Font Size: a A A

Automatic Test Framework Technology For High-level Language Of AI Chip

Posted on:2021-04-22Degree:MasterType:Thesis
Country:ChinaCandidate:L F ZhengFull Text:PDF
GTID:2428330602977689Subject:Electronic and communication engineering
Abstract/Summary:
In this era of rapid development of technology,if a certain product wants to have a strong competitiveness in the market,it must have fast and high-quality delivery capabilities during the development process.Therefore,automated testing during the development process is pivotal If a lot of manual testing is used in the development process,the test task may not be completed in time,increasing the risk that the project cannot be delivered on time,so the automatic test framework technology is used,which is a continuation of manual testing and can reduce manual testing.The repetitive work greatly improves the test efficiency.The automatic test framework in this subject is divided into two parts,one is the automatic test framework for Bangc's built-in function test set,and the other is the automatic test framework for deep learning operators,which is for a company's platform chip MLU(Machine Learning Unit)compiler,MLU is an AI chip for deep learning.The main work of the thesis includes:(1)Manually construct a test set for the built-in function of BangC to ensure the correctness of the implementation of the built-in function,and construct a description file of the test set.Each line in the file contains the name of the built-in function and the parameters and parameter sizes required by the built-in function.Then use python language to design automated scripts to realize the function call form of test set use cases,and finally use python language and shell scripting language to implement the automatic test framework of the test set to complete the testing and verification of the framework.(2)Extract several commonly used deep learning operators,complete the basic implementation of the MLU terminal operator,vectorize the operator according to the built-in function of BangC,design the description file of the operator,and implement the automatic test framework for the operator,Complete the testing and verification of the framework.This thesis designs an automatic test framework based on the BangC language of the AI chip,which implements automated testing of all built-in function test sets and operator test sets,reducing manual repetitive work,and can quickly check the correctness of one or more built-in functions,And comparison of the results of the operators,so that you can have more time to design test cases,optimize operators and improve the performance of the test framework.
Keywords/Search Tags:MLU, automatic test framework, vectorization, python, BangC
Related items