CN109634843A - A kind of distributed automatization method for testing software and platform towards AI chip platform - Google Patents
A kind of distributed automatization method for testing software and platform towards AI chip platform Download PDFInfo
- Publication number
- CN109634843A CN109634843A CN201811285022.1A CN201811285022A CN109634843A CN 109634843 A CN109634843 A CN 109634843A CN 201811285022 A CN201811285022 A CN 201811285022A CN 109634843 A CN109634843 A CN 109634843A
- Authority
- CN
- China
- Prior art keywords
- client
- test
- task
- server end
- chip
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3692—Test management for test results analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3664—Environments for testing or debugging software
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3688—Test management for test execution, e.g. scheduling of test suites
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/546—Message passing systems or structures, e.g. queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2209/00—Indexing scheme relating to G06F9/00
- G06F2209/54—Indexing scheme relating to G06F9/54
- G06F2209/548—Queue
Abstract
The invention discloses a kind of distributed automatization method for testing software and platform towards AI chip platform.The platform can go to run by the different AI chip platform client computer that is distributed to that server automates AI software test task, it is after the completion of test that test result is unified to server admin, then it issues environmental renewal task automatically according to demand and initiates second of identical test assignment then to X86 client, the task carries out pure software operation, and the unified secondary test result of collection management in X86 client.After test job twice, automatically test result twice can be compared and analyzed, and export analysis result；Also, the change of front and back off-line model twice is traced, the node or input for positioning off-line model lead to mistake, and helper applications engineer and Hardware Engineer check problem.
Description
Technical field
The invention belongs to computer software technical fields；It is related to once carrying out AI core in multiple stage computers or development board
The scene of piece platform software test proposes one kind towards AI chip platform distributed automatization method for testing software and platform.
Background technique
Software test (or being software detection) is for identifying the correctness of software, integrality, safety and quality
Process.Software test is the audit or comparison procedure between a kind of reality output and anticipated output, in order to defined
Under the conditions of program is operated, to find program error, measure software quality, and design requirement progress whether is able to satisfy to it
The process of assessment.Software test is generated along with the generation of software, is responsible for completion by software developer oneself in early days, now
There is special tester to complete.
The increasingly diversification for calculating acceleration function chip and support algorithm with having AI, causes the function of AI software more next
It is more complicated, corresponding test job also become increasingly complex with it is heavy.In actual test job, often have for a certain
Software carries out the demand of repeatedly different tests, such as the test of the application scenarios such as ResNet SSD target detection, needs
It will be for various parameters and image input test AI chip platform to, to the correctness of deduction and performance, being related to before neural network
Disparate modules have a dozens of, each module has dozens of even a test points up to a hundred, the test case meeting finally referred to again
Have several hundred a or thousands of.
The method of usual AI software test mainstream is manually or operation simple script is tested, and ties to test
Fruit is recorded.Usually after devising test case and passing through evaluation, as tester according to described in test case
Regulation executes test step by step, obtains the actual result of AI chip platform, and compared with CPU expected result.Manual test
Advantage be people can the result to test case reasonably handled and coped with, especially when test result is uncertain
When, but disadvantage is also apparent from --- and under-efficiency is relatively low, and insufficient test likely results in the operation mistake of AI software
Accidentally, for example the consequences such as tesla's traffic accident, the present invention are the concept that AI chip platform just introduces distributed automatization test, are used
Large-scale machines carry out AI software automated testing.
Software automated testing is will to be converted into a kind of process that machine executes taking human as the behavioral test of driving.Software is certainly
Dynamicization testing research field is concentrated mainly on automation (such as unit of automatic management and the dynamic test of software testing flow
Test, functional test and aspect of performance).In the two fields compared with manual test, the advantage of test automation is obvious
's.Testing efficiency can be improved in automatic test first, and tester is made to focus more on the foundation of new test module and open
Hair, to improve test coverage；Secondly, automatic test is more convenient for testing the digital management of assets, so that test assets
The available multiplexing in entire software test lifecycle, this feature are particularly relevant in functional test and regression test；
In addition, testing process automatic management can make the test activity development of mechanism more proceduring.It in actual operation, is test
The correctness of algorithm and AI chip platform often has the scene tested and compared twice, for example realizes one based on CPU nerve
Network algorithm needs to be tested on the machine for having AI chip platform again to examine the validity of AI chip platform,
Then data, the performance of front and back twice are compared.Traditional automation software testing does not support this scene.
Summary of the invention
For the technical problems in the prior art, it is distributed that the purpose of the present invention is to provide a kind of AI chip platforms
Automated software testing method and platform.The platform can be distributed to not by server by what AI software test task automated
With AI chip platform client computer go to run, after the completion of test that test result is unified to server admin, then root
It issues environmental renewal task automatically according to demand and initiates second of identical test assignment then to X86 client, which exists
Pure software operation, and the unified secondary test result of collection management are carried out in X86 client.When twice, test job terminates
Afterwards, automatically test result twice can be compared and analyzed, and exports analysis result.Also, to front and back offline mould twice
The change of type is traced, and the node or input for positioning off-line model lead to mistake, helper applications engineer and hardware engineering
Teacher checks problem.
It is of the invention towards AI chip platform distributed automatization software test platform, its technical solution is as follows.
The technical program mainly includes disappearing of communicating between server end, x86 client, AI chip client and three
Cease the technological development of this three parts of queue.Pass through a server end Master and multiple attached client Slave computers
To construct the automatic test platform jointly.Server end Master controls the whole flow process of software test, according to current visitor
The current test assignment executive condition at family end distributes corresponding test assignment for client, and receives the survey of client return
Test result, and result is analyzed；Client tests the test assignment received according to the instruction of server, and will
Test result feeds back to server；Between server and client, communicated by message queue.
1) as shown in Figure 1, server end mainly include system initialization module, task sending module, results acquisition module,
Comparative result analysis module.Initialization module mainly completes the load of software test task list, and (test assignment is saved by Master
Point initialization and unified management), (information includes the IP address of client Slave machine, X86/ for the load of client configuration file
AI chip client, CPU model, and it is tied to the test assignment list of the machine), and heartbeat request is initiated to client,
Ensure that each client is in available state；Task distribution module is responsible for searching the test assignment not being performed and idle visitor
Family end, and task is distributed for idle client；Results acquisition module receives the test result of client feedback, and by the result
Storage is in the database；Comparative result analysis module reads the result that front and back is tested twice from database and compares, and analyzes
And export the test result item for having gap.Difference analysis, retrospect finally are carried out to the data of derived binary system off-line model
The change of front and back off-line model twice, including calculating the change of the model datas such as figure changes, weight, arrangement variation of memory etc.,
The node or input for positioning off-line model lead to mistake, and helper applications engineer and Hardware Engineer check problem.
2) as shown in Fig. 2, client mainly includes context initialization, more new environment and the execution big portion of neural computing three
Point.Context initialization mainly to the work such as context initialization and deployment, carries out the preparation of some general detection work.For
X86 client then initializes the component of neural computing, then disposes its environment for AI chip client.More new environment is main
It is installing and deploying for some specific test platforms to be done for specific test assignment, including off-line model updates deployment etc..
Executing neural computing is mainly the task Distribution List according to server end, is run to specific test assignment, is run
After, it will test the information such as result and be passed into message queue, device end to be serviced obtains and analyzes it.
3) message queue carries out the transmitting of message for both sides mainly as the bridge of server end and client.Server
Test assignment can be submitted to message queue by end, and each client is registered on message queue, then received message queue
In issue the task of oneself, and execute.After client executing complete task, also test result can be sent to message queue, etc.
Server is waited to go to obtain.
Specific step is as follows:
1) server-side system initializes: loading use-case test table first, client allocation list, automatically generates extensive survey
Data are tried, and compiling off-line model is generated using AI chip tools chain with information above.The off-line model is AI chip client
It uses, and X86 client then be used directly the test data of network storage.After the completion of above, heartbeat is issued to each client
Information, checks whether each client is in normal condition.
2) after client receives heartbeat message, the basic environment of itself can be checked, to clothes if all going well
Reply ready response in business end.
3) server end can update client state table according to the response of X86 and AI chip client, then to each
Client sends the order of initialization, allows its preparation.
4) for client according to the initialization command of server end, command parameter is a json file, including off-line model
With test data store path, Hardware match library version and environmental variance etc., client is updated according to the statement of json file
Locally execute environment, and feedback server end Ready state.
5) server distributed tasks: according to the client Slave metadata of maintenance (including client hardware type, Ready
State, Process status information) and test assignment metadata (including the state that each task is in, distributed, be completed,
It is not carried out), being not carried out for task and idle client are therefrom chosen, task is sent to idle client executing.And with one
Wheel of fixing time interrogates client state, safeguards Slave metadata.
6) client obtains task: monitoring information queue is jumped when getting the task dispatch messages of server end
7)。
7) client parse task: parsing task, with obtain task number, test list, Client_id, off-line model or
The information such as person's test data store path, and check that store path whether there is.
8) client executing is tested: executing specified test assignment, and feedback server end Process status information.
9) after obtaining test result, current test client feedback test result: is sent to server by message queue
The information such as task number, Client_id, state and the final result of task.
10) server results acquire: receiving the test result from client feedback and the result is stored in database
In.
11) it CPU task execution: in order to judge whether AI chip platform correctly executes off-line model, needs to construct same net
The CPU task of network and model data.Update the configuration of each FTP client FTP according to demand, at the same according to 5) to 10) the step of send out
Play the processing of same cpu test task, this subtask issues the X86 client in cluster, execute same test data and
Calculated result and performance information are saved into JSON file, return to server by off-line model.
12) server results comparative analysis: test result twice is subjected to automation comparison, and is exported automatically variant
Part.
13) server error checking: server is obtained in relatively cpu test with the test of AI chip platform as a result, confirmation
As a result after mistake, the modification of off-line model is analyzed, it is offline to this test to test correct off-line model to the last time
Model is analyzed.Specifically, difference analysis is carried out to the data of the two export off-line model, retrospect front and back is offline twice
The change of model, including calculating the change of the model datas such as figure change, weight, arrangement variation of memory etc., auxiliary positioning is offline
The node of model or input lead to mistake, determine modification and the erroneous association of off-line model.Helper applications engineer and hardware
Engineer checks problem.
Compared with prior art, the positive effect of the present invention are as follows:
(1) traditional automatic test software is based on single cpu mode more, and the present invention is based on distributed modes, uses
Master-Slave framework, it is more flexible and efficient；
(2) traditional AI automatic test software is without providing test twice and carrying out to test result twice automatic
Change the function of analysis, the present invention is directed to AI chip platform, provides contrast test and automatically carries out to test result twice
The function of analysis carries out analysis solution to corresponding problem convenient for developer.
(3) error result of AI chip platform may be caused by hardware or software, and off-line model node and side are divided in addition
Analysis is complicated, and defect repair is difficult to carry out.The present invention tracks the version of off-line model, to calculating in this off-line model
The reason of variation of arrangement etc. of the model datas such as figure, weight, memory is analyzed and exported, lead to error result, helps soft
Part engineer and Hardware Engineer efficiently check problem.
Detailed description of the invention
Fig. 1 is the distributed automatization software test platform server end architecture diagram towards AI chip platform；
Fig. 2 is the distributed automatization software test platform client framework figure towards AI chip platform.
Specific embodiment
Below by example, the present invention is further illustrated, the range of but do not limit the invention in any way.
Set following test scene: server one, server；Client 8, respectively AI_client1, AI_
client2、AI_client3、AI_client4、x86_client1、x86_client2、x86_client3、x86_
client4；Use-case test table has 5 detections, is case1, case2, case3, case4, case5,5 test items it is specific
Content such as table 1:
Table 1 is the description of 5 case
| Case | Description |
| Case1 | ResNet is tested based on TensorFlow |
| Case2 | SSD is tested based on TensorFlow |
| Case3 | VGG is tested based on TensorFlow |
| Case4 | AlexNet is tested based on TensorFlow |
| Case5 | Conv is tested based on TensorFlow |
Implementation steps are as follows:
1) server-side system initializes: the configuration files such as load test table case1~case5, client allocation list, complete
The initialization (including generating test data and off-line model etc.) of pairs of server end, then to customer end A I_client1~
AI_client4, x86_client1~x86_client4 send the order of initialization, and containing script in order can allow it to download
Corresponding dependent software package TensorFlow is simultaneously installed；
2) client environment is disposed: after receiving the context initialization order from server end, operation order, and installation
The relevant environment of TensorFlow platform.
3) after client completes initialization, server end ready for sending signal can the ready feedback of client: be given.
4) server task is distributed: according to the feedback of client, discovery currently has 8 free time client, and 5 tests are appointed
Business；Server end can carry out layout to client and test assignment, and case1 is then distributed to AI_client1 and x86_
Client1, case2 is distributed to AI_client2 and x86_client2, case3 is distributed to AI_client3 and x86_
Client3, case4 is distributed to AI_client4 and x86_client4, and 8 clients are all in busy condition at this time, to mesh
Mark network is tested, and does not have suitable client that can execute test assignment case5, then case5 is continued waiting for.
5) customer end A I_client1~AI_client4, x86_client1~x86_client4 obtain task: monitoring
6) message queue is jumped when getting the task dispatch messages of server end.
6) client parses task: parsing task, to obtain task number, the test information such as list and Client_id, such as
AI_client1 and x86_client1 will receive the task of case1, i.e., tests ResNet.
7) client executing is tested: in the test assignment that corresponding client executing is specified, such as AI_client1 and x86_
Client1, which can start to execute, tests ResNet based on TensorFlow, and wherein AI_client1 loads off-line model
It is tested, and x86_client1 parsing network and test data are tested.
8) client feedback test result: AI_client1~AI_client4, x86_client1~x86_client4
After obtaining test result, by message queue to server send the task number of current test assignment, Client_id, state and
The information such as final result.Assuming that herein, AI_client3 and x86_client3 are initially completed test feedback, then server is simultaneously
It jumps 4), test assignment case5 is distributed in AI_client3 and x86_client3 and executes test.
9) server results acquire: receiving the test result from client feedback and the result is stored in database
In.
10) server results comparative analysis:, can be in the database by the AI of all case after the completion of all test assignments
The test result twice of chip platform and x86 platform reads out analysis, it is assumed that analyze herein result or performance it is wrong or
The detection that person degenerates is case2, then will finally export an analysis as a result, the front and back of case2 as the result is shown test
As a result wrong.If saved the neural network off-line model of the secondary inspection without mistake, for subsequent contrast point
Analysis.
11) server error checking: according to the analysis of the front and back 10) obtained test assignment twice as a result, first with mind
Through directed graph cutting method, the last time is checked and is correctly carried out with internet off-line model and this error off-line model
Comparative analysis, obtain off-line model used in error result fall into a trap the model datas such as nomogram, weight, memory arrangement etc. variation feelings
Condition, auxiliary program person carry out wrong investigation.
The above embodiments are merely illustrative of the technical solutions of the present invention rather than is limited, the ordinary skill of this field
Personnel can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the spirit and scope of the present invention, this
The protection scope of invention should be subject to described in claims.
Claims (10)
1. a kind of distributed automatization method for testing software towards AI chip platform, step include:
1) server end load use-case test table, client allocation list and generate test data, and with information above compile from
Line model；Wherein client includes x86 client, AI chip client；
2) client locally executes environment according to the update of the initialization command of server end；The initialization command includes offline mould
Type and test data store path, Hardware match library version and environmental variance；
3) being not carried out for task is sent to idle AI chip client executing by server end；The parsing of AI chip client receives
Task, obtain task number, test list, client number Client_id, off-line model information or test data store road
Diameter；Then specified test assignment is executed and by test result feedback server end；Received server-side comes from AI chip client
It holds the test result of feedback and stores it in lane database；
4) the cpu test task being not carried out is sent to idle x86 client executing by server end；The parsing of x86 client is received
Arriving for task obtains task number, test list, client number Client_id, off-line model information or test data storage road
Diameter；Then specified test assignment is executed and by test result feedback server end；Received server-side is anti-from x86 client
The test result of feedback simultaneously stores it in lane database；
5) test result twice is carried out automation comparison and exports discrepant part by server end.
2. the method as described in claim 1, which is characterized in that server end exports off-line model from test result twice
Data simultaneously carry out difference analysis, obtain the change information of off-line model, change including calculating figure, weight changes, the row of memory
Cloth changes.
3. method according to claim 2, which is characterized in that the server end is by each test result and preset survey
Test result compares, and checks whether off-line model is correct；Then neural network digraph cutting method is utilized, it is correct to checking
Off-line model and the off-line model of error analyzed, obtain the change information of off-line model.
4. the method as described in claim 1, which is characterized in that the server end issues heartbeat message to each client,
Check whether each client is in normal condition；After client receives heartbeat message, the basic environment of itself can be examined
It looks into, replys ready response to server-side if all going well；Then server end can be updated according to the response of client
Then client state table sends the order of initialization to each client.
5. the method as described in claim 1, which is characterized in that the initialization command is a json file.
6. the method as described in claim 1, which is characterized in that the metadata of the server end maintenance client and test are appointed
The metadata of business；The metadata of the client includes client hardware type, Ready state, Process status information, institute
The metadata for stating test assignment includes state that each task is in.
7. a kind of distributed automatization software test platform towards AI chip platform, which is characterized in that including server end and
Client, the client include x86 client, AI chip client；Wherein,
The server end, load use-case test table, client allocation list and generation test data, and compile off-line model；
Being not carried out for task is sent to idle AI chip client executing, the cpu test task being not carried out is sent to the free time
X86 client executing；And automation comparison is carried out according to the test result received and exports discrepant part；
The AI chip client obtains task number, test list, client number Client_ for parsing receiving for task
Id, off-line model information or test data store path；Then specified test assignment is executed and by test result back services
Device end；Test result of the received server-side from AI chip client feedback simultaneously stores it in lane database；
The x86 client, for parsing receiving for task, obtain task number, test list, client number Client_id,
Off-line model information or test data store path；Then specified test assignment is executed and by test result feedback server
End；Test result of the received server-side from x86 client feedback simultaneously stores it in lane database.
8. platform as claimed in claim 7, which is characterized in that the information in the client allocation list includes the IP of client
Address, X86/AI chip client, CPU model and the test assignment list for being tied to the client.
9. platform as claimed in claim 7, which is characterized in that the server end and the x86 client, AI chip client
It is communicated between end by message queue.
10. platform as claimed in claim 9, which is characterized in that test assignment is submitted to message queue by the server end,
Each client is registered on message queue, oneself task and execution are then issued in received message queue；Work as client
After having executed task, test result is sent to message queue, waiting for service device goes to obtain.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811285022.1A CN109634843B (en) | 2018-10-31 | 2018-10-31 | Distributed automatic software testing method and platform for AI chip platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811285022.1A CN109634843B (en) | 2018-10-31 | 2018-10-31 | Distributed automatic software testing method and platform for AI chip platform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109634843A true CN109634843A (en) | 2019-04-16 |
| CN109634843B CN109634843B (en) | 2021-09-21 |
Family
ID=66066994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811285022.1A Active CN109634843B (en) | 2018-10-31 | 2018-10-31 | Distributed automatic software testing method and platform for AI chip platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109634843B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110070176A (en) * | 2019-04-18 | 2019-07-30 | 北京中科寒武纪科技有限公司 | The processing method of off-line model, the processing unit of off-line model and Related product |
| CN110910945A (en) * | 2019-11-19 | 2020-03-24 | 深圳忆联信息系统有限公司 | Method and device for testing robustness of SSD (solid State disk) signal, computer equipment and storage medium |
| CN111158967A (en) * | 2019-12-31 | 2020-05-15 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
| CN111506508A (en) * | 2020-04-17 | 2020-08-07 | 北京百度网讯科技有限公司 | Edge calculation test method, device, equipment and readable storage medium |
| CN111880433A (en) * | 2020-07-02 | 2020-11-03 | 上海机电工程研究所 | System and method for automatically realizing remote heterogeneous semi-physical simulation test task |
| CN111949329A (en) * | 2020-08-07 | 2020-11-17 | 苏州浪潮智能科技有限公司 | AI chip task processing method and device based on x86 architecture |
| CN112527676A (en) * | 2020-12-23 | 2021-03-19 | 中移（杭州）信息技术有限公司 | Model automation test method, device and storage medium |
| CN113779913A (en) * | 2021-11-12 | 2021-12-10 | 浙江大学 | Verification platform structure and test method for AI multi-chip system |
| WO2023036280A1 (en) * | 2021-09-10 | 2023-03-16 | 华为技术有限公司 | Model test method and apparatus |
| WO2023050885A1 (en) * | 2021-09-28 | 2023-04-06 | 华为技术有限公司 | Application performance testing method, method and apparatus for establishing performance testing model |
| CN117407299A (en) * | 2023-10-18 | 2024-01-16 | 北京大学 | Model test method and system |
| US11983535B2 (en) | 2019-03-22 | 2024-05-14 | Cambricon Technologies Corporation Limited | Artificial intelligence computing device and related product |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030131088A1 (en) * | 2002-01-10 | 2003-07-10 | Ibm Corporation | Method and system for automatic selection of a test system in a network environment |
| CN101615159A (en) * | 2009-07-31 | 2009-12-30 | 中兴通讯股份有限公司 | Off-line test system and local data management method thereof and corresponding device thereof |
| CN105786667A (en) * | 2016-02-29 | 2016-07-20 | 惠州Tcl移动通信有限公司 | Distributed automated testing method and system |
| CN106970880A (en) * | 2017-04-28 | 2017-07-21 | 中国科学院软件研究所 | A kind of distributed automatization method for testing software and system |
-
2018
- 2018-10-31 CN CN201811285022.1A patent/CN109634843B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030131088A1 (en) * | 2002-01-10 | 2003-07-10 | Ibm Corporation | Method and system for automatic selection of a test system in a network environment |
| CN101615159A (en) * | 2009-07-31 | 2009-12-30 | 中兴通讯股份有限公司 | Off-line test system and local data management method thereof and corresponding device thereof |
| CN105786667A (en) * | 2016-02-29 | 2016-07-20 | 惠州Tcl移动通信有限公司 | Distributed automated testing method and system |
| CN106970880A (en) * | 2017-04-28 | 2017-07-21 | 中国科学院软件研究所 | A kind of distributed automatization method for testing software and system |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11983535B2 (en) | 2019-03-22 | 2024-05-14 | Cambricon Technologies Corporation Limited | Artificial intelligence computing device and related product |
| CN110070176A (en) * | 2019-04-18 | 2019-07-30 | 北京中科寒武纪科技有限公司 | The processing method of off-line model, the processing unit of off-line model and Related product |
| CN110910945B (en) * | 2019-11-19 | 2021-08-17 | 深圳忆联信息系统有限公司 | Method and device for testing robustness of SSD (solid State disk) signal, computer equipment and storage medium |
| CN110910945A (en) * | 2019-11-19 | 2020-03-24 | 深圳忆联信息系统有限公司 | Method and device for testing robustness of SSD (solid State disk) signal, computer equipment and storage medium |
| CN111158967A (en) * | 2019-12-31 | 2020-05-15 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
| US11714128B2 (en) | 2019-12-31 | 2023-08-01 | Kunlunxin Technology (Beijing) Company Limited | Method and apparatus for testing artificial intelligence chip, device and storage medium |
| CN111158967B (en) * | 2019-12-31 | 2021-06-08 | 北京百度网讯科技有限公司 | Artificial intelligence chip testing method, device, equipment and storage medium |
| CN111506508A (en) * | 2020-04-17 | 2020-08-07 | 北京百度网讯科技有限公司 | Edge calculation test method, device, equipment and readable storage medium |
| CN111880433A (en) * | 2020-07-02 | 2020-11-03 | 上海机电工程研究所 | System and method for automatically realizing remote heterogeneous semi-physical simulation test task |
| CN111949329B (en) * | 2020-08-07 | 2022-08-02 | 苏州浪潮智能科技有限公司 | AI chip task processing method and device based on x86 architecture |
| CN111949329A (en) * | 2020-08-07 | 2020-11-17 | 苏州浪潮智能科技有限公司 | AI chip task processing method and device based on x86 architecture |
| CN112527676A (en) * | 2020-12-23 | 2021-03-19 | 中移（杭州）信息技术有限公司 | Model automation test method, device and storage medium |
| WO2023036280A1 (en) * | 2021-09-10 | 2023-03-16 | 华为技术有限公司 | Model test method and apparatus |
| WO2023050885A1 (en) * | 2021-09-28 | 2023-04-06 | 华为技术有限公司 | Application performance testing method, method and apparatus for establishing performance testing model |
| CN116069603A (en) * | 2021-09-28 | 2023-05-05 | 华为技术有限公司 | Performance test method of application, method and device for establishing performance test model |
| CN116069603B (en) * | 2021-09-28 | 2023-12-08 | 华为技术有限公司 | Performance test method of application, method and device for establishing performance test model |
| CN113779913A (en) * | 2021-11-12 | 2021-12-10 | 浙江大学 | Verification platform structure and test method for AI multi-chip system |
| CN117407299A (en) * | 2023-10-18 | 2024-01-16 | 北京大学 | Model test method and system |
| CN117407299B (en) * | 2023-10-18 | 2024-05-07 | 北京大学 | Model test method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109634843B (en) | 2021-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109634843A (en) | A kind of distributed automatization method for testing software and platform towards AI chip platform | |
| CN109960643B (en) | Code testing method and device | |
| US10430319B1 (en) | Systems and methods for automatic software testing | |
| KR101132560B1 (en) | System and method for automatic interface testing based on simulation for robot software components | |
| CN109684215A (en) | A kind of automated software Quality System Review and iteratively faster method | |
| Maâlej et al. | Distributed and Resource-Aware Load Testing of WS-BPEL Compositions. | |
| CN106970880B (en) | Distributed automatic software testing method and system | |
| US20090089039A1 (en) | System and method of emulating functionality of a web service | |
| US20170220324A1 (en) | Data communication accelerator system | |
| CN106612204B (en) | Service checking method and device | |
| CN102566443A (en) | Simulation verification system and method for integrated avionics system model based on artifact design description language (ADDL) | |
| WO2021097824A1 (en) | Code quality and defect analysis method, server and storage medium | |
| CN109977012B (en) | Joint debugging test method, device, equipment and computer readable storage medium of system | |
| CN110413316A (en) | Non-intrusion type test macro and method towards block chain intelligence contract | |
| de Oliveira et al. | Comparative evaluation of the maintainability of RESTful and SOAP-WSDL web services | |
| CN113704095A (en) | Automatic generation method and device for automatic interface test case | |
| Xiao et al. | A framework for verifying sla compliance in composed services | |
| US9354962B1 (en) | Memory dump file collection and analysis using analysis server and cloud knowledge base | |
| US11113105B1 (en) | Computer implemented system and method for generating platform agnostic digital worker | |
| Cooray et al. | Test reconfiguration for service oriented applications | |
| Zhang et al. | The implementation of a dependency matrix-based QoS diagnosis support in SOA middleware | |
| Jacob et al. | Scrutiny on various approaches of software performance testing tools | |
| CN110908918A (en) | Unit testing method and device for multiple interdependent node | |
| Traore et al. | Software performance modeling using the UML: a case study | |
| Lübke | Calculating test coverage for BPEL processes with process log analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |