recommenders-team/recommenders

Provides implementations of classical and deep learning algorithms (ALS, xDeepFM, DKN, sequential models) alongside Jupyter notebooks covering the full recommendation pipeline: data preparation, model training, offline evaluation, hyperparameter optimization, and Azure deployment. Includes utility functions for dataset loading, metric computation, and train/test splitting across multiple backends (CPU, GPU, PySpark), supporting both collaborative filtering and content-based approaches.

21,514 stars and 20,023 monthly downloads. Available on PyPI.