Communication-Efficient TeraByte-Scale Model Training Framework for Online Advertising

by   Weijie Zhao, et al.

Click-Through Rate (CTR) prediction is a crucial component in the online advertising industry. In order to produce a personalized CTR prediction, an industry-level CTR prediction model commonly takes a high-dimensional (e.g., 100 or 1000 billions of features) sparse vector (that is encoded from query keywords, user portraits, etc.) as input. As a result, the model requires Terabyte scale parameters to embed the high-dimensional input. Hierarchical distributed GPU parameter server has been proposed to enable GPU with limited memory to train the massive network by leveraging CPU main memory and SSDs as secondary storage. We identify two major challenges in the existing GPU training framework for massive-scale ad models and propose a collection of optimizations to tackle these challenges: (a) the GPU, CPU, SSD rapidly communicate with each other during the training. The connections between GPUs and CPUs are non-uniform due to the hardware topology. The data communication route should be optimized according to the hardware topology; (b) GPUs in different computing nodes frequently communicates to synchronize parameters. We are required to optimize the communications so that the distributed system can become scalable. In this paper, we propose a hardware-aware training workflow that couples the hardware topology into the algorithm design. To reduce the extensive communication between computing nodes, we introduce a k-step model merging algorithm for the popular Adam optimizer and provide its convergence rate in non-convex optimization. To the best of our knowledge, this is the first application of k-step adaptive optimization method in industrial-level CTR model training. The numerical results on real-world data confirm that the optimized system design considerably reduces the training time of the massive model, with essentially no loss in accuracy.


Distributed Hierarchical GPU Parameter Server for Massive Scale Deep Learning Ads Systems

Neural networks of ads systems usually take input from multiple resource...

Heterogeneous Acceleration Pipeline for Recommendation System Training

Recommendation systems are unique as they show a conflation of compute a...

ZeRO-Offload: Democratizing Billion-Scale Model Training

Large-scale model training has been a playing ground for a limited few r...

More Bang for Your Buck: Improved use of GPU Nodes for GROMACS 2018

We identify hardware that is optimal to produce molecular dynamics traje...

Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training

The data ingestion pipeline, responsible for storing and preprocessing t...

M6-10T: A Sharing-Delinking Paradigm for Efficient Multi-Trillion Parameter Pretraining

Recent expeditious developments in deep learning algorithms, distributed...

Effective GPU Parallelization of Distributed and Localized Model Predictive Control

To effectively control large-scale distributed systems online, model pre...

Please sign up or login with your details

Forgot password? Click here to reset