arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.14912cs.ARcs.LG
收藏

ML流水线的高效表格数据预处理

Efficient Tabular Data Preprocessing of ML Pipelines

ETH Zurich(苏黎世联邦理工学院)

Yu Zhu, Wenqi Jiang, Gustavo Alonso

AI总结:

本文提出FPGA硬件加速器Piper用于表格数据预处理,解决CPU与GPU性能差距导致的瓶颈,在商业推荐系统训练中实现比128核CPU最高71.3倍加速。

摘要:

数据预处理流水线(包括数据解码、清理和转换)是机器学习(ML)训练的关键组件。由于用于预处理的CPU与用于模型训练的GPU之间的性能差距日益扩大,其计算密集且常成为主要瓶颈。近期研究表明,需要跨多台机器的大量CPU才能实现足够吞吐量以使GPU饱和,导致资源和能耗增加。当流水线涉及词汇表生成时,由于不同CPU核心和服务器间存在显著的逐行同步开销,预处理性能扩展性较差。为解决此局限,本文提出硬件加速器Piper用于表格数据预处理,在FPGA上构建原型,并展示其在商业推荐系统训练流水线中的潜力。使用二进制输入时,Piper在延迟上比128核CPU服务器实现4.7~71.3倍加速,并优于数据中心GPU 4.8~20.3倍。此性能彰显了Piper提升预处理流水线效率并显著降低资源消耗的潜力。

英文摘要:

Data preprocessing pipelines, which includes data decoding, cleaning, and transforming, are a crucial component of Machine Learning (ML) training. Thy are computationally intensive and often become a major bottleneck, due to the increasing performance gap between the CPUs used for preprocessing and the GPUs used for model training. Recent studies show that a significant number of CPUs across several machines are required to achieve sufficient throughput to saturate the GPUs, leading to increased resource and energy consumption. When the pipeline involves vocabulary generation, the preprocessing performance scales poorly due to significant row-wise synchronization overhead between different CPU cores and servers. To address this limitation, in this paper we present the design of Piper, a hardware accelerator for tabular data preprocessing, prototype it on FPGAs, and demonstrate its potential for training pipelines of commercial recommender systems. Piper achieves 4.7 $\sim$ 71.3$\times$ speedup in latency over a 128-core CPU server and outperforms a data-center GPU by 4.8$\sim$ 20.3$\times$ when using binary input. The impressive performance showcases Piper's potential to increase the efficiency of data preprocessing pipelines and significantly reduce their resource consumption.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑