Introduction

Recommendation models

Model-based recommendation systems play key roles in internet industry, from social network to e-commerce platform. Recommendation models are getting deeper in recent years, which makes training on GPUs a good choice.

However, industrial-scale recommendation models are not only deeper, but also much wider. Training wide-and-deep recommendation models on GPUs with real-world datasets still suffers from low utilization and high cost.

wide-and-deep

HybridBackend

HybridBackend is a high-performance framework for training wide-and-deep recommendation model on heterogeneous cluster.

HybridBackend provides following features:

Memory-efficient loading of categorical data
GPU-efficient orchestration of embedding layers
Communication-efficient training and evaluation at scale
Easy to use with existing AI workflows

HybridBackend speeds up training of wide-and-deep recommendation models dramatically:

performance