Understanding Training Efficiency of Deep Learning Recommendation Models at Scale

Abstract

The use of GPUs has proliferated for machine learning workflows and is now considered mainstream for many deep learning models. Meanwhile, when training state-of-the-art personal recommendation models, which consume the highest number of compute cycles at our large-scale datacenters, the use of GPUs came with various challenges due to having both compute-intensive and memory-intensive components. GPU performance and efficiency of these recommendation models are largely affected by model architecture configurations such as dense and sparse features, MLP dimensions. Furthermore, these models often contain large embedding tables that do not fit into limited GPU memory. The goal of this paper is to explain the intricacies of using GPUs for training recommendation models, factors affecting hardware efficiency at scale, and learnings from a new scale-up GPU server design, Zion.

To appear in IEEE International Symposium on High-Performance Computer Architecture (HPCA 2021)

References in corpus (11)

• Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
• Deep Learning Recommendation Model for Personalization and Recommendation Systems
• Extremely Large Minibatch SGD: Training ResNet-50 on ImageNet in 15 Minutes
• MLPerf Training Benchmark
• Distributed Hierarchical GPU Parameter Server for Massive Scale Deep Learning Ads Systems
• Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems
• How to scale distributed deep learning?
• DeepRecSys: A System for Optimizing End-To-End At-scale Neural Recommendation Inference
• Mixed Dimension Embeddings with Application to Memory-Efficient Recommendation Systems
• Developing a Recommendation Benchmark for MLPerf Training and Inference
• CPR: Understanding and Improving Failure Tolerant Training for Deep Learning Recommendation with Partial Recovery