Implementation method and equipment of training model and storage medium

A technology for training models and implementation methods, applied in the field of communication, can solve problems such as numerous versions of training tools, achieve the effects of saving compilation time, facilitating development and maintenance, and improving version development efficiency

Inactive Publication Date: 2019-06-04
ZTE CORP
View PDF1 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of this, the purpose of the embodiments of the present invention is to provide a training model implementation method, device and computer-readable storage medium to solve the problem that there are many versions of training tools, resulting in running all training models on the same platform during the product integration and joint debugging phase. Problems in production environment

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Implementation method and equipment of training model and storage medium
  • Implementation method and equipment of training model and storage medium
  • Implementation method and equipment of training model and storage medium

Examples

Experimental program
Comparison scheme
Effect test

no. 1 example

[0023] In order to better understand this embodiment, the following describes the compilation process of adding a new layer to an existing training tool:

[0024] Such as figure 1 Shown is a schematic diagram of the internal framework of the existing training tool. The training tool includes a core layer, the underlying dependency function library, and several new added layers. The core layer includes a convolutional layer, a pooling layer, and a fully connected layer.

[0025] Taking caffe as an example, the caffe version compilation process is as follows figure 2 shown. Caffe uses the protoc tool to generate code caffe.pb.cc and caffe.pb.h before compiling the configuration items of all layers. The code of each layer of caffe includes the configuration item code with include, and then compiles and generates libcaffe.so. Compile caffe.cpp and link libcaffe.so to generate the executable file caffe.

[0026] Please refer to image 3 As shown, if you want to add a layer (BN...

no. 2 example

[0081] Such as Figure 9 As shown, the second embodiment of the present invention provides a device for implementing a training model, which includes: a memory 21, a processor 22, and a training model stored in the memory 21 and operable on the processor 22 The implementation program of the training model, when the implementation program of the training model is executed by the processor 22, is used to implement the steps of the implementation method of the training model described below:

[0082] The training program loads the dynamic library generated by the business algorithm logic layer;

[0083] During forward propagation and backpropagation calculation, the forward propagation program and the backpropagation program of the dynamic library are called through the forward propagation interface and the backpropagation interface for training to obtain the training model.

[0084] When the implementation program of the training model is executed by the processor 22, ...

no. 3 example

[0102] The third embodiment of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a training model implementation program, and the training model implementation program is used to implement the first embodiment when executed by a processor The steps of the implementation method of the training model.

[0103] It should be noted that the computer-readable storage medium in this embodiment belongs to the same idea as the method in the first embodiment, and its specific implementation process is detailed in the method embodiment, and the technical features in the method embodiment are all described in this embodiment. Correspondingly applicable, no more details here.

[0104] In the computer-readable storage medium of the embodiment of the present invention, the training program and the logic layer of the business algorithm are decoupled, and the training program and the logic layer of the business algorithm can ru...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The embodiment of the invention discloses a training model implementation method and device and a computer readable storage medium. The method comprises the steps that a training program loads a dynamic library generated by a service algorithm logic layer; And during forward propagation and backward propagation calculation, a forward propagation program and a backward propagation program of the dynamic library is called for training through a forward propagation interface and a backward propagation interface to obtain a training model. The training program and the service algorithm logic layerare decoupled, the training program and the service algorithm logic layer can run independently, the training program can load the service algorithm logic layer to run, and respective development andmaintenance are facilitated; Compilation and operation only involve service algorithm logic layer codes, the compilation time is saved, and the version development efficiency is improved.

Description

technical field [0001] The present invention relates to the field of communication technologies, and in particular to a training model implementation method, device and computer-readable storage medium. Background technique [0002] In the process of teamwork, due to the many versions of training tools, different R&D engineers may obtain different versions of training tools, and different training tools depend on different software and hardware environments. The software environment and hardware environment of the environment are different. Once all training models are run in the same production environment during the product integration and joint debugging phase, various unexpected exceptions will appear. Contents of the invention [0003] In view of this, the purpose of the embodiments of the present invention is to provide a training model implementation method, device and computer-readable storage medium to solve the problem that there are many versions of training too...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F9/445G06F9/448
CPCG06F9/445G06F9/448
Inventor 李忠良屠要峰陈小强郭斌黄震江高洪
Owner ZTE CORP
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap