MONOLITH

Summary from the TLDR newsletter

Monolith is a deep learning framework for large scale recommendation modeling. It features collisionless embedding tables, which guarantees unique representation for different ID features, and real time training, which captures the latest hotspots and helps users to discover new interests rapidly. Monolith supports batch/real-time training and serving.