CN109478251A - 处理方法及加速装置 - Google Patents

处理方法及加速装置 Download PDF

Info

Publication number
CN109478251A
CN109478251A CN201880002821.5A CN201880002821A CN109478251A CN 109478251 A CN109478251 A CN 109478251A CN 201880002821 A CN201880002821 A CN 201880002821A CN 109478251 A CN109478251 A CN 109478251A
Authority
CN
China
Prior art keywords
weight
neural network
less
instruction
equal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201880002821.5A
Other languages
English (en)
Other versions
CN109478251B (zh
Inventor
不公告发明人
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui Cambricon Information Technology Co Ltd
Original Assignee
Shanghai Cambricon Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN201710370905.1A external-priority patent/CN108960420B/zh
Priority claimed from CN201710456759.4A external-priority patent/CN109146069B/zh
Priority claimed from CN201710677987.4A external-priority patent/CN109389218B/zh
Priority claimed from CN201710678038.8A external-priority patent/CN109389208B/zh
Application filed by Shanghai Cambricon Information Technology Co Ltd filed Critical Shanghai Cambricon Information Technology Co Ltd
Priority to CN201910474387.7A priority Critical patent/CN110175673B/zh
Priority claimed from PCT/CN2018/088033 external-priority patent/WO2018214913A1/zh
Publication of CN109478251A publication Critical patent/CN109478251A/zh
Application granted granted Critical
Publication of CN109478251B publication Critical patent/CN109478251B/zh
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Neurology (AREA)
  • Feedback Control In General (AREA)
  • Image Analysis (AREA)

Abstract

本公开提供了一种处理装置，包括粗粒度剪枝单元，用于对神经网络的权值进行粗粒度剪枝，以得到剪枝后的权值；运算单元，用于根据剪枝后的权值对所述神经网络进行训练；其中，所述粗粒度剪枝单元具体用于：通过滑动窗口从神经网络的权值中选取出M个权值；当所述M个权值满足预设条件时，将所述M个权值的全部或者部分置为零。上述处理装置能减少访存同时减少运算量，从而获得加速比并降低能耗。

Description

PCT国内申请，说明书已公开。

Claims (22)

  1. PCT国内申请，权利要求书已公开。
CN201880002821.5A 2017-05-23 2018-05-23 处理方法及加速装置 Active CN109478251B (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910474387.7A CN110175673B (zh) 2017-05-23 2018-05-23 处理方法及加速装置

Applications Claiming Priority (9)

Application Number Priority Date Filing Date Title
CN201710370905.1A CN108960420B (zh) 2017-05-23 2017-05-23 处理方法及加速装置
CN2017103709051 2017-05-23
CN201710456759.4A CN109146069B (zh) 2017-06-16 2017-06-16 运算装置、运算方法和芯片
CN2017104567594 2017-06-16
CN201710677987.4A CN109389218B (zh) 2017-08-09 2017-08-09 数据压缩方法及压缩装置
CN2017106780388 2017-08-09
CN201710678038.8A CN109389208B (zh) 2017-08-09 2017-08-09 数据的量化装置及量化方法
CN2017106779874 2017-08-09
PCT/CN2018/088033 WO2018214913A1 (zh) 2017-05-23 2018-05-23 处理方法及加速装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
CN201910474387.7A Division CN110175673B (zh) 2017-05-23 2018-05-23 处理方法及加速装置

Publications (2)

Publication Number Publication Date
CN109478251A true CN109478251A (zh) 2019-03-15
CN109478251B CN109478251B (zh) 2021-01-05

Family

ID=65678543

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201880002821.5A Active CN109478251B (zh) 2017-05-23 2018-05-23 处理方法及加速装置

Country Status (1)

Country Link
CN (1) CN109478251B (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111626430A (zh) * 2019-04-18 2020-09-04 中科寒武纪科技股份有限公司 一种数据处理方法及相关产品
CN112115095A (zh) * 2020-06-12 2020-12-22 苏州浪潮智能科技有限公司 一种用于哈希算法的可重构硬件及运行方法
CN113253724A (zh) * 2021-04-30 2021-08-13 深圳市优必选科技股份有限公司 一种步态规划方法、装置、计算机可读存储介质及机器人
CN114217688A (zh) * 2021-12-24 2022-03-22 中国科学技术大学 一种基于神经网络结构的npu功耗优化系统和方法

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105512723A (zh) * 2016-01-20 2016-04-20 南京艾溪信息科技有限公司 一种用于稀疏连接的人工神经网络计算装置和方法
CN105787500A (zh) * 2014-12-26 2016-07-20 日本电气株式会社 基于人工神经网络的特征选择方法和装置

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105787500A (zh) * 2014-12-26 2016-07-20 日本电气株式会社 基于人工神经网络的特征选择方法和装置
CN105512723A (zh) * 2016-01-20 2016-04-20 南京艾溪信息科技有限公司 一种用于稀疏连接的人工神经网络计算装置和方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
SONG HAN, ET AL.: "Deep Compression: Compression Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding", 《ARXIV:1510.00149V5》 *
黄红梅，胡寿松: "基于在线学习RBF神经网络的故障预报", 《南京航空航天大学学报》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111626430A (zh) * 2019-04-18 2020-09-04 中科寒武纪科技股份有限公司 一种数据处理方法及相关产品
US11762690B2 (en) 2019-04-18 2023-09-19 Cambricon Technologies Corporation Limited Data processing method and related products
CN111626430B (zh) * 2019-04-18 2023-09-26 中科寒武纪科技股份有限公司 一种数据处理方法及相关产品
CN112115095A (zh) * 2020-06-12 2020-12-22 苏州浪潮智能科技有限公司 一种用于哈希算法的可重构硬件及运行方法
CN112115095B (zh) * 2020-06-12 2022-07-08 苏州浪潮智能科技有限公司 一种用于哈希算法的可重构硬件及运行方法
CN113253724A (zh) * 2021-04-30 2021-08-13 深圳市优必选科技股份有限公司 一种步态规划方法、装置、计算机可读存储介质及机器人
CN113253724B (zh) * 2021-04-30 2024-05-21 深圳市优必选科技股份有限公司 一种步态规划方法、装置、计算机可读存储介质及机器人
CN114217688A (zh) * 2021-12-24 2022-03-22 中国科学技术大学 一种基于神经网络结构的npu功耗优化系统和方法
CN114217688B (zh) * 2021-12-24 2024-05-10 中国科学技术大学 一种基于神经网络结构的npu功耗优化系统和方法

Also Published As

Publication number Publication date
CN109478251B (zh) 2021-01-05

Similar Documents

Publication Publication Date Title
EP3657340B1 (en) Processing method and accelerating device
CN111221578B (zh) 计算装置以及计算方法
EP3786786A1 (en) Processing device, processing method, chip, and electronic apparatus
CN109478251A (zh) 处理方法及加速装置
CN109032669A (zh) 神经网络处理装置及其执行向量最小值指令的方法
CN109902816A (zh) 板卡和神经网络运算方法
CN110163356A (zh) 一种计算装置及方法
WO2019129070A1 (zh) 一种集成电路芯片装置
US11562216B2 (en) Integrated circuit chip apparatus
CN109389208A (zh) 数据的量化装置及量化方法
CN108960420B (zh) 处理方法及加速装置
CN109389218A (zh) 数据压缩方法及压缩装置
CN109389209A (zh) 处理装置及处理方法
CN110175673B (zh) 处理方法及加速装置
CN109977446A (zh) 集成电路芯片装置及相关产品
WO2019165940A1 (zh) 集成电路芯片装置、板卡及相关产品

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
TA01 Transfer of patent application right

Effective date of registration: 20201207

Address after: Room 611-194, R & D center building, China (Hefei) international intelligent voice Industrial Park, 3333 Xiyou Road, hi tech Zone, Hefei City, Anhui Province

Applicant after: Anhui Cambrian Information Technology Co., Ltd

Address before: 201306 floor 6, block B, 168 Tonghui Road, Pudong New Area, Shanghai.

Applicant before: Shanghai Cambricon Information Technology Co.,Ltd.

TA01 Transfer of patent application right
GR01 Patent grant
GR01 Patent grant