TensorFlow Recommenders

An open source library for building, evaluating, and serving sophisticated recommender system models using TensorFlow 2.x.

TensorFlow Recommenders (TFRS) streamlines the full recommendation lifecycle by providing modular components for two-stage retrieval and ranking architectures. It integrates seamlessly with the Keras API to handle complex tasks like multi-task learning, feature preprocessing, and ScaNN-based approximate nearest neighbor search. Developers use TFRS to build production-ready systems that scale across millions of items, leveraging specialized layers for Factorized Top-K metrics and DCN V2 (Deep & Cross Network) architectures to capture high-order feature interactions.

https://github.com/tensorflow/recommenders