HNPU: An Adaptive DNN Training Processor Utilizing Stochastic Dynamic Fixed-Point and Active Bit-Precision Searching

Autor:	Seokchan Song, Juhyoung Lee, Hoi-Jun Yoo, Donghyeon Han, Dongseok Im, Youngwoo Kim, Gwangtae Park
Rok vydání:	2021
Předmět:	Computer engineering Artificial neural network Computer science Adaptive system Bandwidth (signal processing) Process (computing) Reconfigurability System on a chip Throughput Electrical and Electronic Engineering Efficient energy use
Zdroj:	IEEE Journal of Solid-State Circuits. 56:2858-2869
ISSN:	1558-173X 0018-9200
Popis:	This article presents HNPU, which is an energy-efficient deep neural network (DNN) training processor by adopting algorithm-hardware co-design. The HNPU supports stochastic dynamic fixed-point representation and layer-wise adaptive precision searching unit for low-bit-precision training. It additionally utilizes slice-level reconfigurability and sparsity to maximize its efficiency both in DNN inference and training. Adaptive bandwidth reconfigurable accumulation network enables reconfigurable DNN allocation and maintains its high core utilization even in various bit-precision conditions. Fabricated in a 28-nm process, the HNPU accomplished at least $5.9\times $ higher energy efficiency and $2.5\times $ higher area efficiency in actual DNN training compared with the previous state-of-the-art on-chip learning processors.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::810593c84b6733a41f87fd519cf951ec https://doi.org/10.1109/jssc.2021.3066400 Zobrazit plný text záznamu