公开(公告)号：US20220138528A1

公开(公告)日：2022-05-05

申请号：US17526755

申请日：2021-11-15

Applicant: BEIJING BAIDU NETCOM SCIENCE TECHNOLOGY CO., LTD.

Inventor： Chao TIAN ,  Lei JAI ,  Junhui WEN ,  Qiang LI

IPC: G06N3/04 ,  G06F9/50

Abstract: A data processing method for a neural network accelerator, an electronic device and a storage medium are provided. The technical solution includes: obtaining data to be processed and an operation to be executed; obtaining a real-number full-connection operation corresponding to the operation to be executed; and performing the real-number full-connection operation on the data based on a real-number full-connection unit of the neural network accelerator to obtain a result of the operation to be executed for the data.

公开(公告)号：US20220292337A1

公开(公告)日：2022-09-15

申请号：US17832303

申请日：2022-06-03

Applicant: BEIJING BAIDU NETCOM SCIENCE TECHNOLOGY CO., LTD.

Inventor： Chao TIAN ,  Lei JIA ,  Xiaoping YAN ,  Junhui WEN ,  Guanglai DENG ,  Qiang LI

IPC: G06N3/063 ,  G06F9/30 ,  G06F9/50 ,  G06K9/62

Abstract: A neural network processing method, a neural network processing unit (NPU) and a processing device are provided. The method includes: obtaining by a quantizing unit in the NPU float type input data, quantizing the float type input data to obtain quantized input data, and providing the quantized input data to an operation unit; performing by the operation unit of the NPU a matrix-vector operation and/or a convolution operation to the quantized input data to obtain an operation result of the quantized input data; and performing by the quantizing unit inverse quantization to the operation result output by the operation unit to obtain an inverse quantization result.