Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Deep learning model file conversion method and system, computer equipment and computer readable storage medium

A model file and deep learning technology, applied in the field of deep learning, can solve the problems of unrevealed model files and limitations of the application environment, and achieve the effect of ensuring smooth loading

Pending Publication Date: 2020-06-12
ONE CONNECT SMART TECH CO LTD SHENZHEN
View PDF1 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the Chinese patent application number CN201710669779 only discloses the parameters of converting the general deep learning model into a single NPU model, but does not disclose the conversion of the general deep learning model into model files under other deep learning frameworks, and the application environment is still limited

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
  • Deep learning model file conversion method and system, computer equipment and computer readable storage medium
  • Deep learning model file conversion method and system, computer equipment and computer readable storage medium
  • Deep learning model file conversion method and system, computer equipment and computer readable storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0048] see figure 1 As shown, the conversion method of the deep learning model file of the present invention comprises the following steps:

[0049] Step S10: Receive the model file to be converted obtained under the original deep learning framework.

[0050] In this embodiment, the model file to be converted has weight files and network structure codes, and the model file to be converted is a complete model file.

[0051] Step S20: According to the file type of the model file to be converted, select the corresponding readable frame to read the model file to be converted, so as to obtain the loading information of the model file to be converted.

[0052] In this embodiment, the loading information is the information loaded into the network structure code and weight file, and the loading information includes data types, computing units, calculation graphs, activators, and channel dimensions.

[0053] In this embodiment, the key code is intelligently located, and the original ...

Embodiment 2

[0110] see figure 2 As shown, the conversion method of the deep learning model file of the present invention comprises the following steps:

[0111] Step S10: Receive the model file to be converted obtained under the original deep learning framework.

[0112] In this embodiment, the model file to be converted includes a complete model file and an incomplete model file, the complete model file includes a weight file and network structure code, and the incomplete model file includes a weight file but not a network structure code.

[0113] Step S210: Identify the original deep learning framework corresponding to the model file to be converted.

[0114] Step S220: Determine whether the model file to be converted is a complete model file; if yes, execute step S230; if not, execute step S240;

[0115] Step S230: Loading processing to obtain loading information of the model file to be converted.

[0116] In this embodiment, the loading information of a complete model file (inclu...

Embodiment 3

[0123] see image 3 As shown, the conversion method of the deep learning model file of the present invention comprises the following steps:

[0124] Step S10: receiving the model file to be converted obtained under the original deep learning framework;

[0125] Step S210: Identify the original deep learning framework corresponding to the model file to be converted;

[0126] Step S2210: Determine whether the model file to be converted needs further integrity check by identifying the saving code of the model file to be converted; if yes, execute step S2220; if not, execute step S230;

[0127] Step S2220: Check whether the content of the model file to be converted has the characteristics of the network structure code; if so, execute step S230; if not, execute step S240;

[0128] Step S230: loading processing to obtain the loading information of the model file to be converted;

[0129] Step S240: Prompt for supplementation and wait for supplementation, then return to step S2210...

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 invention discloses a deep learning model file conversion method and system, computer equipment and a computer readable storage medium, and the method comprises the following steps: receiving a to-be-converted model file obtained under an original deep learning framework; according to the file type of the to-be-converted model file, selecting a corresponding readable framework to read the to-be-converted model file so as to obtain loading information of the to-be-converted model file; analyzing and / or translating the loaded information to obtain intermediate standard information; integrating into an intermediate specification model file according to the information of the intermediate specification; the intermediate specification model file is opened through the target deep learning framework, the storage engine of the target deep learning framework is started to store the obtained intermediate specification model file as the target model file, and the target model file can be applied to the target deep learning framework, so that the application environment of the target model file is expanded.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a conversion method, system, computer equipment and computer-readable storage medium of a deep learning model file that expands the application environment of the target model file. Background technique [0002] There are many commonly used deep learning frameworks, and different frameworks are often used in the process from algorithm development to algorithm deployment. Because the model files output by different frameworks are incompatible with each other, the first model file trained by the user in the first learning framework cannot be used in the second learning framework. For example, the user has trained an image recognition model with the Pytorch framework. But the production environment uses the TensorFlow framework for inference. The traditional method needs to rewrite the model with the TensorFlow framework, and then train the parameters, which will be a...

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): G06N20/00
CPCG06N20/00
Inventor 朱锦祥单以磊臧磊
Owner ONE CONNECT SMART TECH CO LTD SHENZHEN
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com