公开(公告)号：US20210357735A1

公开(公告)日：2021-11-18

申请号：US17250890

申请日：2019-05-21

Applicant: Institute of Computing Technology, Chinese Academy of Sciences

Inventor： Xiaowei LI ,  Xin WEI ,  Hang LU

IPC: G06N3/063 ,  G06N3/04

Abstract: Disclosed embodiments relate to a split accumulator for a convolutional neural network accelerator, comprising: arranging original weights in a computation sequence and aligning by bit to obtain a weight matrix, removing slack bits in the weight matrix, allowing essential bits in each column of the weight matrix to fill the vacancies according to the computation sequence to obtain an intermediate matrix, removing null rows in the intermediate matrix, obtain a kneading matrix, wherein each row of the kneading matrix serves as a kneading weight; obtaining positional information of the activation corresponding to each bit of the kneading weight; divides the kneading weight by bit into multiple weight segments, processing summation of the weight segments and the corresponding activations according to the positional information, and sending a processing result to an adder tree to obtain an output feature map by means of executing shift-and-add on the processing result.

公开(公告)号：US20210350204A1

公开(公告)日：2021-11-11

申请号：US17250889

申请日：2019-05-21

Applicant: Institute of Computing Technology, Chinese Academy of Sciences

Inventor： Xiaowei LI ,  Xin WEI ,  Hang LU

IPC: G06N3/04

Abstract: Disclosed embodiments relate to a convolutional neural network accelerator, comprising: arranging original weights in a computation sequence and aligning by bit to obtain a weight matrix, removing slack bits in the weight matrix, allowing essential bits in each column of the weight matrix to fill the vacancies according to the computation sequence to obtain an intermediate matrix, removing null rows in the intermediate matrix, obtain a kneading matrix, wherein each row of the kneading matrix serves as a kneading weight; obtaining positional information of the activation corresponding to each bit of the kneading weight; divides the kneading weight by bit into multiple weight segments, processing summation of the weight segments and the corresponding activations according to the positional information, and sending a processing result to an adder tree to obtain an output feature map by means of executing shift-and-add on the processing result.

公开(公告)号：US20210350214A1

公开(公告)日：2021-11-11

申请号：US17250892

申请日：2019-05-21

Applicant: Institute of Computing Technology, Chinese Academy of Sciences

Inventor： Xiaowei LI ,  Xin WEI ,  Hang LU

IPC: G06N3/04 ,  G06F7/50 ,  G06F5/01 ,  H03M7/40

Abstract: Disclosed embodiments relate to a convolutional neural network computing method and system based on weight kneading, comprising: arranging original weights in a computation sequence and aligning by bit to obtain a weight matrix, removing slack bits in the weight matrix, allowing essential bits in each column of the weight matrix to fill the vacancies according to the computation sequence to obtain an intermediate matrix, removing null rows in the intermediate matrix, obtain a kneading matrix, wherein each row of the kneading matrix serves as a kneading weight; obtaining positional information of the activation corresponding to each bit of the kneading weight; divides the kneading weight by bit into multiple weight segments, processing summation of the weight segments and the corresponding activations according to the positional information, and sending a processing result to an adder tree to obtain an output feature map by means of executing shift-and-add on the processing result.