Unlock instant, AI-driven research and patent intelligence for your innovation.

AI accelerator card simulation test system based on ResNet50 network and working method thereof

A technology of simulation testing and working methods, applied in biological neural network models, detection of faulty computer hardware, functional testing, etc., can solve problems such as low design efficiency of AI accelerator cards

Active Publication Date: 2020-06-09
SUZHOU LANGCHAO INTELLIGENT TECH CO LTD
View PDF4 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The application provides a ResNet50 network-based AI accelerator card simulation test system and its working method to solve the problem of low design efficiency of AI accelerator cards in the prior art

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
  • AI accelerator card simulation test system based on ResNet50 network and working method thereof
  • AI accelerator card simulation test system based on ResNet50 network and working method thereof
  • AI accelerator card simulation test system based on ResNet50 network and working method thereof

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0084] see figure 1 , figure 1 It is a schematic structural diagram of an AI accelerator card simulation test system based on the ResNet50 network provided by the embodiment of the present application.

[0085] Depend on figure 1 It can be seen that the AI ​​accelerator card simulation test system based on the ResNet50 network in this embodiment mainly includes: a convolution module, an activation module, a pooling module, a residual module, a fully connected module, a softmax module and a quantization module.

[0086] This embodiment uses C++ programming to implement basic modules such as convolution module, activation module, pooling module, residual module, fully connected module, softmax module and quantization module in the network respectively, and each module is a C++ class implementation. Classify and divide the ResNet50 network, which is divided into res1, res2, res3, res4, res5, full connection and softmax, and use the basic modules to form branch1 and branch2 bran...

Embodiment 2

[0111] exist Figure 1-Figure 4 On the basis of the illustrated embodiment see Figure 5 , Figure 5 It is a schematic flowchart of the working method of a ResNet50 network-based AI accelerator card simulation test system provided by the embodiment of the present application.

[0112] Depend on Figure 5 As can be seen, the working method in the present embodiment mainly includes the following processes:

[0113] Build the basic modules, including: convolution module, activation module, pooling module, residual module, full connection module, softmax module and quantization module.

[0114] S1: Use the basic module to construct branch1 branch and branch2 branch.

[0115] S2: Use branch1 branch and branch2 branch to form res.

[0116] S3: Use res to build a ResNet50 network.

[0117] S4: Use the convolution module to perform convolution calculation.

[0118] Specifically, step S4 also includes the following process:

[0119] S40: Acquire quantized FeatureMap data, weigh...

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 an AI accelerator card simulation test system based on a ResNet50 network and a working method thereof. The simulation test system mainly comprises a convolution module, an activation module, a pooling module, a residual error module, a full connection module, a softmax module and a quantification module. The working method of the simulation test system mainly comprises thesteps of: building a basic module, building a barch1 branch and a barch2 branch through the basic module, and forming res through the barch1 branch and the barch2 branch; constructing a ResNet50 network by utilizing res; performing convolution calculation by using a convolution module; quantizing the float32 data subjected to convolution calculation by adopting a quantization module to obtain quantized data; and according to the quantized data, carrying out simulation test on the AI accelerator card in the ResNet50 network. According to the method and the device, the calculation amount can beeffectively reduced, and the test time is saved, so that the system design problem of the AI accelerator card and the quantized algorithm precision are quickly verified, and the software debugging and hardware development speeds are increased.

Description

technical field [0001] The application relates to the technical field of artificial intelligence, in particular to a ResNet50 network-based AI accelerator card emulation test system and its working method. Background technique [0002] With the widespread application of artificial intelligence in various fields, people have higher and higher requirements for the calculation speed and accuracy of artificial intelligence. In order to meet the people's demand for computing accuracy and speed, major hardware manufacturers specialize in AI accelerator cards. For the accelerator card that has been designed, it is necessary to verify whether its performance is qualified before making the board. Therefore, how to verify the AI ​​accelerator card so as to improve the success rate of board production is an important technical issue. [0003] The current method of verifying the AI ​​accelerator card is mainly to directly verify on the board, through the test of the board to confirm w...

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
IPC IPC(8): G06F11/26G06N3/04G06N3/10
CPCG06F11/26G06N3/10G06N3/045
Inventor 曹其春赵雅倩董刚梁玲燕尹文枫
Owner SUZHOU LANGCHAO INTELLIGENT TECH CO LTD
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com