TWI857493B - 用於神經網路運算之電腦實施方法，系統以及非暫時性電腦可讀媒體 - Google Patents

用於神經網路運算之電腦實施方法，系統以及非暫時性電腦可讀媒體 Download PDF

Info

Publication number
TWI857493B
TWI857493B TW112105472A TW112105472A TWI857493B TW I857493 B TWI857493 B TW I857493B TW 112105472 A TW112105472 A TW 112105472A TW 112105472 A TW112105472 A TW 112105472A TW I857493 B TWI857493 B TW I857493B
Authority
TW
Taiwan
Prior art keywords
filters
ifm
array
tensor
convolution
Prior art date
Application number
TW112105472A
Other languages
English (en)
Chinese (zh)
Other versions
TW202343310A (zh
Inventor
張曉謙
嚴恩勖
肖志斌
Original Assignee
香港商墨子國際有限公司
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
Application filed by 香港商墨子國際有限公司 filed Critical 香港商墨子國際有限公司
Publication of TW202343310A publication Critical patent/TW202343310A/zh
Application granted granted Critical
Publication of TWI857493B publication Critical patent/TWI857493B/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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
    • G06N3/0495—Quantised networks; Sparse networks; Compressed networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Complex Calculations (AREA)
  • Image Processing (AREA)
TW112105472A 2022-02-16 2023-02-16 用於神經網路運算之電腦實施方法，系統以及非暫時性電腦可讀媒體 TWI857493B (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US17/673,490 US20230259758A1 (en) 2022-02-16 2022-02-16 Adaptive tensor compute kernel for sparse neural network
US17/673,490 2022-02-16

Publications (2)

Publication Number Publication Date
TW202343310A TW202343310A (zh) 2023-11-01
TWI857493B true TWI857493B (zh) 2024-10-01

Family

ID=87558678

Family Applications (1)

Application Number Title Priority Date Filing Date
TW112105472A TWI857493B (zh) 2022-02-16 2023-02-16 用於神經網路運算之電腦實施方法，系統以及非暫時性電腦可讀媒體

Country Status (7)

Country Link
US (1) US20230259758A1 (https=)
EP (1) EP4479887A4 (https=)
JP (1) JP2025505291A (https=)
KR (1) KR20240149907A (https=)
CN (1) CN118715527A (https=)
TW (1) TWI857493B (https=)
WO (1) WO2023155748A1 (https=)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116261736B (zh) * 2020-06-12 2024-08-16 墨芯国际有限公司 用于双稀疏卷积处理和并行化的方法和系统
CN112925644B (zh) * 2021-02-26 2024-08-13 北京小米松果电子有限公司 深度学习算子优化方法、装置、设备及存储介质
CN116662330A (zh) * 2022-02-21 2023-08-29 中兴通讯股份有限公司 数据处理方法、转发芯片、存储介质及程序产品
US12567122B1 (en) 2022-04-19 2026-03-03 Nvidia Corporation Application programming interface to modify tensor dimensions
US20230140173A1 (en) * 2022-08-19 2023-05-04 Arnab Raha Deep neural network (dnn) accelerators with heterogeneous tiling
TWI873681B (zh) * 2023-06-14 2025-02-21 緯創資通股份有限公司 物件檢測方法、機器學習方法及電子裝置
CN117707791B (zh) * 2024-02-02 2024-05-14 北京壁仞科技开发有限公司 用于进行注意力运算的方法、设备和存储介质
CN118152713B (zh) * 2024-05-10 2024-08-06 北京壁仞科技开发有限公司 数据处理方法、装置、电子设备和计算机可读存储介质
TWI884041B (zh) * 2024-07-19 2025-05-11 國立清華大學 基於混合精度演算法和記憶體內運算加速器之軟硬體協同運作方法及其系統及非暫態電腦可讀儲存媒體
CN121233886B (zh) * 2025-12-01 2026-03-20 上海壁仞科技股份有限公司 卷积计算方法、电子设备、存储介质及程序产品

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190147372A1 (en) * 2017-11-15 2019-05-16 Uber Technologies, Inc. Systems and Methods for Object Detection, Tracking, and Motion Prediction
US20200118307A1 (en) * 2018-10-10 2020-04-16 New York University System, method, and computer-accessible medium for generating magnetic resonance imaging-based anatomically guided positron emission tomography reconstruction images with a convolutional neural network
TW202014202A (zh) * 2018-06-01 2020-04-16 美商格瑞爾公司 用於資料分類之卷積神經網路系統及方法
US20200175095A1 (en) * 2018-11-29 2020-06-04 Adobe Inc. Object recognition and tagging based on fusion deep learning models

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10169298B1 (en) * 2017-05-11 2019-01-01 NovuMind Limited Native tensor processor, using outer product unit
US11443176B2 (en) * 2018-05-17 2022-09-13 International Business Machines Corporation Acceleration of convolutional neural networks on analog arrays
US11429850B2 (en) * 2018-07-19 2022-08-30 Xilinx, Inc. Performing consecutive mac operations on a set of data using different kernels in a MAC circuit
EP3654247B1 (en) * 2018-11-15 2025-01-01 IMEC vzw Convolution engine for neural networks
US11604958B2 (en) * 2019-03-13 2023-03-14 Samsung Electronics Co., Ltd. Method and apparatus for processing computation of zero value in processing of layers in neural network
WO2021071930A1 (en) * 2019-10-07 2021-04-15 Google Llc Redistributing tensor elements between machine learning computing units
US12554962B2 (en) * 2019-12-24 2026-02-17 Intel Corporation Configurable processor element arrays for implementing convolutional neural networks
CN115456161A (zh) * 2020-03-27 2022-12-09 华为技术有限公司 一种数据处理方法和数据处理系统
KR102914873B1 (ko) * 2020-12-14 2026-01-16 삼성전자 주식회사 채널 수에 기초하여 컨볼루션 연산을 수행하는 npu 장치 및 이의 동작 방법
KR102602584B1 (ko) * 2021-04-14 2023-11-16 한국전자통신연구원 인공 지능 반도체 프로세서 및 인공 지능 반도체 프로세서의 동작 방법
US20230195419A1 (en) * 2021-12-17 2023-06-22 Arm Limited System and Method for Accelerating Neural Networks

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190147372A1 (en) * 2017-11-15 2019-05-16 Uber Technologies, Inc. Systems and Methods for Object Detection, Tracking, and Motion Prediction
TW202014202A (zh) * 2018-06-01 2020-04-16 美商格瑞爾公司 用於資料分類之卷積神經網路系統及方法
US20200118307A1 (en) * 2018-10-10 2020-04-16 New York University System, method, and computer-accessible medium for generating magnetic resonance imaging-based anatomically guided positron emission tomography reconstruction images with a convolutional neural network
US20200175095A1 (en) * 2018-11-29 2020-06-04 Adobe Inc. Object recognition and tagging based on fusion deep learning models

Also Published As

Publication number Publication date
CN118715527A (zh) 2024-09-27
EP4479887A4 (en) 2026-01-07
EP4479887A1 (en) 2024-12-25
TW202343310A (zh) 2023-11-01
WO2023155748A1 (en) 2023-08-24
KR20240149907A (ko) 2024-10-15
US20230259758A1 (en) 2023-08-17
JP2025505291A (ja) 2025-02-21

Similar Documents

Publication Publication Date Title
TWI857493B (zh) 用於神經網路運算之電腦實施方法，系統以及非暫時性電腦可讀媒體
JP7752199B2 (ja) 階層的重み疎畳み込み処理のための方法とシステム
Liang et al. Evaluating fast algorithms for convolutional neural networks on FPGAs
CN108765247B (zh) 图像处理方法、装置、存储介质及设备
CN114026569B (zh) 使用脉动阵列的扩张卷积
Lu et al. SpWA: An efficient sparse winograd convolutional neural networks accelerator on FPGAs
CN111831254B (zh) 图像处理加速方法、图像处理模型存储方法及对应装置
CN117273101B (zh) 用于均衡权重稀疏卷积处理的方法及系统
KR102316670B1 (ko) 연산 가속기
US9886377B2 (en) Pipelined convolutional operations for processing clusters
KR20220129107A (ko) 행렬 곱셈기
JP2021509747A (ja) ハードウェアベースのプーリングのシステムおよび方法
CN106846235B (zh) 一种利用NVIDIA Kepler GPU汇编指令加速的卷积优化方法及系统
US20240273163A1 (en) Accelerator for sparse matrix multiplication in neural networks
Zlateski et al. ZNNi: maximizing the inference throughput of 3D convolutional networks on CPUs and GPUs
KR102372869B1 (ko) 인공 신경망을 위한 행렬 연산기 및 행렬 연산 방법
CN119166287A (zh) 计算任务优化方法、装置、设备、介质和程序产品
Song et al. Design and implementation of convolutional neural networks accelerator based on multidie
CN120226017A (zh) 具有卷积计算单元的向量运算加速
KR20230110355A (ko) 계층별 분석을 통한 신경망 프루닝 방법 및 시스템
CN119416850B (zh) 一种适配硬件张量指令及内存的神经网络推理优化方法
US20240126617A1 (en) Deep fusion of kernel execution
CN114692841B (zh) 数据处理装置、数据处理方法及相关产品
CN114764608B (zh) 执行神经网络模型的数据处理装置、方法及相关产品
CN119961559A (zh) 矩阵乘性能优化方法、装置、电子设备和存储介质