System for testing allopatric closure of artificial intelligence chip

Abstract

The invention discloses a system for testing the remote closure of an artificial intelligence chip, and belongs to the technical field of chip testing. Comprising a test set management module used for generating a customized test set according to test requirements; the intelligent gateway module is used for monitoring the network connection state of the machine room and the test room, and the test room is internally provided with test terminal equipment; and the intelligent test module is connected with the test set management module and the intelligent gateway module, and the test terminal equipment analyzes the artificial intelligence chip according to the customized test set, generates a test report and outputs the test report. The technical scheme has the beneficial effects that the remote test of the chip in an enterprise is realized, the confidentiality requirements of business and technology are met, the cost of the adaptation test of the domestic chip is reduced, and the security of the adaptation test is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of chip testing technology, and in particular to a chip off-site closedness testing system. Background Art

[0002] With the advent of the era of large artificial intelligence models, the demand for computing power is increasing. When building computing clusters, companies usually face the problem of how to choose AI (Artificial Intelligence) chips. When purchasing chips, they usually have to do a lot of testing, considering whether the performance and compatibility can meet their own business, as well as cost-effectiveness and other issues.

[0003] However, due to the confidentiality requirements of enterprises in terms of business and technology, the remote access methods provided by chip manufacturers are usually not used, because this will bring uncontrollable risks. Chip manufacturers are usually required to provide batch servers and related supporting equipment to complete the testing work in the enterprise's computer room, but long-distance equipment relocation, installation, and debugging will extend the testing cycle, increase risks, and bring additional costs and burdens to enterprises and manufacturers. Summary of the invention

[0004] The purpose of the present invention is to provide a system for off-site closed testing of artificial intelligence chips to solve the above technical problems; a system for off-site closed testing of artificial intelligence chips, comprising: a test set management module, used to generate customized test sets according to test requirements; an intelligent gateway module, used to monitor the network connection status between a computer room and a test room, wherein the test room is provided with a test terminal device; an intelligent test module, connecting the test set management module and the intelligent gateway module, wherein the test terminal device analyzes the artificial intelligence chip according to the customized test set, generates a test report and outputs it.

[0005] Preferably, the test set management module includes a project configuration library for pre-establishing multiple performance test projects; a customization unit connected to the project configuration library for combining the performance test projects according to the test requirements to obtain the customized test set.

[0006] Preferably, the intelligent gateway module includes a computer room gateway connected to a server cluster and used to collect network status data of the server cluster; a test room gateway connected to the computer room gateway and used to collect and send the network status data to the intelligent test module.

[0007] Preferably, the intelligent gateway module also includes an alarm unit, which is respectively connected to the computer room gateway and the test room gateway, and is used to monitor the resource utilization of the computer room gateway and the test room gateway by adopting a preset alarm trigger mechanism.

[0008] Preferably, the preset alarm triggering mechanism includes that when the resource utilization exceeds a set threshold, the alarm unit sends an alarm message to the test terminal device.

[0009] Preferably, the intelligent testing module includes: a performance monitoring unit, which uses a data acquisition card to collect performance data of the artificial intelligence chip; a real-time analysis unit, which is connected to the performance monitoring unit and uses a data analysis algorithm to analyze the performance data to obtain analysis results; and a report generation unit, which is connected to the real-time analysis unit and uses a report writing terminal to generate the test report according to the analysis results.

[0010] Preferably, the intelligent test module further comprises an encrypted storage unit, which is connected to the report generating unit and is used to store the physical environment data, power status and operation log of the test terminal device.

[0011] Preferably, it also includes an intelligent monitoring module connected to the intelligent gateway module, and the intelligent monitoring module includes a multi-dimensional monitoring unit, including a first monitoring component connected to the computer room gateway and used to record the environmental data of the computer room; a second monitoring component connected to the test room gateway and used to record the environmental data of the conference room in the test room; and a third monitoring component connected to the test room gateway and used to record the environmental data of the test terminal device.

[0012] Preferably, the intelligent monitoring module also includes a first training unit, connected to the multidimensional monitoring unit, for collecting historical monitoring data of the multidimensional monitoring unit to form first training data, and constructing an abnormality recognition model based on the first training data; an abnormality recognition unit, connected to the first training unit, for identifying abnormal behaviors of the computer room and the test room based on the abnormality recognition model and the real-time monitoring data of the multidimensional monitoring unit.

[0013] Preferably, it also includes a review and encryption module, which is connected to the intelligent testing module. The review and encryption module includes: a second training unit, which is used to collect historical test data of the intelligent testing module to form second training data, and build a review model based on the second training data; a review unit, which is connected to the second training unit and is used to review the test report according to the review model to generate a post-review report and an audit opinion; and an encrypted transmission unit, which is connected to the review unit and outputs the post-review report and the audit opinion through a secure channel.

[0014] The beneficial effects of the present invention are: by adopting the above technical solution, remote chip testing can be achieved within the enterprise, meeting the confidentiality requirements of business and technology, reducing the cost of domestic chip adaptation testing, and enhancing the security of adaptation testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a system diagram of the off-site closedness test of the artificial intelligence chip of the present invention; Figure 2 It is a system connection block diagram of the remote closedness test of the artificial intelligence chip of the present invention; Figure 3 is a schematic diagram of a test set management module of the present invention; Figure 4 is a schematic diagram of the intelligent gateway module of the present invention; Figure 5 is a schematic diagram of an intelligent test module of the present invention; Figure 6 is a schematic diagram of an intelligent monitoring module of the present invention; Figure 7 It is a schematic diagram of the review and encryption module of the present invention.

[0016] In the attached figure: 1. Test set management module; 11. Project configuration library; 12. Customization unit; 2. Intelligent gateway module; 21. Computer room gateway; 22. Test room gateway; 23. Alarm unit; 3. Intelligent testing module; 31. Performance monitoring unit; 32. Real-time analysis unit; 33. Report generation unit; 34. Encrypted storage unit; 4. Intelligent monitoring module; 41. Multi-dimensional monitoring unit; 411. First monitoring component; 412. Second monitoring component; 413. Third monitoring component; 42. First training unit; 43. Abnormal identification unit; 5. Review and encryption module; 51. Second training unit; 52. Review unit; 53. Encrypted transmission unit; 6. Computer room; 61. Server cluster; 7. Test room; 71. Test terminal equipment; 72. Report writing terminal. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0019] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0020] A system for off-site closed testing of artificial intelligence chips, such as Figure 1 , Figure 2As shown, it includes a test set management module 1, which is used to generate a customized test set according to the test requirements; an intelligent gateway module 2, which is used to monitor the network connection status of the computer room 6 and the test room 7, and the test room 7 is equipped with a test terminal device 71; an intelligent test module 3, which connects the test set management module 1 and the intelligent gateway module 2, and the test terminal device 71 analyzes the artificial intelligence chip according to the customized test set, generates a test report and outputs it.

[0021] Specifically, the present invention provides a system for off-site closed testing of artificial intelligence chips, which is helpful for the research and development and verification of domestic AI chips. The remote connection between the computer room 6 and the test room 7 is realized through the intelligent gateway module 2, which solves the security and efficiency problems in the traditional testing method.

[0022] In a preferred embodiment, referring to Figure 3 The test set management module 1 includes a project configuration library 11 for pre-establishing multiple performance test projects; a customization unit 12, connected to the project configuration library 11, for combining performance test projects according to test requirements to obtain a customized test set.

[0023] Specifically, the performance test items can be formulated according to the user's performance requirements for the chip, or can be formulated based on industry standards. The test content of the performance test items includes temperature, load conditions, and test time.

[0024] In a preferred embodiment, referring to Figure 4 The intelligent gateway module 2 includes a computer room gateway 21 connected to the server cluster 61 and used to collect network status data of the server cluster 61; a test room gateway 22 connected to the computer room gateway 21 and used to collect and send network status data to the intelligent test module 3.

[0025] Specifically, the computer room gateway 21 collects data from the server cluster 61 in the computer room 6, including status information and logs, and converts the different communication protocols used into standardized protocols to enable effective communication with the test room 7. The test room gateway 22 receives data from the computer room 6, processes it, and forwards it to the test equipment for analysis and testing.

[0026] In a preferred embodiment, the intelligent gateway module 2 also includes an alarm unit 23, which is respectively connected to the computer room gateway 21 and the test room gateway 22, and is used to monitor the resource utilization of the computer room gateway 21 and the test room gateway 22 using a preset alarm trigger mechanism; the preset alarm trigger mechanism includes that when the resource utilization exceeds a set threshold, the alarm unit 23 sends an alarm message to the test terminal device 71.

[0027] Specifically, the alarm unit 23 can monitor the operating status of the equipment in the computer room 6 in real time, and manage and control it, such as restarting the equipment, updating the configuration, etc.

[0028] For example, CPU utilization exceeding 90% or memory utilization exceeding 80% will be considered abnormal behavior. The alarm unit 23 compares the status data received in real time with the set threshold. If a certain status data exceeds the preset threshold, the alarm unit 23 will identify it as abnormal behavior.

[0029] In a preferred embodiment, referring to Figure 5 The intelligent testing module 3 includes a performance monitoring unit 31, which uses a data acquisition card to collect performance data of the artificial intelligence chip; a real-time analysis unit 32, which is connected to the performance monitoring unit 31 and uses a data analysis algorithm to analyze the performance data to obtain analysis results; a report generation unit 33, which is connected to the real-time analysis unit 32 and uses a report writing terminal 72 to generate a test report according to the analysis results; the intelligent testing module 3 also includes an encryption storage unit 34, which is connected to the report generation unit 33 and is used to store the physical environment data, power status and operation log of the test terminal device 71.

[0030] Specifically, the encryption storage unit 34 uses public keys and private keys to encrypt and decrypt data, which can prevent information leakage. The encryption storage unit 34 stores the physical environment data, power supply status and operation log of the test terminal device 71 to facilitate subsequent auditing.

[0031] Data analysis algorithms (such as statistical analysis and machine learning models) observe the changing trends of performance data over time through time series analysis and process performance data in real time.

[0032] In a preferred embodiment, it also includes an intelligent monitoring module 4, connected to the intelligent gateway module 2, referring to Figure 6 The intelligent monitoring module 4 includes a multi-dimensional monitoring unit 41, including a first monitoring component 411, connected to the computer room gateway 21, used to record the environmental data of the computer room 6; a second monitoring component 412, connected to the test room gateway 22, used to record the environmental data of the conference room in the test room 7; a third monitoring component 413, connected to the test room gateway 22, used to record the environmental data of the test terminal device 71.

[0033] Specifically, the multi-dimensional monitoring unit 41 supports remote access. The user logs in to the management interface of the multi-dimensional monitoring unit 41 (usually accesses the IP address of the device through a web browser) and sets the user account and password. Find and enable the remote access function (for example, HTTP, HTTPS, RTSP and other protocols) in the device settings and configure it according to the requirements of the device manual.

[0034] The first monitoring component 411 is arranged in the computer room 6, the first monitoring component 411 is wirelessly connected to the computer room gateway 21, the computer room gateway 21 is wiredly connected to the server cluster 61, and the computer room gateway 21 and the test room gateway 22 encrypt data transmission via a public network.

[0035] The second monitoring component 412 and the third monitoring component 413 are wirelessly connected to the test room gateway 22 respectively.

[0036] The first monitoring component 411, the second monitoring component 412, and the third monitoring component 413 can record the temperature, humidity, air quality, and equipment status in the space.

[0037] In a preferred embodiment, the intelligent monitoring module 4 also includes a first training unit 42, connected to the multi-dimensional monitoring unit 41, for collecting historical monitoring data of the multi-dimensional monitoring unit 41 to form first training data, and constructing an abnormality recognition model based on the first training data; an abnormality recognition unit 43, connected to the first training unit 42, for identifying abnormal behaviors of the computer room 6 and the test room 7 based on the abnormality recognition model and the real-time monitoring data of the multi-dimensional monitoring unit 41.

[0038] Specifically, the first training unit 42 stores the historical monitoring data collected by the multi-dimensional monitoring unit 41 in a database or data warehouse, removes noise, outliers and missing values ​​in turn, and then converts the data into a unified format for subsequent processing. Finally, features that are helpful for abnormality identification, such as peak values, trend changes, etc., are extracted from the original data.

[0039] The processed real-time monitoring data is input into the anomaly recognition model to generate prediction results. According to the output of the model, it is judged whether the real-time data is abnormal. If so, the corresponding behavior and status are marked. If not, monitoring continues.

[0040] In a preferred embodiment, it also includes a review and encryption module 5, connected to the intelligent test module 3, referring to Figure 7 The review and encryption module 5 includes a second training unit 51, which is used to collect historical test data of the intelligent test module 3, form second training data, and build a review model based on the second training data; a review unit 52, which is connected to the second training unit 51, and is used to review the test report according to the review model, and generate a post-review report and review opinions; an encrypted transmission unit 53, which is connected to the review unit 52 and outputs the post-review report and review opinions through a secure channel.

[0041] Specifically, historical test data including test results, test parameters, environment settings and other information are extracted from the intelligent test module 3 and collected through database query, API call and other methods.

[0042] Clean and organize the collected test data to ensure data quality. For example, remove duplicate data, fill in missing values, and standardize data formats. Divide into training sets and test sets to facilitate subsequent model training and verification.

[0043] Use machine learning or deep learning algorithms (such as decision trees, random forests, neural networks, etc.) to train the model using the training set. Use the test set to verify the performance of the model, and evaluate the accuracy and robustness of the model through indicators such as cross-validation and confusion matrix.

[0044] The test report to be reviewed is used as input and analyzed by the review model. The model will evaluate the data in the test report based on historical test data and learned rules.

[0045] Generate post-audit reports and audit opinions. Based on the evaluation results of the model, a post-audit report is automatically generated, which contains the evaluation conclusions, possible errors, and suggested modifications. At the same time, an audit opinion is formed to provide more in-depth analysis and suggestions for subsequent decision-making reference.

[0046] Specifically, the present invention is applied to the reasoning performance test of a domestic artificial intelligence chip, and the off-site closed test of the chip is implemented as follows: Step S0, deploy the computer room 6 and the test room 7 in a closed test environment; Step S1, establish a project configuration library 11 through the test set management module 1, and the project configuration library 11 includes multiple performance test items to be used; Step S2, based on the test requirements, select the required performance test items in the project configuration library 11 to form a customized test set; Step S3, monitor the network connection status of the computer room 6 and the test room 7 based on the intelligent gateway module 2 to prevent unauthorized access; Step S4, the intelligent testing module 3 uses the test terminal device 71 to analyze the artificial intelligence chip according to the customized test set, the report writing terminal 72 (vpn client) and the test terminal device 71 (ssh client) are remotely on the same screen, and the report writing terminal 72 (vpn client) outputs the test report; Step S5, the intelligent monitoring module 4 monitors the environmental data of the computer room 6 and the test room 7 in real time to ensure the isolation and security of the test.

[0047] During the entire test process, the intelligent monitoring module 4 monitors the test environment in real time to ensure the isolation and security of the test, monitors the physical environment and power status of the test equipment, and records all operation logs for subsequent audits. The intelligent gateway monitors the network connection status to prevent unauthorized access. After the test is completed, the system automatically cleans up the test environment, encrypts and stores the test data, and sends the result report to the relevant parties for review through a secure channel.

[0048] In this way, the domestically produced closed off-site testing system for artificial intelligence chips can comprehensively and efficiently evaluate the various performance indicators of artificial intelligence chips while ensuring a high degree of security and isolation, providing reliable data support for chip optimization and application scenario matching.

[0049] The above description is only a preferred embodiment of the present invention, and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A system for off-site closed testing of artificial intelligence chips, characterized in that: It includes, a test set management module, which is used to generate customized test sets according to test requirements; An intelligent gateway module, used to monitor the network connection status of a computer room and a test room, wherein the test room is provided with a test terminal device; The intelligent testing module is connected to the test set management module and the intelligent gateway module. The test terminal device analyzes the artificial intelligence chip according to the customized test set, generates a test report and outputs it.

2. The system for remote closedness testing of artificial intelligence chips according to claim 1 is characterized in that: The test set management module includes a project configuration library for pre-establishing multiple performance test projects; a customization unit connected to the project configuration library for combining the performance test projects according to the test requirements to obtain the customized test set.

3. The system for remote closedness testing of artificial intelligence chips according to claim 1 is characterized in that: The intelligent gateway module includes a computer room gateway connected to a server cluster and used to collect network status data of the server cluster; The test room gateway is connected to the computer room gateway and is used to collect and send the network status data to the intelligent test module.

4. The system for remote closedness testing of artificial intelligence chips according to claim 3 is characterized in that: The intelligent gateway module also includes an alarm unit, which is connected to the computer room gateway and the test room gateway respectively, and is used to monitor the resource utilization of the computer room gateway and the test room gateway by using a preset alarm trigger mechanism.

5. The system for remote closedness testing of artificial intelligence chips according to claim 4 is characterized in that: The preset alarm triggering mechanism includes that when the resource utilization exceeds a set threshold, the alarm unit sends an alarm message to the test terminal device.

6. The system for remote closedness testing of artificial intelligence chips according to claim 1 is characterized in that: The intelligent testing module includes a performance monitoring unit, which uses a data acquisition card to collect performance data of the artificial intelligence chip; A real-time analysis unit, connected to the performance monitoring unit, uses a data analysis algorithm to analyze the performance data to obtain an analysis result; A report generating unit is connected to the real-time analyzing unit and generates the test report according to the analysis result by using a report writing terminal.

7. The system for remote closedness testing of artificial intelligence chips according to claim 6 is characterized in that: The intelligent test module further includes an encrypted storage unit, which is connected to the report generating unit and is used to store the physical environment data, power supply status and operation log of the test terminal device.

8. The system for remote closedness testing of artificial intelligence chips according to claim 3 is characterized in that: It also includes an intelligent monitoring module connected to the intelligent gateway module, and the intelligent monitoring module includes a multi-dimensional monitoring unit, including a first monitoring component connected to the computer room gateway and used to record the environmental data of the computer room; A second monitoring component, connected to the test room gateway, for recording environmental data of the conference room in the test room; The third monitoring component is connected to the test room gateway and is used to record the environmental data of the test terminal equipment.

9. The system for remote closedness testing of artificial intelligence chips according to claim 8 is characterized in that: The intelligent monitoring module further includes a first training unit connected to the multi-dimensional monitoring unit, configured to collect historical monitoring data of the multi-dimensional monitoring unit to form first training data, and to construct an abnormality recognition model according to the first training data; An abnormality identification unit is connected to the first training unit and is used to identify abnormal behaviors of the computer room and the test room according to the abnormality identification model and the real-time monitoring data of the multi-dimensional monitoring unit.

10. The system for remote closedness testing of artificial intelligence chips according to claim 8, characterized in that: It also includes a review and encryption module connected to the intelligent test module, the review and encryption module includes a second training unit for collecting historical test data of the intelligent test module to form second training data, and constructing a review model according to the second training data; A review unit, connected to the second training unit, for reviewing the test report according to the review model and generating a post-review report and review opinions; The encrypted transmission unit is connected to the review unit and outputs the post-review report and the review opinion through a secure channel.

Images