RecStudio is a unified, highly-modularized and recommendation-efficient recommendation library based on PyTorch.

RecStudio is also equipped with a web service, where the recommendation pipeline can be quickly established and visually evaluated on selected datasets, and the evaluation results are automatically archived and visualized in a leaderboard.

Modular Model Design and Unified Dataset Processing

Customize your model like building blocks using the modules in RecStudio, such as query/item encoders, loss functions, scorers, samplers and so on.

Get different types of datasets for different types of models through unified dataset config.

  • General Dataset Structure
  • Modular Model Structure
  • GPU Acceleration
  • Simple Model Categorization
  • Simple and Complex Negative Samplers

Run Models Quickly

RecStudio is also equipped with a web service.
Select one dataset and drag models to form your own recommendation pipline quickly, and then you can run it visually.

Query the Leaderboard

We save some results of models on different datasets into database.
Select a dataset and a metric, you can see the leaderboard of models on this dataset.