Merlin HugeCTR: GPU-accelerated Recommender System Training and Inference

TL;DR

Merlin HugeCTR is an open-source GPU-accelerated framework that significantly speeds up training and inference of recommendation models, enabling scalable, low-latency online deployment with high performance on single and multi-node systems.

Contribution

It introduces a novel GPU-optimized architecture with hierarchical storage and a parameter server, achieving substantial speedups over CPU-based methods for recommendation systems.

Findings

Up to 24.6x speedup on MLPerf benchmark with a single DGX A100

5-62x inference speedup over CPU implementations using hierarchical parameter server

Effective multi-node training acceleration for large-scale recommendation models

Abstract

In this talk, we introduce Merlin HugeCTR. Merlin HugeCTR is an open source, GPU-accelerated integration framework for click-through rate estimation. It optimizes both training and inference, whilst enabling model training at scale with model-parallel embeddings and data-parallel neural networks. In particular, Merlin HugeCTR combines a high-performance GPU embedding cache with an hierarchical storage architecture, to realize low-latency retrieval of embeddings for online model inference tasks. In the MLPerf v1.0 DLRM model training benchmark, Merlin HugeCTR achieves a speedup of up to 24.6x on a single DGX A100 (8x A100) over PyTorch on 4x4-socket CPU nodes (4x4x28 cores). Merlin HugeCTR can also take advantage of multi-node environments to accelerate training even further. Since late 2021, Merlin HugeCTR additionally features a hierarchical parameter server (HPS) and supports deployment via the NVIDIA Triton server framework, to leverage the computational capabilities of GPUs for high-speed recommendation model inference. Using this HPS, Merlin HugeCTR users can achieve a 5~62x speedup (batch size dependent) for popular recommendation models over CPU baseline implementations, and dramatically reduce their end-to-end inference latency.