Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
DEEP LEARNING MODEL DISTRIBUTED OPERATION METHOD AND APPARATUS
Document Type and Number:
WIPO Patent Application WO/2021/098269
Kind Code:
A1
Abstract:
A deep learning model distributed operation method and apparatus. The method comprises: registering a virtual processor in a device management list; registering and compiling an operator supported by the virtual processor; detecting hardware resources associated with the virtual processor, and determining respective allocation proportions of the hardware resources according to the computing power of the associated hardware resources; configuring a deep learning model on the basis of the operator supported by the virtual processor, and specifying the virtual processor for an operator used in the deep learning model; and the virtual processor allocating, according to the allocation proportions, input data of a corresponding operator to the hardware resources associated with the virtual processor so as to perform operation, and merging operation results of the hardware resources into an output of the corresponding operator. The concept of a virtual processor is introduced, the virtual processor is specified as a computing device for a corresponding operator, operation is allocated to different hardware devices to achieve parallel execution, such that the heterogeneous acceleration of deep learning model operation is achieved.

Inventors:
ZHAO QIANQIAN (CN)
TONG PEILIN (CN)
ZHAO HONGBO (CN)
Application Number:
PCT/CN2020/104006
Publication Date:
May 27, 2021
Filing Date:
July 24, 2020
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
INSPUR SUZHOU INTELLIGENT TECHNOLOGY CO LTD (CN)
International Classes:
G06N20/00
Foreign References:
CN110866610A2020-03-06
CN108369537A2018-08-03
CN108805798A2018-11-13
CN110121747A2019-08-13
US20180302291A12018-10-18
Attorney, Agent or Firm:
UNITALEN ATTORNEYS AT LAW (CN)
Download PDF: