of recommender system algorithms, including:
  • Matching based methods: NAML, NRMS, LSTUR, etc.
  • Ranking based methods: DCN, DeepFM, PNN, etc.

Legommenders have served as a fundamental framework for several research projects, including ONCE, SPAR,GreenRec, and UIST. If you find Legommenders useful in your research, please consider citing our project:

@online{legommenders,
  author = {Liu, Qijiong},
  title = {Legommenders: A Modular Framework for Recommender Systems},
  year = {2023},
  url = {https://github.com/Jyonn/Legommenders}
}