arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2206.13601cs.ARcs.LG
收藏

使用非易失性存储器技术的高效深度学习

Efficient Deep Learning Using Non-Volatile Memory Technology

Carnegie Mellon University(卡内基梅隆大学);Apple Inc.(苹果公司);The University of Texas at Austin(德克萨斯大学奥斯汀分校)

Ahmet Inci, Mehmet Meric Isgenc, Diana Marculescu

摘要:

嵌入式机器学习(ML)系统现在已经成为部署ML服务任务的主导平台,并且预计将成为训练ML模型的同等重要性。随之而来的挑战是在严格的内存约束下进行整体高效部署,特别是低功耗和高吞吐量的实现。在此上下文中,与常规SRAM相比,诸如STT-MRAM和SOT-MRAM的非易失性存储器(NVM)技术由于其非易失性、更高的单元密度和可缩放性特征而具有显著的优点。虽然之前的工作已经研究了NVM对通用应用的几种架构影响,但在这项工作中,我们提出了DeepNVM++,这是一个综合框架,通过结合特定技术的电路级模型和各种深度学习工作负载的实际内存行为,来表征,建模和分析GPU架构中基于NVM的缓存,用于深度学习(DL)应用。DeepNVM++依赖于使用传统SRAM和新兴STT-MRAM和SOT-MRAM技术实现的最后一级缓存的等容量和等面积性能和能量模型。在等容量情况下,与传统SRAM相比,STT-MRAM和SOT-MRAM分别提供高达3.8倍和4.7倍的能量延迟乘积(EDP)减少以及2.4倍和2.8倍的面积减少。在等面积假设下,与SRAM相比,STT-MRAM和SOT-MRAM分别提供高达2.2倍和2.4倍的EDP减少,并容纳2.3倍和3.3倍的缓存容量。我们还进行了可扩展性分析,并表明,STT-MRAM和SOT-MRAM实现的数量级EDP减少相比,SRAM的大缓存容量。DeepNVM++在STT-/SOT-MRAM技术上进行了演示,可用于DL应用中GPU末级缓存的任何NVM技术的表征、建模和分析。

英文摘要:

Embedded machine learning (ML) systems have now become the dominant platform for deploying ML serving tasks and are projected to become of equal importance for training ML models. With this comes the challenge of overall efficient deployment, in particular low power and high throughput implementations, under stringent memory constraints. In this context, non-volatile memory (NVM) technologies such as STT-MRAM and SOT-MRAM have significant advantages compared to conventional SRAM due to their non-volatility, higher cell density, and scalability features. While prior work has investigated several architectural implications of NVM for generic applications, in this work we present DeepNVM++, a comprehensive framework to characterize, model, and analyze NVM-based caches in GPU architectures for deep learning (DL) applications by combining technology-specific circuit-level models and the actual memory behavior of various DL workloads. DeepNVM++ relies on iso-capacity and iso-area performance and energy models for last-level caches implemented using conventional SRAM and emerging STT-MRAM and SOT-MRAM technologies. In the iso-capacity case, STT-MRAM and SOT-MRAM provide up to 3.8x and 4.7x energy-delay product (EDP) reduction and 2.4x and 2.8x area reduction compared to conventional SRAM, respectively. Under iso-area assumptions, STT-MRAM and SOT-MRAM provide up to 2.2x and 2.4x EDP reduction and accommodate 2.3x and 3.3x cache capacity when compared to SRAM, respectively. We also perform a scalability analysis and show that STT-MRAM and SOT-MRAM achieve orders of magnitude EDP reduction when compared to SRAM for large cache capacities. DeepNVM++ is demonstrated on STT-/SOT-MRAM technologies and can be used for the characterization, modeling, and analysis of any NVM technology for last-level caches in GPUs for DL applications.

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。

↑