Method, device and electronic device for determining parameter tuning information

By training the neural network model to process historical data and environmental information, the parameter tuning search space is reduced, and the Bayesian algorithm is inefficient in high-dimensional parameter space is solved, and fast and accurate parameter tuning is achieved.

CN118396038BActive Publication Date: 2024-11-12ALIBABA CLOUD COMPUTING CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410825446.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-24
Publication Date
2024-11-12
Estimated Expiration
2044-06-24

AI Technical Summary

Technical Problem

In the prior art, Bayesian algorithms are inefficient when determining parameter tuning information in high-dimensional parameter space, resulting in a long time-consuming parameter tuning process.

Method used

By training a neural network model, the parameter value range is determined using historical parameter tuning data, and the current environment information is processed based on the model, reducing the search space to improve efficiency and accuracy.

Benefits of technology

It realizes the rapid and accurate determination of parameter tuning information in high-dimensional parameter space, improves parameter tuning efficiency and accuracy, and solves the problem of low efficiency of Bayesian algorithms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118396038B_ABST
    Figure CN118396038B_ABST
Patent Text Reader

Abstract

The present application discloses a method, device and electronic device for determining parameter tuning information. Relating to the field of artificial intelligence, the method comprises: determining the environmental information of the current operating environment of the target application, wherein the target application is the application to be parameter tuned; processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; determining the parameter tuning information of the target application based on the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application. The present application solves the technical problem of low efficiency in determining parameter tuning information in the related art using the Bayesian algorithm to determine parameter tuning information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and more specifically, to a method, device and electronic device for determining parameter tuning information. Background Art

[0002] There are requirements for parameter tuning in scenarios such as operating system kernels, compilers, and databases. Effective parameter settings can optimize service performance and improve overall service quality. However, the number of parameters involved in the above fields is large, which may be as high as hundreds or thousands of dimensions, and traditional tuning methods often rely on experienced engineers to perform manual tuning based on domain knowledge. Therefore, high-dimensional parameters pose a great challenge to manual tuning, which may not only lead to low tuning efficiency, but also affect the tuning effect.

[0003] The current mainstream tuning algorithm is based on Bayesian theory. According to the tuning effect of real-time feedback, it dynamically searches in the preset parameter space to determine the parameter tuning information. This algorithm framework is relatively general. In recent years, it has not only been used in parameter tuning scenarios related to computer systems, but also widely used in hyperparameter tuning scenarios of deep learning models. However, when there are many parameters and the dimension of the parameter space is high, the Bayesian algorithm often requires a long search cycle, resulting in the problem of low efficiency in determining parameter tuning information.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide a method, device and electronic device for determining parameter tuning information, so as to at least solve the technical problem of low efficiency in determining parameter tuning information when using Bayesian algorithm to determine parameter tuning information in related technologies.

[0006] According to one aspect of an embodiment of the present application, a method for determining parameter tuning information is provided, including: determining environmental information of a current operating environment of a target application, wherein the target application is an application to be parameter tuned; processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; and determining the parameter tuning information of the target application based on the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application.

[0007] Further, at least one application includes a target application, and the first neural network model learns the parameters required for application parameter tuning and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning, and processes the environmental information through the first neural network model to obtain the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning, including: determining the multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning learned by the first neural network model; determining the target operating environment from the operating environment learned by the first neural network model based on the environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; determining the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0008] Furthermore, determining the parameter tuning information of the target application based on the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning includes: generating multiple groups of parameter combinations based on the multiple parameters required for the target application tuning and the parameter value ranges corresponding to the parameters; determining the tuning data information based on the parameter combinations in the multiple groups of parameter combinations, and constructing a first training sample set based on the tuning data information corresponding to the multiple groups of parameter combinations; training the first neural network model based on the first training sample set to obtain the second neural network model, and determining the parameter tuning information through the second neural network model.

[0009] Furthermore, determining tuning data information according to a parameter combination in a plurality of parameter combinations, and constructing a first training sample set according to the tuning data information corresponding to the plurality of parameter combinations includes: for a parameter combination in a plurality of parameter combinations, performing parameter tuning on a target application based on the parameter combination, and performing tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; combining the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain the tuning data information; and using the tuning data information corresponding to the plurality of parameter combinations as training samples and the application type of the target application as a true label to construct a first training sample set.

[0010] Furthermore, a tuning evaluation is performed on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination, including: determining multiple evaluation indicators required for tuning evaluation of the target application to obtain multiple target evaluation indicators; and a tuning evaluation is performed on the target application after parameter tuning based on the multiple target evaluation indicators to obtain a tuning evaluation result corresponding to the parameter combination.

[0011] Furthermore, the data information of historical parameter tuning includes historical tuning evaluation results of the application. The first neural network model learns, based on the historical tuning evaluation results, multiple evaluation indicators used in historical tuning evaluation of at least one application, and determines multiple evaluation indicators required for tuning evaluation of the target application. Obtaining multiple target evaluation indicators includes: determining multiple target evaluation indicators based on the multiple evaluation indicators used in historical tuning evaluation that have been learned by the first neural network model.

[0012] Furthermore, determining multiple target evaluation indicators based on multiple evaluation indicators used in historical tuning evaluations that have been learned by the first neural network model includes: determining multiple evaluation indicators used in historical tuning evaluations of the target application that have been learned by the first neural network model as multiple initial evaluation indicators; upon receiving an indicator update instruction from the user, extracting indicator update information from the indicator update instruction, and determining multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; and upon not receiving an indicator update instruction from the user, determining the multiple initial evaluation indicators as multiple target evaluation indicators.

[0013] Furthermore, training the first neural network model based on the first training sample set to obtain a second neural network model, and determining parameter tuning information through the second neural network model includes: determining weights corresponding to target evaluation indicators among multiple target evaluation indicators; training the first neural network model based on the first training sample set and the weights corresponding to multiple target evaluation indicators to obtain a second neural network model, and determining parameter tuning information through the second neural network model.

[0014] Furthermore, the data information of historical parameter tuning includes historical tuning evaluation results of the application, and the first neural network model learns the weight of the evaluation indicator according to the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results, and determines the weight corresponding to the target evaluation indicator among multiple target evaluation indicators, including: determining the weights of multiple target evaluation indicators according to the weights of the evaluation indicators learned by the first neural network model.

[0015] Furthermore, the first neural network model is obtained in the following manner: obtaining data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; using the data information of historical parameter tuning as training samples and the application type of the application as the true label to construct a second training sample set; training the initial neural network model based on the second training sample set to obtain the first neural network model.

[0016] According to another aspect of an embodiment of the present application, a method for determining parameter tuning information is also provided, including: obtaining environmental information of a current operating environment of a target application uploaded by a client, wherein the target application is an application to be parameter tuned; processing the environmental information through a first neural network model in a cloud server to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, and determining the parameter tuning information of the target application based on the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; and feeding back the parameter tuning information of the target application to the client.

[0017] According to another aspect of an embodiment of the present application, a device for determining parameter tuning information is also provided, including: a first determination unit, used to determine environmental information of a current operating environment of a target application, wherein the target application is an application to be parameter tuned; a processing unit, used to process the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; a second determination unit, used to determine the parameter tuning information of the target application based on the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application.

[0018] Furthermore, at least one application includes a target application, and the first neural network model learns the parameters required for application parameter tuning and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning. The processing unit includes: a first determination subunit, used to determine multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning that have been learned by the first neural network model; a second determination subunit, used to determine the target operating environment from the operating environment that has been learned by the first neural network model based on the environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; a third determination subunit, used to determine the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0019] Furthermore, the second determination unit includes: a generation subunit, used to generate multiple groups of parameter combinations according to multiple parameters required for target application tuning and parameter value ranges corresponding to the parameters; a construction subunit, used to determine tuning data information based on parameter combinations in the multiple groups of parameter combinations, and construct a first training sample set based on the tuning data information corresponding to the multiple groups of parameter combinations; a fourth determination subunit, used to train the first neural network model based on the first training sample set to obtain a second neural network model, and determine the parameter tuning information through the second neural network model.

[0020] Furthermore, the construction subunit includes: a first processing module, which is used to perform parameter tuning on a target application based on a parameter combination in a plurality of parameter combinations, and to perform tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; a second processing module, which is used to combine the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain tuning data information; and a construction module, which is used to use the tuning data information corresponding to the plurality of parameter combinations as training samples and the application type of the target application as a true label to construct a first training sample set.

[0021] Furthermore, the second processing module includes: a first determination submodule, used to determine multiple evaluation indicators required for tuning and evaluating the target application, and obtain multiple target evaluation indicators; a first processing submodule, used to perform tuning and evaluation on the target application after parameter tuning based on multiple target evaluation indicators, and obtain tuning evaluation results corresponding to the parameter combination.

[0022] Furthermore, the data information of historical parameter tuning includes historical tuning evaluation results of the application, and the first neural network model learns multiple evaluation indicators used in historical tuning evaluation of at least one application based on the historical tuning evaluation results. The first determination submodule includes: a second determination submodule, which is used to determine multiple target evaluation indicators based on the multiple evaluation indicators used in the historical tuning evaluation that have been learned by the first neural network model.

[0023] Furthermore, the second determination submodule includes: a third determination submodule, used to determine multiple evaluation indicators used in historical tuning evaluation of the target application that have been learned by the first neural network model as multiple initial evaluation indicators; a fourth determination submodule, used to extract indicator update information from the indicator update instruction when receiving the user's indicator update instruction, and determine multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; a fifth determination submodule, used to determine multiple initial evaluation indicators as multiple target evaluation indicators when no indicator update instruction is received from the user.

[0024] Furthermore, the fourth determination subunit includes: a first determination module, used to determine the weights corresponding to the target evaluation indicators among multiple target evaluation indicators; a second determination module, used to train the first neural network model based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain the second neural network model, and determine the parameter tuning information through the second neural network model.

[0025] Furthermore, the data information of historical parameter tuning includes historical tuning evaluation results of the application, the first neural network model learns the weight of the evaluation indicator according to the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results, and the first determination module includes: a sixth determination submodule, which is used to determine the weights of multiple target evaluation indicators according to the weights of the evaluation indicators learned by the first neural network model.

[0026] Furthermore, the device for determining parameter tuning information also includes: an acquisition unit, used to acquire data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; a construction unit, used to use the data information of historical parameter tuning as a training sample and the application type of the application as a true label to construct a second training sample set; and a training unit, used to train the initial neural network model based on the second training sample set to obtain a first neural network model.

[0027] According to another aspect of an embodiment of the present invention, an electronic device is provided, including: a memory storing an executable program; and a processor for running the program, wherein when the program is running, any one of the above-mentioned methods for determining parameter tuning information is executed.

[0028] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for determining parameter tuning information.

[0029] In an embodiment of the present application, by determining the environmental information of the current operating environment of the target application, wherein the target application is an application to be parameter tuned; processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; determining the parameter tuning information of the target application according to the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, the first neural network model is trained based on data information of historical parameter tuning of at least one application, so that the first neural network model can learn the information required for parameter tuning of different applications from historical data. By processing the environmental information through the first neural network model to obtain the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, on the one hand, the search space for parameter tuning information is narrowed, and the efficiency of determining parameter tuning information is improved; on the other hand, the search space for parameter tuning information is determined in a targeted manner based on historical data and environmental information, and the accuracy of determining parameter tuning information is improved. Thereby, the purpose of determining the parameter tuning information of the target application based on a specific search space is achieved, and the technical effect of effectively improving the efficiency of determining the parameter tuning information is realized, thereby solving the technical problem of low efficiency in determining the parameter tuning information by using the Bayesian algorithm in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0031] Figure 1 is a schematic diagram of a computer terminal provided according to Embodiment 1 of the present application;

[0032] Figure 2 is a flow chart of a method for determining parameter tuning information provided in the related art;

[0033] Figure 3 The process of determining the parameter tuning information according to the first embodiment of the present application is as follows Figure 1 ;

[0034] Figure 4 The process of determining the parameter tuning information according to the first embodiment of the present application is as follows Figure 2 ;

[0035] Figure 5 is a flowchart of a method for determining parameter tuning information provided in Embodiment 2 of the present application;

[0036] Figure 6is a schematic diagram of a device for determining parameter tuning information provided in Embodiment 3 of the present application;

[0037] Figure 7 It is a structural block diagram of an electronic device provided according to Embodiment 4 of the present application. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0040] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards in the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0041] First, some nouns or terms that appear in the description of the embodiments of the present application are subject to the following explanations:

[0042] Parameter tuning: The process of continuously improving the system's score on a benchmark program by adjusting system parameters.

[0043] System parameters: refers to configuration items in the operating system that can be dynamically set and affect application performance, such as kernel configuration parameters and application configuration parameters.

[0044] Benchmark program: An evaluation program used to estimate the performance of a tuning application and give a performance score.

[0045] Parameter tuning algorithm: A program that implements a certain algorithm strategy and continuously gives system parameter combinations to find parameter configurations that improve application performance.

[0046] Proxy model: A mathematical model that constructs a mapping relationship between system parameter combinations and benchmark program scores, and uses this to guide the operation of the parameter tuning algorithm.

[0047] LLM: Large Language Model, large language model.

[0048] Example 1

[0049] According to an embodiment of the present application, a method for determining parameter tuning information is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0050] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for determining parameter tuning information. Figure 1 As shown, the computer terminal (or mobile device) 10 may include a processor set 102 (the processor set 102 may include but is not limited to a processing device such as a microprocessor (Microcontroller Unit, referred to as MCU) or a programmable logic device (Field Programmable Gate Array, referred to as FPGA), and the processor set 102 may include a processor set, Figure 1 102a, 102b, ..., 102n are used to illustrate), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. Those skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0051] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuits may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device).

[0052] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method for determining parameter tuning information in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the method for determining the parameter tuning information described above is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0053] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0054] The display may be a touch screen type liquid crystal display, which may enable a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0055] There are requirements for parameter tuning in scenarios such as operating system kernels, compilers, and databases. Effective parameter settings can optimize service performance and improve overall service quality. However, the number of parameters involved in the above fields is large, which may be as high as hundreds or thousands of dimensions, and traditional tuning methods often rely on experienced engineers to perform manual tuning based on domain knowledge. Therefore, high-dimensional parameters pose a great challenge to manual tuning, which may not only lead to low tuning efficiency, but also affect the tuning effect.

[0056] The current mainstream tuning algorithm is based on Bayesian theory, using Gaussian Processes (GP) as a proxy model to simulate the behavior of the objective function. The key to the Bayesian algorithm is how to decide the next parameter point to be sampled. It uses a proxy model to achieve a balance between exploration and mining; exploration refers to finding potential better points in areas where the proxy model has high uncertainty in prediction; while mining refers to verifying whether the areas where the proxy model has better prediction effects (i.e., lower or higher function values) are truly the global best solution. Therefore, the Bayesian algorithm is a typical regression series algorithm, which is unsupervised, but requires real-time feedback to obtain evaluation indicators to generate the next sampling point.

[0057] Figure 2 is a flow chart of a method for determining parameter tuning information provided in the related art, Figure 2 The specific process of using the Bayesian algorithm in the field of parameter tuning is shown. Figure 2 As shown, in the process of determining the parameter tuning information, a system parameter combination is given based on a parameter tuning algorithm (such as random search, Bayesian optimization, a tree-structured Bayesian optimization algorithm TPE (Tree-structured Parzen Estimator)), and then parameter tuning is performed based on the system parameter combination, and a benchmark program is run to determine the benchmark program score of this tuning, and then a new system parameter combination is determined based on the relationship between the system parameter combination and the benchmark program score, and it is continuously iterated in this way to determine the parameter tuning information.

[0058] The above algorithm framework is relatively general. In recent years, it has been used not only in computer system-related parameter tuning scenarios, but also in deep learning model hyperparameter tuning scenarios. Figure 2 In the process of determining parameter tuning information shown, the Bayesian algorithm needs to execute the benchmark program or obtain online performance indicators for each round of sampling points generated. Therefore, when there are many parameters and the parameter space dimension is high, the above Bayesian algorithm often requires a longer search cycle. If you want to obtain relatively effective tuning results, the number of running rounds needs to be more than twice the number of parameters, and preferably 3-5 times. This means that if 500 parameters are tuned for performance, the number of running rounds each time is at least 1000 rounds, and preferably 1500-2500 rounds. The execution time of many benchmark programs for one round is from several minutes to hundreds of minutes, which will result in too long time required to actually determine the parameter tuning information, and there is a problem of low efficiency in determining the parameter tuning information.

[0059] In the above technical background, the present application provides Figure 3 The method for determining the parameter tuning information shown. Figure 3 The process of determining the parameter tuning information according to the first embodiment of the present application is as follows Figure 1 The method comprises:

[0060] Step S301 : determining the environment information of the current running environment of the target application, wherein the target application is an application to be parameter-tuned.

[0061] Optionally, electronic devices, application systems, servers and other devices can be used as the execution subject of the present application. In this embodiment, the parameter tuning system is used as the execution subject to execute the method for determining parameter tuning information. The parameter tuning system at least includes a first neural network model, and the first neural network model is used to process the environmental information corresponding to the target application. Among them, the target application includes but is not limited to database applications, server applications, big data applications, etc. Database applications include but are not limited to relational databases MySQL, Redis (Remote Dictionary Server, remote dictionary service), etc. Server applications include but are not limited to reverse proxy Web (World Wide Web, World Wide Web) servers, and big data applications include but are not limited to applications based on distributed storage and computing frameworks, or open source computing frameworks for large-scale data processing and other frameworks.

[0062] Optionally, application information of the application to be parameter tuned can be preset in the parameter tuning system, so that when the training of the first neural network model is completed, the target application is automatically determined according to the application information of the application to be parameter tuned; the parameter tuning system can also determine the target application according to the application tuning instruction when receiving the application tuning instruction sent by the user, wherein the application tuning instruction is used to indicate the application to be parameter tuned.

[0063] After determining the target application, the parameter tuning system can collect environmental information of the current operating environment of the target application. Environmental information includes, but is not limited to, CPU (Central Processing Unit) information, memory information, information about the operating system where the application is located, disk space, graphics card information, etc. of the current operating environment of the target application, where CPU information may refer to the number of CPU cores, memory information may refer to memory capacity, information about the operating system where the application is located may refer to the system type and version number of the operating system where the application is located, and graphics card information may refer to the graphics card model or performance. The parameter tuning system can collect the above environmental information by using a system detection tool to detect the current operating environment of the target application, or by checking the configuration file of the target application, or by executing system commands in the current operating environment of the target application.

[0064] Step S302, processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for target application parameter tuning, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application.

[0065] Optionally, a second training sample set can be constructed based on data information of historical parameter tuning of at least one application, so as to obtain the first neural network model by training based on the second training sample set. For example, the data information of historical parameter tuning of the application includes the parameters, parameter values ​​and historical tuning evaluation results used by the application in historical parameter tuning, and the data information of historical parameter tuning is used as a training sample, and the application type of the application is used as a true label to construct the second training sample set. For another example, the data information of historical parameter tuning of the application includes the parameters and parameter values ​​used by the application in historical parameter tuning, or includes the parameters, parameter values ​​and historical tuning evaluation results used by the application in historical parameter tuning, and the data information of historical parameter tuning is used as a training sample, and the environmental information of the historical operating environment of the application is used as a true label to construct the second training sample set. For another example, the data information of historical parameter tuning of the application includes the parameters and parameter values ​​used by the application in historical parameter tuning, or includes the parameters, parameter values ​​and historical tuning evaluation results used by the application in historical parameter tuning, and the data information of historical parameter tuning is used as a training sample, and the application type and the environmental information of the historical operating environment of the application are used as true labels to construct the second training sample set.

[0066] After determining the second training sample set, the first neural network model can be trained based on the second training sample set, and the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application can be determined by the trained first neural network model. In some embodiments, the multiple parameters required for parameter tuning of the target application can be preset, or can be learned by the first neural network model from data information of historical parameter tuning of at least one application, so that after determining the multiple parameters required for parameter tuning of the target application, the parameter value ranges corresponding to the parameters are determined.

[0067] For example, in the process of training the first neural network model based on the second training sample set, the first neural network model can learn the parameters required for application parameter tuning and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning. Therefore, multiple parameters required for parameter tuning of the target application can be determined based on the learned information, and after determining the environmental information of the current operating environment of the target application, the parameter value ranges corresponding to the multiple parameters are determined based on the learned information and the environmental information of the current operating environment of the target application. Optionally, the parameter value ranges corresponding to different parameters required for parameter tuning of the target application in the same operating environment are not exactly the same, and the parameter value ranges corresponding to the same parameters required for parameter tuning of the target application in different operating environments are not exactly the same.

[0068] Step S303: determining parameter tuning information of the target application according to parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application.

[0069] Optionally, the parameter tuning information may include multiple parameters required for tuning the parameters of the target application and parameter values ​​corresponding to the parameters.

[0070] Optionally, the parameter tuning information of the target application can be determined based on the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning. For example, for a parameter among the multiple parameters, select any one value from the parameter value range corresponding to the parameter as the parameter value of the parameter, so that after the parameter values ​​corresponding to the multiple parameters required for the target application parameter tuning are determined, the parameter tuning information is obtained. For another example, for a parameter among the multiple parameters, select an intermediate value from the parameter value range corresponding to the parameter as the parameter value of the parameter, so that after the parameter values ​​corresponding to the multiple parameters required for the target application parameter tuning are determined, the parameter tuning information is obtained. For another example, the multiple parameters required for the target application parameter tuning and the parameter value ranges corresponding to the parameters are processed by other neural network models to determine the parameter tuning information. Among them, other neural network models refer to models other than the first neural network model, and other neural network models may be obtained by training the first neural network model, or may not be obtained by training the first neural network model.

[0071] It should be noted that when the Bayesian algorithm is used to determine the parameter tuning information in the related art, each initiated tuning is independent and the data of the historical tuning tasks cannot be used. For example, 1000 rounds of parameter tuning of MySQL are performed 100 times, and each tuning is independently run for 1000 rounds. Then, repeated sampling points may be repeated in 100 times, and it is impossible to find a better sampling point based on the 100*1000 tunings of the full run. In the present application, a first neural network model is obtained by training based on the data information of the historical parameter tuning of at least one application, so that the first neural network model can learn the tuning parameter information required for different applications from the historical data, thereby improving the efficiency and accuracy of determining the parameter tuning information.

[0072] In this solution, a first neural network model is obtained by training data information of historical parameter tuning of at least one application, so that the first neural network model can learn the information required for parameter tuning of different applications from historical data. By processing the environmental information by the first neural network model, the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application are obtained. On the one hand, the search space for parameter tuning information is narrowed and the efficiency of determining parameter tuning information is improved. On the other hand, the search space for parameter tuning information is determined in a targeted manner based on historical data and environmental information, and the accuracy of determining parameter tuning information is improved. Thereby, the purpose of determining the parameter tuning information of the target application based on a specific search space is achieved, and the technical effect of effectively improving the efficiency of determining parameter tuning information is achieved, thereby solving the technical problem of low efficiency in determining parameter tuning information when using the Bayesian algorithm to determine parameter tuning information in related technologies.

[0073] In order to obtain a first neural network model with better performance, in the method for determining parameter tuning information provided in Example 1 of the present application, the first neural network model is obtained in the following manner: obtaining data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; using the data information of historical parameter tuning as training samples and the application type of the application as the true label to construct a second training sample set; training the initial neural network model based on the second training sample set to obtain the first neural network model.

[0074] Optionally, at least one application includes a target application. There are multiple historical parameter tunings for an application in at least one application, that is, there is data information for multiple historical parameter tunings for an application in at least one application. For the same application, the operating environment of the application in different historical parameter tunings may be different. For example, the operating environment of the application in the first historical parameter tuning is 4 CPU cores and 16G memory capacity, and the operating environment in the second historical parameter tuning is 4 CPU cores and 8G memory capacity. The historical parameter tuning of at least one application can be implemented based on algorithms such as random search and Bayesian algorithm.

[0075] Optionally, the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning. The historical tuning evaluation results may be obtained by evaluating the application after historical tuning of the application based on the parameters and parameter values ​​used by the application during historical parameter tuning. The historical tuning evaluation results include multiple evaluation indicators and indicator values ​​corresponding to the evaluation indicators. The evaluation indicators include but are not limited to response time, throughput, throughput rate, concurrency, throughput bandwidth, etc. The tuning evaluation can be implemented through a benchmark program or other system detection tools.

[0076] After determining the historical parameter tuning data information of at least one application, the historical parameter tuning data information can be used as training samples, and the application type of the application can be used as a true label to construct a second training sample set. Different applications have different application types, and the application types of the applications include but are not limited to MySQL, Redis, etc.

[0077] After constructing the second training sample set, the parameter tuning system can train the initial neural network model based on the second training sample set to obtain the first neural network model. Optionally, the initial neural network model and the first neural network model can be LLM models or other neural network models, such as deep neural network models.

[0078] In an optional embodiment, it is possible to use Figure 4 The schematic diagram shown implements the training of the first neural network model. Figure 4 The process of determining the parameter tuning information according to the first embodiment of the present application is as follows Figure 2 ,like Figure 4As shown, a second training sample set is constructed based on data information of historical parameter tuning of at least one application, and the second training sample set and the full adjustable parameter library are used as inputs of the initial neural network model to train the initial neural network model to obtain a first neural network model. Among them, the full adjustable parameter library contains multiple levels and domain parameters of the kernel, kernel driver, compiler, runtime, and application, and the number of parameters can be in the range of 2000-100000. By using the full adjustable parameter library as the input of the initial neural network model, the first neural network model can learn information about all parameters.

[0079] Optionally, in the process of training the first neural network model based on the second training sample set, the first neural network model can learn the parameters required for application parameter tuning of different application types and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning and the true labels. Among them, the true labels are used for the first neural network model to learn the relevant knowledge corresponding to different application types (i.e., the parameters required for the above-mentioned application parameter tuning and the parameter value ranges of the application parameters in different operating environments).

[0080] Optionally, after training the first neural network model, the user may send an application tuning instruction to the parameter tuning system so that the parameter tuning system determines the target application and further determines the environment information of the target application. Optionally, the application tuning instruction may be a command line in a preset format, such as "keentuneparam tune ${app_name}", where "app_name" represents the name of the target application, for example, Figure 4 The "keentuneparam tune ${MySQL}" in the command is an application tuning instruction, in which the target application is MySQL. Optionally, the application tuning instruction can also use the Prompt operation to facilitate the specification of the output results and format of the parameter tuning system. The Prompt operation is used to guide the model to generate a response text fragment.

[0081] It should be noted that by setting the data information of historical parameter tuning to include the parameters, parameter values ​​and historical tuning evaluation results used in the historical parameter tuning, the richness of the data information of historical parameter tuning is improved. By constructing the second training sample set in the above manner and training the initial neural network model based on the second training sample set, on the one hand, the trained first neural network model can learn the information required for application parameter tuning based on the rich historical data, and on the other hand, the trained first neural network model can learn the information required for application parameter tuning of different application types based on the real labels, thereby improving the accuracy of determining the parameter tuning information.

[0082] How to determine the multiple parameters required for target application parameter tuning and the parameter value ranges corresponding to the parameters is crucial. Therefore, in the method for determining parameter tuning information provided in Example 1 of the present application, at least one application includes a target application, and the first neural network model learns the parameters required for application parameter tuning and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning. The environmental information is processed by the first neural network model to obtain the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning, including: determining the multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning learned by the first neural network model; determining the target operating environment from the operating environment learned by the first neural network model based on the environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; determining the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0083] Optionally, the first neural network model can learn the parameters required for application parameter tuning based on the data information of the historical parameter tuning of the application. Therefore, after determining the target application, the first neural network model can determine the multiple parameters corresponding to the target application from the parameters required for application parameter tuning that it has learned, thereby determining the multiple parameters corresponding to the target application as the multiple parameters required for target application parameter tuning.

[0084] In an optional embodiment, for at least one application in the application, the number of parameters required for parameter tuning of the application learned by the first neural network model is less than the preset number of parameters. For example, the preset number of parameters is 20. Assuming that 100 parameters appear in the data information of the historical parameter tuning of the application, the first neural network model selects 20 parameters from the 100 parameters (i.e., the parameters appearing in the data information of the historical parameter tuning of the application) based on the correlation between the learned parameters and the historical tuning evaluation results, and determines that the 20 parameters are the learned parameters required for parameter tuning of the application, thereby achieving a reduction in the number of parameters and avoiding the problem of redundant parameters and excessive computational complexity when all the parameters that have appeared are used as the parameters required for tuning the application.

[0085] Optionally, in the process of training the initial neural network model based on the second training sample set to obtain the first neural network model, a first prompt statement may be input into the initial neural network model, wherein the first prompt statement is used to guide the initial neural network model to learn parameters less than the preset number of parameters from the data information of the historical parameter tuning of the application. Optionally, a first code may also be built into the initial neural network model, and the first code is used to guide the initial neural network model to learn parameters less than the preset number of parameters from the data information of the historical parameter tuning of the application.

[0086] Since there is data information of multiple historical parameter tunings for an application in at least one application, and for the same application, the operating environment of the application in different historical parameter tunings may be different, therefore, for the application in at least one application, the first neural network model can also learn the parameter value range of the application under different operating environments based on the data information of historical parameter tuning.

[0087] In an optional embodiment, for at least one application in the application, the parameter value range of the application's parameters learned by the first neural network model in different operating environments is smaller than the target value range, and the target value range refers to the parameter value range of the application's parameter in the data information of the historical parameter tuning of the application. For example, assuming that the parameter value of a certain parameter in the data information of the historical parameter tuning of the application is in the interval [10, 20], then the target value range is [10, 20], and the parameter value range learned by the first neural network model may be [10, 12], or [15, 20], or other value ranges smaller than the target value range. In this way, the parameter value range can be reduced, avoiding the problem of excessive search space and resulting in excessive calculation when the target value range is used as the parameter value range of the parameters required for tuning the application.

[0088] Optionally, in the process of training the initial neural network model based on the second training sample set to obtain the first neural network model, a second prompt statement can be input into the initial neural network model, wherein the second prompt statement is used to guide the initial neural network model to learn from the data information of the historical parameter tuning of the application to obtain a parameter value range that is smaller than the target value range. Optionally, a second code can also be built into the initial neural network model, and the second code is used to guide the initial neural network model to learn from the data information of the historical parameter tuning of the application to obtain a parameter value range that is smaller than the target value range.

[0089] After acquiring the environmental information, the first neural network model can determine an operating environment with a high degree of similarity to the current operating environment of the target application from the operating environments that the first neural network model has learned, and obtain the target operating environment. Optionally, the first neural network model can calculate the similarity between the current operating environment of the target application and the learned operating environment based on the environmental information of the current operating environment and the environmental information of the learned operating environment based on a preset similarity calculation rule. The target operating environment can be one or more.

[0090] For example, assuming that the environmental information includes the number of CPU cores and the memory capacity, the first neural network model can determine the similarity of the two environmental information based on the difference between the number of CPU cores and the difference between the memory capacity in the two environmental information to be calculated for similarity (for example, the smaller the difference, the higher the similarity). For example, assuming that the environmental information of the target application is 4 CPU cores and 6G memory capacity, the environmental information of the first operating environment that has been learned is 4 CPU cores and 16G memory capacity, and the environmental information of the second operating environment that has been learned is 8 CPU cores and 128G memory capacity, then the similarity between the first environmental information and the environmental information of the target application is higher. For another example, the first neural network model can also perform vector conversion on the environmental information, and determine the similarity of the two environmental information by calculating the similarity of the vectors corresponding to the two environmental information to be calculated for similarity. The aforementioned preset threshold value can be a preset value, or it can be determined after determining the similarity between the multiple operating environments that the first neural network model has learned and the target operating environment, sorting the multiple operating environments that the first neural network model has learned based on the similarity, and determining it according to the similarity corresponding to the Nth operating environment in the sorting, where N is a positive integer.

[0091] After determining the target operating environment, the first neural network model can determine the parameter value ranges of multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of multiple parameters required for target application parameter tuning in the target operating environment.

[0092] Optionally, the parameter value ranges of multiple parameters required for tuning the target application parameters in the target operating environment may be directly determined as the parameter value ranges corresponding to the multiple parameters of the target application in the current operating environment.

[0093] Optionally, it is also possible to determine whether there is a target operating environment that is the same as the current operating environment, so as to determine the parameter value ranges corresponding to the multiple parameters required for tuning the target application parameters in the current operating environment based on the judgment result. For example, in a case where the judgment result indicates that there is a target operating environment that is the same as the current operating environment, the parameter value ranges of the multiple parameters of the target application in the target operating environment are directly determined as the parameter value ranges corresponding to the multiple parameters of the target application in the current operating environment. In a case where the judgment result indicates that there is no target operating environment that is the same as the current operating environment, optionally, the first neural network model can adjust the parameter value ranges of the multiple parameters of the target application parameters in the target operating environment based on the association between the learned parameter value ranges and the operating environment, and obtain the parameter value ranges corresponding to the multiple parameters of the target application in the current operating environment.

[0094] In an optional embodiment, it is possible to use Figure 4 The schematic diagram shown in FIG. 1 realizes the output of multiple parameters and the parameter value ranges corresponding to the parameters required for parameter tuning of the target application. Figure 4 As shown, the first neural network model can output the first file (i.e. Figure 4 json in the ), the first file includes multiple parameters required for target application parameter tuning and the corresponding parameter value ranges of the multiple parameters in the current running environment.

[0095] It should be noted that in the above process, by determining multiple parameters required for the target application from the parameters learned from the first neural network model, it is possible to accurately determine the multiple parameters required for the target application based on the parameter information in the historical tuning, and by determining the corresponding parameter value range in the current operating environment based on the parameter value range in the target operating environment similar to the current operating environment, it is possible to specifically determine the parameter value range based on the operating environment, thereby effectively improving the accuracy of the determined parameter value range.

[0096] In order to accurately determine the parameter tuning information of the target application, after determining the target application, the first neural network model is trained a second time to determine the parameter tuning information of the target application based on the second neural network model obtained by the secondary training. In the method for determining parameter tuning information provided in Example 1 of the present application, determining the parameter tuning information of the target application according to the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application includes: generating multiple groups of parameter combinations according to the multiple parameters required for tuning the target application and the parameter value ranges corresponding to the parameters; determining tuning data information according to the parameter combinations in the multiple groups of parameter combinations, and constructing a first training sample set according to the tuning data information corresponding to the multiple groups of parameter combinations; training the first neural network model based on the first training sample set to obtain the second neural network model, and determining the parameter tuning information through the second neural network model.

[0097] Optionally, the parameter tuning system can generate multiple sets of parameter combinations based on multiple parameters required for target application tuning and parameter value ranges corresponding to the parameters. For example, for a parameter among multiple parameters, any value is selected from the parameter value range corresponding to the parameter as the parameter value of the parameter, thereby generating a set of parameter combinations based on multiple parameters and their corresponding parameter values. Multiple sets of parameter combinations are obtained by random multiple times.

[0098] After determining multiple groups of parameter combinations, the parameter tuning system can determine the tuning data information based on the parameter combinations, and the tuning data information includes the parameter combinations and the tuning evaluation results corresponding to the parameter combinations. For example, for a parameter combination in a plurality of parameter combinations, the parameter tuning system can perform parameter tuning on the target application according to the parameter combination under the current operating environment of the target application, and perform tuning evaluation on the target application after parameter tuning to obtain the tuning evaluation results corresponding to the parameter combination, thereby determining the tuning data information based on the parameter combination and the tuning evaluation results corresponding to the parameter combination. For another example, for a parameter combination in a plurality of parameter combinations, the parameter tuning system can input the parameter combination into a target prediction model, predict the predicted tuning evaluation results corresponding to the parameter combination through the target prediction model, thereby determining the tuning data information based on the parameter combination and the predicted tuning evaluation results corresponding to the parameter combination.

[0099] Afterwards, the parameter tuning system can train the first neural network model based on the first training sample set to obtain the second neural network model, and determine the parameter tuning information through the second neural network model. Optionally, the second neural network model automatically outputs the parameter tuning information when the training is completed, and the parameter tuning system includes the second neural network model. For example, when training the first neural network model based on the first training sample set, the parameter tuning system can input a third prompt statement to the first neural network model to guide the first neural network model to output the parameter tuning information when the second neural network model is trained through the third prompt statement. In some embodiments, the initial neural network model can be a pre-trained LLM model, and the process of training the first neural network model and the second neural network model is equivalent to a fine-tuning process.

[0100] It should be noted that, by determining the first training sample set based on multiple parameters required for target application parameter tuning and the parameter value ranges corresponding to the parameters, it is possible to construct training samples based on parameter combinations in the search space determined by the first neural network model, and obtain the second neural network model through training the first training sample set to determine parameter tuning information, so that the second neural network model can learn multiple parameter combinations in the search space and optimize the parameter information, thereby improving the efficiency and accuracy of determining the parameter tuning information.

[0101] In order to effectively construct the first training sample set, in the method for determining parameter tuning information provided in Example 1 of the present application, tuning data information is determined according to a parameter combination in a plurality of parameter combinations, and the first training sample set is constructed according to the tuning data information corresponding to the plurality of parameter combinations, including: for the parameter combinations in the plurality of parameter combinations, performing parameter tuning on the target application based on the parameter combination, and performing tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; combining the parameter combination and the tuning evaluation results corresponding to the parameter combination to obtain the tuning data information; and using the tuning data information corresponding to the plurality of parameter combinations as training samples, and the application type of the target application as a true label to construct the first training sample set.

[0102] Optionally, for a parameter combination in the plurality of parameter combinations, the parameter tuning system may perform parameter tuning on the target application under the current operating environment of the target application based on the parameter combination, thereby performing tuning evaluation on the target application after parameter tuning, and obtaining a tuning evaluation result corresponding to the parameter combination. The parameter tuning system may use a benchmark program or other methods to perform tuning evaluation on the target application after parameter tuning.

[0103] Afterwards, the parameter tuning system can combine the parameter combinations and the tuning evaluation results corresponding to the parameter combinations to obtain tuning data information, thereby using the tuning data information corresponding to multiple groups of parameter combinations as training samples and the application type of the target application as the true label to construct the first training sample set.

[0104] In an optional embodiment, it is possible to use Figure 4 The schematic diagram shown in the figure realizes the secondary training of the first neural network model. Figure 4 In the example, the target application is MySQL. Therefore, the parameter tuning system can perform parameter tuning on MySQL based on the parameter combination in the current running environment of MySQL to determine the real-time tuning data information, and train the first neural network model through the obtained tuning data information to obtain the second neural network model.

[0105] Optionally, during the training process, the second neural network model can determine the tuning effect score corresponding to the parameter combination according to the parameter combination and the corresponding tuning evaluation result, thereby determining the better parameter value corresponding to the parameter according to the tuning effect score to obtain parameter tuning information. Figure 4 As shown, the second neural network model outputs parameter tuning information after training is completed, and the parameter tuning information can be fed back to the user who issued the parameter tuning instruction through the user interface. According to different actual needs, the parameter tuning information output by the second neural network model can be one group or multiple groups for user selection.

[0106] It should be noted that, by performing parameter tuning on the target application based on the parameter combination, and constructing a first training sample set based on the tuning evaluation results of the parameter tuning and the parameter combination to train the second neural network model, it is possible to train the second neural network model based on the tuning data information in the actual operating environment, so that the second neural network model can give better tuning suggestions for the current operating environment of the target application, thereby improving the accuracy of the parameter tuning information.

[0107] In order to determine the tuning evaluation result corresponding to the parameter combination, in the method for determining parameter tuning information provided in Example 1 of the present application, a tuning evaluation is performed on the target application after parameter tuning to obtain the tuning evaluation result corresponding to the parameter combination, including: determining multiple evaluation indicators required for tuning and evaluating the target application to obtain multiple target evaluation indicators; and performing tuning and evaluation on the target application after parameter tuning based on the multiple target evaluation indicators to obtain the tuning evaluation result corresponding to the parameter combination.

[0108] Optionally, the parameter tuning system may obtain a plurality of preset evaluation indicators, thereby determining the plurality of preset evaluation indicators as target evaluation indicators.

[0109] Optionally, the data information of historical parameter tuning may include historical tuning evaluation results of the application. The first neural network model can learn, based on the historical tuning evaluation results, multiple evaluation indicators used in historical tuning evaluation of at least one application, thereby determining multiple target evaluation indicators based on the learned multiple evaluation indicators used in historical tuning evaluation.

[0110] Optionally, the parameter tuning system can also determine multiple target evaluation indicators based on multiple evaluation indicators learned through the first neural network model and in combination with the user's personalized update instructions for the evaluation indicators.

[0111] After determining multiple target evaluation indicators, the parameter tuning system can evaluate the target application after parameter tuning based on the multiple target evaluation indicators to obtain the tuning evaluation results corresponding to the parameter combination. For example, the target application is evaluated by using a benchmark program or other methods to obtain an initial tuning evaluation result, and then the evaluation information corresponding to the target evaluation indicator is filtered out from the initial tuning evaluation result to obtain the tuning evaluation result. For another example, the target evaluation indicator of the target application is evaluated by using a benchmark program or other methods to obtain the tuning evaluation result.

[0112] It should be noted that, by determining the tuning evaluation results based on the target evaluation indicators, it is achieved that the tuning evaluation results are determined only based on the evaluation indicators that are concerned during parameter tuning. Therefore, when the second neural network model is trained based on the first training sample set generated based on the tuning evaluation results, the second neural network model can determine multiple parameters with better parameter values ​​based on the evaluation indicators of concern, thereby improving the flexibility and accuracy of the determined parameter tuning information.

[0113] In order to effectively determine multiple target evaluation indicators, in the method for determining parameter tuning information provided in Example 1 of the present application, the data information of historical parameter tuning includes historical tuning evaluation results of the application, and the first neural network model learns multiple evaluation indicators used in historical tuning evaluation of at least one application based on the historical tuning evaluation results, and determines multiple evaluation indicators required for tuning evaluation of the target application. Obtaining multiple target evaluation indicators includes: determining multiple target evaluation indicators based on the multiple evaluation indicators used in the historical tuning evaluation that have been learned by the first neural network model.

[0114] Optionally, the data information of historical parameter tuning includes historical tuning evaluation results of the application. Therefore, for at least one application in the application, the first neural network model can determine the evaluation indicators appearing in the historical tuning evaluation results corresponding to the application as the learned evaluation indicators of the application.

[0115] In the process of determining multiple target evaluation indicators based on multiple evaluation indicators used in historical tuning evaluations learned by the first neural network model, optionally, multiple evaluation indicators used in historical tuning evaluations of target applications learned by the first neural network model can be determined as target evaluation indicators. Optionally, evaluation indicators with a frequency of occurrence greater than a preset frequency can be determined from multiple evaluation indicators used in historical tuning evaluations of target applications learned by the first neural network model, and the evaluation indicators with a frequency of occurrence greater than the preset frequency can be determined as target evaluation indicators. Optionally, multiple target evaluation indicators can also be determined in combination with the user's personalized update instructions for evaluation indicators.

[0116] It should be noted that by determining multiple target evaluation indicators based on the first neural network model, the efficiency and accuracy of determining the target evaluation indicators are improved, thereby improving the efficiency and accuracy of determining parameter tuning information.

[0117] In order to more flexibly determine multiple target evaluation indicators, in the method for determining parameter tuning information provided in Example 1 of the present application, determining multiple target evaluation indicators based on multiple evaluation indicators used in historical tuning evaluations that have been learned by the first neural network model includes: determining multiple evaluation indicators used in historical tuning evaluations of target applications that have been learned by the first neural network model as multiple initial evaluation indicators; when receiving an indicator update instruction from the user, extracting indicator update information from the indicator update instruction, and determining multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; when not receiving an indicator update instruction from the user, determining the multiple initial evaluation indicators as multiple target evaluation indicators.

[0118] Optionally, multiple evaluation indicators that have been learned by the first neural network model for use in historical tuning and evaluation of the target application can be determined as initial evaluation indicators. Optionally, evaluation indicators that appear more than a preset number of times can be determined from the multiple evaluation indicators that have been learned by the first neural network model for use in historical tuning and evaluation of the target application, and the evaluation indicators that appear more than a preset number of times can be determined as initial evaluation indicators.

[0119] Optionally, the user may send a personalized update instruction (i.e., an indicator update instruction) of the evaluation indicator to the first neural network model to personalize the target evaluation indicator. The indicator update instruction includes indicator update information, which is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators. The indicator update information may include indicator information of the evaluation indicators to be added and / or indicator information of the evaluation indicators to be deleted. In the case of receiving the user's indicator update instruction, if the indicator update information includes the indicator information of the evaluation indicators to be added, the parameter tuning system may determine the evaluation indicators to be added and multiple initial evaluation indicators as target evaluation indicators. If the indicator update information includes the indicator information of the evaluation indicators to be deleted, the parameter tuning system may determine the initial evaluation indicators other than the evaluation indicators to be deleted from the multiple initial evaluation indicators as target evaluation indicators. In the case of not receiving the user's indicator update instruction, the multiple initial evaluation indicators are directly determined as multiple target evaluation indicators.

[0120] Optionally, the above-mentioned indicator information at least includes an indicator identifier. When the indicator update instruction includes indicator update information for adding a new evaluation indicator, if the evaluation indicator to be added is an evaluation indicator unknown to the parameter tuning system, the indicator information of the evaluation indicator to be added also includes a method for obtaining the indicator value of the evaluation indicator to be added. For example, the acquisition method may be "obtaining the indicator value of the evaluation indicator from the sixth line of the initial tuning evaluation result output by the benchmark program".

[0121] Optionally, the parameter tuning system may classify multiple initial evaluation indicators into two categories: throughput and latency, and display the initial evaluation indicators under the two categories to the user through a human-computer interaction interface based on the categories, so that the user can more conveniently issue indicator update instructions.

[0122] It should be noted that in actual applications, different users pay attention to different evaluation indicators when tuning parameters of the same application. Therefore, by combining the evaluation indicators learned by the first network model and the user's personalized indicator update instructions to determine the target evaluation indicators, the flexibility of determining the target evaluation indicators in this application is improved, thereby improving the flexibility of determining parameter tuning information.

[0123] In order to more accurately determine the parameter tuning information, in the method for determining the parameter tuning information provided in Example 1 of the present application, a first neural network model is trained based on a first training sample set to obtain a second neural network model, and the parameter tuning information is determined through the second neural network model, including: determining the weights corresponding to the target evaluation indicators among multiple target evaluation indicators; training the first neural network model based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain the second neural network model, and determining the parameter tuning information through the second neural network model.

[0124] Optionally, for a target evaluation indicator among the multiple target evaluation indicators, the parameter tuning system may obtain a preset weight corresponding to the target evaluation indicator, thereby determining the preset weight as the weight corresponding to the target evaluation indicator.

[0125] Optionally, the data information of historical parameter tuning may include historical tuning evaluation results of the application. The first neural network model may learn the weights of the evaluation indicators according to the frequency of occurrence of the evaluation indicators in the historical tuning evaluation results, thereby determining the weights of multiple target evaluation indicators according to the learned weights of the evaluation indicators.

[0126] Optionally, the parameter tuning system can also determine the weight of the target evaluation indicator based on the weight of the evaluation indicator learned through the first neural network model and in combination with the user's personalized update instruction for the weight of the evaluation indicator.

[0127] After determining the weights of the multiple target evaluation indicators, the parameter tuning system can train the first neural network model based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain the second neural network model. For example, the first training sample set and the weights corresponding to the multiple target evaluation indicators are used as inputs of the first neural network model to train the first neural network model.

[0128] Optionally, in the process of training the first neural network model based on the first training sample set and the weights corresponding to multiple target evaluation indicators to obtain the second neural network model, for the parameter combination in the first training sample set, the second neural network model can determine the tuning effect score corresponding to the parameter combination according to the weights corresponding to the multiple target evaluation indicators and the tuning evaluation results corresponding to the parameter combination, thereby determining multiple relatively optimal parameter combinations from the first training sample set according to the tuning effect scores, and then determining the optimal values ​​of multiple parameters according to the spatial distribution of the parameter values ​​of the relatively optimal multiple parameter combinations to obtain parameter tuning information.

[0129] For example, in the process of determining the tuning effect score corresponding to the parameter combination according to the weights corresponding to the multiple target evaluation indicators and the tuning evaluation results corresponding to the parameter combination, for the target evaluation indicators among the multiple target evaluation indicators, the parameter tuning system may preset a scoring rule corresponding to the target evaluation indicator, so that the second neural network model determines the score of the target evaluation indicator according to the scoring rule, and then determines the tuning effect score according to the scores of the multiple target evaluation indicators and the weights corresponding to the multiple target evaluation indicators. For example, assuming that the target evaluation indicator is the response time, the shorter the response time, the higher the score of the target evaluation indicator.

[0130] It should be noted that by training the first neural network model in combination with the weights corresponding to the target evaluation indicators, the second neural network model is obtained, so that the second neural network model can focus more on learning relatively important target evaluation indicators during the training process, thereby further improving the accuracy of the determined parameter tuning information.

[0131] In order to effectively determine the weight corresponding to the target evaluation indicator, in the method for determining parameter tuning information provided in Example 1 of the present application, the data information of historical parameter tuning includes the historical tuning evaluation results of the application, and the first neural network model learns the weight of the evaluation indicator according to the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results. Determining the weight corresponding to the target evaluation indicator among multiple target evaluation indicators includes: determining the weights of multiple target evaluation indicators according to the weights of the evaluation indicators already learned by the first neural network model.

[0132] Optionally, the data information of historical parameter tuning includes historical tuning evaluation results of the application. Therefore, for at least one application in the application, the first neural network model can learn the weight of the evaluation indicator of the application based on the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results corresponding to the application.

[0133] Optionally, the weights of the same evaluation index corresponding to different applications learned by the first neural network model may be different. For example, for an evaluation index corresponding to a certain application, the frequency of occurrence of the evaluation index is determined based on the ratio between the number of occurrences of the evaluation index in the historical tuning evaluation results corresponding to the application and the number of historical tuning evaluation results corresponding to the application. Afterwards, the frequencies of occurrence of multiple evaluation indexes corresponding to the application may be normalized, and then the normalized values ​​corresponding to the evaluation index may be determined as the weight of the evaluation index under the application.

[0134] Optionally, the weights of the same evaluation indicators corresponding to different applications learned by the first neural network model may be the same. For example, for a certain evaluation indicator, the frequency of occurrence of the evaluation indicator is determined based on the ratio between the number of occurrences of the evaluation indicator in the historical tuning evaluation results corresponding to at least one application and the number of historical tuning evaluation results corresponding to at least one application. Afterwards, the frequencies of occurrence of multiple evaluation indicators corresponding to at least one application may be normalized, and then the normalized values ​​corresponding to the evaluation indicators may be determined as the weights of the evaluation indicators.

[0135] In the process of determining the weights of multiple target evaluation indicators based on the weights of the evaluation indicators learned by the first neural network model, optionally, the weights corresponding to the multiple target evaluation indicators learned by the first neural network model can be determined as the weights of the multiple target evaluation indicators.

[0136] Optionally, the weights of multiple target evaluation indicators can also be determined in combination with the user's personalized update instructions for weights. For example, the weights that the first neural network model has learned for multiple target evaluation indicators are determined as the initial weights of the multiple target evaluation indicators. When the user's weight update instruction is received, the weight update information is extracted from the weight update instruction, and the weights of the multiple target evaluation indicators are determined based on the weight update information and the initial weights of the multiple target evaluation indicators. Among them, the weight update information includes the target evaluation indicator to be updated and the weight value required for the target evaluation indicator. In the case where the user's weight update instruction is not received, the initial weights of the multiple target evaluation indicators are determined as the weights of the multiple target evaluation indicators.

[0137] In an optional embodiment, if Figure 4 As shown, the first neural network model can output the second file (i.e. Figure 4 json in the parameter tuning system), and the second file includes multiple evaluation indicators required for tuning and evaluating the target application and the weights of the target evaluation indicators. The second file may also include a method for obtaining the indicator value of the target evaluation indicator. Optionally, the correspondence between the indicator value of the evaluation indicator and the acquisition method may be preset in the parameter tuning system. For example, the acquisition method may be "obtaining the indicator value of throughput (i.e., the target evaluation indicator) from the 11th line in the initial tuning evaluation result output by the benchmark program."

[0138] It should be noted that by determining the weights of multiple target evaluation indicators according to the first neural network model, the efficiency and accuracy of determining the weights of the target evaluation indicators are improved, thereby improving the efficiency and accuracy of determining parameter tuning information.

[0139] In an embodiment of the present application, by determining the environmental information of the current operating environment of the target application, wherein the target application is an application to be parameter tuned; processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; determining the parameter tuning information of the target application according to the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, the first neural network model is trained based on data information of historical parameter tuning of at least one application, so that the first neural network model can learn the information required for parameter tuning of different applications from historical data. By processing the environmental information through the first neural network model to obtain the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, on the one hand, the search space for parameter tuning information is narrowed, and the efficiency of determining parameter tuning information is improved; on the other hand, the search space for parameter tuning information is determined in a targeted manner based on historical data and environmental information, and the accuracy of determining parameter tuning information is improved. Thereby, the purpose of determining the parameter tuning information of the target application based on a specific search space is achieved, and the technical effect of effectively improving the efficiency of determining the parameter tuning information is realized, thereby solving the technical problem of low efficiency in determining the parameter tuning information by using the Bayesian algorithm in the related technology.

[0140] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0141] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0142] Example 2

[0143] According to an embodiment of the present application, a method for determining parameter tuning information is also provided. Figure 5 is a flow chart of a method for determining parameter tuning information according to the second embodiment of the present application. Figure 5 As shown, the method includes:

[0144] Step S501, obtaining the environment information of the current running environment of the target application uploaded by the client, wherein the target application is the application to be parameter tuned.

[0145] Step S502: Process the environmental information through the first neural network model in the cloud server to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, and determine the parameter tuning information of the target application based on the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application.

[0146] Step S503: Feedback the parameter tuning information of the target application to the client.

[0147] Through the above scheme, the first neural network model is obtained by training according to the data information of the historical parameter tuning of at least one application, so that the first neural network model can learn the information required for the parameter tuning of different applications from the historical data. By processing the environmental information by the first neural network model, the parameter value ranges corresponding to the multiple parameters required for the parameter tuning of the target application are obtained. On the one hand, the search space of the parameter tuning information is narrowed, and the efficiency of determining the parameter tuning information is improved. On the other hand, the search space of the parameter tuning information is determined in a targeted manner based on the historical data and the environmental information, and the accuracy of determining the parameter tuning information is improved. Thereby, the purpose of determining the parameter tuning information of the target application based on a specific search space is achieved, and the technical effect of effectively improving the efficiency of determining the parameter tuning information is achieved, thereby solving the technical problem of low efficiency in determining the parameter tuning information by using the Bayesian algorithm in the related technology.

[0148] In the cloud server, the specific method for determining the parameter tuning information is the same as the method in the first embodiment, and will not be repeated here.

[0149] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0150] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of each embodiment of the present application.

[0151] Example 3

[0152] According to an embodiment of the present application, a device for determining parameter tuning information for implementing the above-mentioned method for determining parameter tuning information is also provided. Figure 6 is a schematic diagram of a device for determining parameter tuning information according to Embodiment 3 of the present application, such as Figure 6 As shown, the device includes: a first determining unit 601, a processing unit 602, and a second determining unit 603.

[0153] A first determining unit 601 is used to determine the environment information of the current running environment of the target application, wherein the target application is the application to be parameter tuned;

[0154] A processing unit 602 is used to process the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of a target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application;

[0155] The second determining unit 603 is configured to determine parameter tuning information of the target application according to parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application.

[0156] In the device for determining parameter tuning information provided in the third embodiment of the present application, the environment information of the current running environment of the target application is determined by the first determination unit 601, wherein the target application is the application to be parameter tuned; the first processing unit 602 processes the environment information through the first neural network model to obtain the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on the data information of the historical parameter tuning of at least one application; the second determination unit 603 determines the parameter tuning information of the target application based on the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application. In this solution, the first neural network model is trained based on the data information of the historical parameter tuning of at least one application, so that the first neural network model can learn the information required for parameter tuning of different applications from the historical data. By processing the environment information by the first neural network model, the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application are obtained. On the one hand, the search space for parameter tuning information is narrowed, and the efficiency of determining parameter tuning information is improved. On the other hand, the search space for parameter tuning information is determined in a targeted manner based on historical data and environment information, and the accuracy of determining parameter tuning information is improved. Thereby, the purpose of determining the parameter tuning information of the target application based on a specific search space is achieved, and the technical effect of effectively improving the efficiency of determining the parameter tuning information is realized, thereby solving the technical problem of low efficiency in determining the parameter tuning information by using the Bayesian algorithm in the related technology.

[0157] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, at least one application includes a target application, and the first neural network model learns the parameters required for application parameter tuning and the parameter value ranges of the application parameters in different operating environments based on the data information of historical parameter tuning, and the processing unit includes: a first determination subunit, used to determine multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning that have been learned by the first neural network model; a second determination subunit, used to determine the target operating environment from the operating environment that has been learned by the first neural network model based on the environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; and a third determination subunit, used to determine the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0158] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the second determination unit includes: a generation subunit, used to generate multiple groups of parameter combinations according to multiple parameters required for target application tuning and parameter value ranges corresponding to the parameters; a construction subunit, used to determine tuning data information based on parameter combinations in the multiple groups of parameter combinations, and construct a first training sample set based on the tuning data information corresponding to the multiple groups of parameter combinations; a fourth determination subunit, used to train the first neural network model based on the first training sample set to obtain a second neural network model, and determine the parameter tuning information through the second neural network model.

[0159] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, a construction subunit includes: a first processing module, used to perform parameter tuning on a target application based on a parameter combination in a plurality of parameter combinations, and to perform tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; a second processing module, used to combine the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain tuning data information; and a construction module, used to use the tuning data information corresponding to the plurality of parameter combinations as training samples and the application type of the target application as a true label to construct a first training sample set.

[0160] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the second processing module includes: a first determination sub-module, used to determine multiple evaluation indicators required for tuning and evaluating the target application, and obtain multiple target evaluation indicators; a first processing sub-module, used to perform tuning and evaluation on the target application after parameter tuning based on multiple target evaluation indicators, and obtain a tuning evaluation result corresponding to the parameter combination.

[0161] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the data information of historical parameter tuning includes historical tuning evaluation results of the application, and the first neural network model learns, based on the historical tuning evaluation results, multiple evaluation indicators used in historical tuning evaluation of at least one application, and the first determination submodule includes: a second determination submodule, which is used to determine multiple target evaluation indicators based on the multiple evaluation indicators used in the historical tuning evaluation that have been learned by the first neural network model.

[0162] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the second determination submodule includes: a third determination submodule, used to determine multiple evaluation indicators used by the first neural network model to perform historical tuning evaluation on the target application as multiple initial evaluation indicators; a fourth determination submodule, used to extract indicator update information from the indicator update instruction when receiving the user's indicator update instruction, and determine multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; a fifth determination submodule, used to determine multiple initial evaluation indicators as multiple target evaluation indicators when no indicator update instruction is received from the user.

[0163] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the fourth determination subunit includes: a first determination module, used to determine the weights corresponding to target evaluation indicators among multiple target evaluation indicators; a second determination module, used to train a first neural network model based on a first training sample set and the weights corresponding to multiple target evaluation indicators to obtain a second neural network model, and determine the parameter tuning information through the second neural network model.

[0164] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the data information of historical parameter tuning includes historical tuning evaluation results of the application, the first neural network model learns the weight of the evaluation indicator according to the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results, and the first determination module includes: a sixth determination submodule, which is used to determine the weights of multiple target evaluation indicators based on the weights of the evaluation indicators learned by the first neural network model.

[0165] Optionally, in the device for determining parameter tuning information provided in Example 3 of the present application, the device for determining parameter tuning information also includes: an acquisition unit, used to acquire data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; a construction unit, used to construct a second training sample set using the data information of historical parameter tuning as a training sample and the application type of the application as a true label; and a training unit, used to train the initial neural network model based on the second training sample set to obtain a first neural network model.

[0166] It should be noted that the first determination unit 601, the processing unit 602, and the second determination unit 603 described above correspond to steps S301 to S303 in Embodiment 1, and the examples and application scenarios implemented by the above units and the corresponding steps are the same, but are not limited to the contents disclosed in the above Embodiment 1. It should be noted that the above modules, as part of the device, can be run in the computer terminal 10 provided in Embodiment 1.

[0167] It should be noted that the preferred implementation scheme involved in the above embodiments of the present application is the same as the scheme provided in Example 1, as well as the application scenario and implementation process, but is not limited to the scheme provided in Example 1.

[0168] Example 4

[0169] The embodiment of the present application may provide an electronic device, which may be any electronic device in a group of electronic devices. Optionally, in this embodiment, the electronic device may also be replaced by a terminal device such as a mobile terminal.

[0170] Optionally, in this embodiment, the electronic device may be located in at least one network device among a plurality of network devices of a computer network.

[0171] In this embodiment, the electronic device can execute the program code of the following steps in the method for determining parameter tuning information: determining environmental information of the current operating environment of the target application, wherein the target application is an application to be parameter tuned; processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; determining the parameter tuning information of the target application based on the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application.

[0172] The above-mentioned electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning that have been learned by the first neural network model; determining the target operating environment from the operating environment that has been learned by the first neural network model based on environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; determining the parameter value ranges corresponding to the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0173] The above-mentioned electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: generate multiple groups of parameter combinations according to multiple parameters required for target application tuning and parameter value ranges corresponding to the parameters; determine tuning data information according to the parameter combinations in the multiple groups of parameter combinations, and construct a first training sample set according to the tuning data information corresponding to the multiple groups of parameter combinations; train the first neural network model based on the first training sample set to obtain a second neural network model, and determine the parameter tuning information through the second neural network model.

[0174] The electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: for a parameter combination in a plurality of parameter combinations, perform parameter tuning on a target application based on the parameter combination, and perform tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; combine the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain tuning data information; use the tuning data information corresponding to the plurality of parameter combinations as training samples, and use the application type of the target application as a true label to construct a first training sample set.

[0175] The above-mentioned electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining multiple evaluation indicators required for tuning and evaluating the target application to obtain multiple target evaluation indicators; performing tuning and evaluation on the target application after parameter tuning based on the multiple target evaluation indicators to obtain the tuning evaluation results corresponding to the parameter combination.

[0176] The electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining multiple target evaluation indicators based on multiple evaluation indicators used in historical tuning evaluations that have been learned by the first neural network model.

[0177] The above-mentioned electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining multiple evaluation indicators used in historical tuning evaluation of the target application that have been learned by the first neural network model as multiple initial evaluation indicators; when receiving an indicator update instruction from the user, extracting indicator update information from the indicator update instruction, and determining multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; when not receiving an indicator update instruction from the user, determining the multiple initial evaluation indicators as multiple target evaluation indicators.

[0178] The above-mentioned electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining the weights corresponding to the target evaluation indicators among multiple target evaluation indicators; training the first neural network model based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain the second neural network model, and determining the parameter tuning information through the second neural network model.

[0179] The electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: determining the weights of multiple target evaluation indicators based on the weights of the evaluation indicators learned by the first neural network model.

[0180] The electronic device can also execute the program code of the following steps in the method for determining parameter tuning information: obtaining data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; using the data information of historical parameter tuning as training samples and the application type of the application as a true label to construct a second training sample set; training the initial neural network model based on the second training sample set to obtain a first neural network model.

[0181] Optionally, Figure 7 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 7 As shown, the electronic device 20 may include: one or more ( Figure 7 Only one is shown in the figure) processor 202, memory 204. The electronic device 20 may also include a storage controller, through which the memory 204 is controlled and managed; the electronic device 20 may also include a peripheral interface, through which the radio frequency module, audio module and display screen are connected.

[0182] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the method and device for determining parameter tuning information in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the method for determining the parameter tuning information described above is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0183] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: determine the environmental information of the current operating environment of the target application, wherein the target application is the application to be parameter tuned; process the environmental information through the first neural network model to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; determine the parameter tuning information of the target application based on the parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application.

[0184] Optionally, the processor may also execute program codes of the following steps: determining multiple parameters required for target application parameter tuning from the parameters required for application parameter tuning that have been learned by the first neural network model; determining the target operating environment from the operating environment that has been learned by the first neural network model based on environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; determining the corresponding parameter value ranges of the multiple parameters required for target application parameter tuning in the current operating environment based on the parameter value ranges of the multiple parameters required for target application parameter tuning in the target operating environment.

[0185] Optionally, the processor may also execute program code for the following steps: generating multiple groups of parameter combinations based on multiple parameters required for target application tuning and parameter value ranges corresponding to the parameters; determining tuning data information based on parameter combinations in the multiple groups of parameter combinations, and constructing a first training sample set based on the tuning data information corresponding to the multiple groups of parameter combinations; training a first neural network model based on the first training sample set to obtain a second neural network model, and determining parameter tuning information through the second neural network model.

[0186] Optionally, the processor may also execute the program code of the following steps: for a parameter combination in a plurality of parameter combinations, perform parameter tuning on a target application based on the parameter combination, and perform tuning evaluation on the target application after parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; combine the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain tuning data information; use the tuning data information corresponding to the plurality of parameter combinations as training samples, and use the application type of the target application as a true label to construct a first training sample set.

[0187] Optionally, the processor may also execute the following program code: determining multiple evaluation indicators required for tuning and evaluating the target application to obtain multiple target evaluation indicators; performing tuning and evaluation on the target application after parameter tuning based on the multiple target evaluation indicators to obtain tuning evaluation results corresponding to the parameter combination.

[0188] Optionally, the processor may also execute the program code of the following steps: determining multiple target evaluation indicators based on multiple evaluation indicators used in historical tuning evaluation that have been learned by the first neural network model.

[0189] Optionally, the processor may also execute program code for the following steps: determining multiple evaluation indicators that have been learned by the first neural network model and used in historical tuning and evaluation of the target application as multiple initial evaluation indicators; extracting indicator update information from the indicator update instruction upon receiving an indicator update instruction from the user, and determining multiple target evaluation indicators based on the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate the addition of new evaluation indicators or the deletion of evaluation indicators; and determining the multiple initial evaluation indicators as multiple target evaluation indicators upon not receiving an indicator update instruction from the user.

[0190] Optionally, the processor may also execute the program code of the following steps: determining the weights corresponding to the target evaluation indicators among multiple target evaluation indicators; training the first neural network model based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain the second neural network model, and determining the parameter tuning information through the second neural network model.

[0191] Optionally, the processor may also execute the program code of the following steps: determining weights of multiple target evaluation indicators based on the weights of the evaluation indicators learned by the first neural network model.

[0192] Optionally, the processor may also execute the program code of the following steps: obtaining data information of historical parameter tuning of at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values ​​and historical tuning evaluation results used by the application during historical parameter tuning; using the data information of historical parameter tuning as training samples and the application type of the application as a true label to construct a second training sample set; training the initial neural network model based on the second training sample set to obtain a first neural network model.

[0193] It can be understood by those skilled in the art that Figure 7 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as , tablet computers, PDAs, mobile Internet devices (MID), PADs and other terminal devices. Figure 7 The structure of the electronic device is not limited. Figure 7 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 7 Different configurations are shown.

[0194] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0195] Example 5

[0196] The embodiment of the present application further provides a computer program product. Optionally, in this embodiment, the computer program product may include a computer program, and when the computer program is executed by a processor, the method for determining parameter tuning information provided in the first embodiment is implemented.

[0197] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0198] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0199] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0200] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0201] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0202] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0203] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining parameter tuning information, It is characterized in that include: Determining environment information of a current running environment of a target application, wherein the target application is an application to be parameter tuned; Processing the environmental information through a first neural network model to obtain parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; Determining parameter tuning information of the target application according to parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application; The at least one application includes the target application, the first neural network model learns the parameter value ranges of the application parameters in different operating environments according to the data information of the historical parameter tuning, and the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning are obtained by processing the environment information through the first neural network model, including: Determine a target operating environment from the operating environments learned by the first neural network model according to the environmental information, wherein a similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; According to the parameter value ranges of the multiple parameters required for the target application parameter tuning in the target operating environment, the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning in the current operating environment are determined.

2. The method according to claim 1, It is characterized in that The first neural network model learns the parameters required for the application parameter tuning according to the data information of the historical parameter tuning, and processes the environmental information through the first neural network model to obtain the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning, including: Determine a plurality of parameters required for tuning the target application parameters from the parameters required for tuning the application parameters learned by the first neural network model.

3. The method according to claim 1, It is characterized in that Determining the parameter tuning information of the target application according to the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application includes: Generate multiple groups of parameter combinations according to multiple parameters required for tuning the target application and parameter value ranges corresponding to the parameters; Determining tuning data information according to a parameter combination in the multiple sets of parameter combinations, and constructing a first training sample set according to the tuning data information corresponding to the multiple sets of parameter combinations; The first neural network model is trained based on the first training sample set to obtain a second neural network model, and the parameter tuning information is determined through the second neural network model.

4. The method according to claim 3, It is characterized in that Determining the tuning data information according to the parameter combinations in the multiple groups of parameter combinations, and constructing the first training sample set according to the tuning data information corresponding to the multiple groups of parameter combinations includes: For a parameter combination in the plurality of parameter combinations, performing parameter tuning on the target application based on the parameter combination, and performing tuning evaluation on the target application after the parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination; Combining the parameter combination and the tuning evaluation result corresponding to the parameter combination to obtain the tuning data information; The first training sample set is constructed by using the tuning data information corresponding to the multiple groups of parameter combinations as training samples and the application type of the target application as a true label.

5. The method according to claim 4, It is characterized in that The target application after parameter tuning is evaluated, and the evaluation results of the tuning corresponding to the parameter combination are obtained, including: Determine a plurality of evaluation indicators required for tuning and evaluating the target application, and obtain a plurality of target evaluation indicators; Based on the multiple target evaluation indicators, a tuning evaluation is performed on the target application after the parameter tuning to obtain a tuning evaluation result corresponding to the parameter combination.

6. The method according to claim 5, It is characterized in that The data information of historical parameter tuning includes historical tuning evaluation results of the application, and the first neural network model learns, based on the historical tuning evaluation results, a plurality of evaluation indicators used in historical tuning evaluation of the application in the at least one application, and determines a plurality of evaluation indicators required for tuning evaluation of the target application, and the obtained plurality of target evaluation indicators include: The multiple target evaluation indicators are determined according to the multiple evaluation indicators used in the historical tuning evaluation that the first neural network model has learned.

7. The method according to claim 6, It is characterized in that Determining the multiple target evaluation indicators according to the multiple evaluation indicators used in the historical tuning evaluation learned by the first neural network model includes: Determining multiple evaluation indicators used in historical tuning evaluation of the target application that have been learned by the first neural network model as multiple initial evaluation indicators; In the case of receiving an indicator update instruction from a user, extracting indicator update information from the indicator update instruction, and determining the multiple target evaluation indicators according to the indicator update information and the multiple initial evaluation indicators, wherein the indicator update information is used to indicate adding a new evaluation indicator or deleting an evaluation indicator; In a case where no indicator update instruction from the user is received, the multiple initial evaluation indicators are determined as the multiple target evaluation indicators.

8. The method according to claim 5, It is characterized in that Training the first neural network model based on the first training sample set to obtain a second neural network model, and determining the parameter tuning information through the second neural network model includes: Determining weights corresponding to target evaluation indicators among the multiple target evaluation indicators; The first neural network model is trained based on the first training sample set and the weights corresponding to the multiple target evaluation indicators to obtain a second neural network model, and the parameter tuning information is determined through the second neural network model.

9. The method according to claim 8, It is characterized in that The data information of the historical parameter tuning includes the historical tuning evaluation results of the application, the first neural network model learns the weight of the evaluation indicator according to the frequency of occurrence of the evaluation indicator in the historical tuning evaluation results, and determines the weight corresponding to the target evaluation indicator among the multiple target evaluation indicators, including: The weights of the multiple target evaluation indicators are determined according to the weights of the evaluation indicators learned by the first neural network model.

10. The method according to claim 1, It is characterized in that The first neural network model is obtained by: Acquire data information of historical parameter tuning of the at least one application, wherein the data information of historical parameter tuning includes parameters, parameter values, and historical tuning evaluation results used by the application during historical parameter tuning; The data information of the historical parameter tuning is used as a training sample, and the application type of the application is used as a true label to construct a second training sample set; An initial neural network model is trained based on the second training sample set to obtain the first neural network model.

11. A method for determining parameter tuning information, It is characterized in that include: Obtaining environment information of a current running environment of a target application uploaded by a client, wherein the target application is an application to be parameter-tuned; The environment information is processed by a first neural network model in the cloud server to obtain parameter value ranges corresponding to multiple parameters required for parameter tuning of the target application, and parameter tuning information of the target application is determined according to the parameter value ranges corresponding to the multiple parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; Feedback parameter tuning information of the target application to the client; The at least one application includes the target application, the first neural network model learns the parameter value ranges of the application parameters in different operating environments according to the data information of the historical parameter tuning, and the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning are obtained by processing the environment information through the first neural network model, including: Determine a target operating environment from the operating environments learned by the first neural network model according to the environmental information, wherein a similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; According to the parameter value ranges of the multiple parameters required for the target application parameter tuning in the target operating environment, the parameter value ranges corresponding to the multiple parameters required for the target application parameter tuning in the current operating environment are determined.

12. A device for determining parameter tuning information, It is characterized in that include: A first determining unit, configured to determine environment information of a current running environment of a target application, wherein the target application is an application to be parameter-tuned; a processing unit, configured to process the environmental information through a first neural network model to obtain parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application, wherein the first neural network model is trained based on data information of historical parameter tuning of at least one application; A second determining unit, configured to determine parameter tuning information of the target application according to parameter value ranges corresponding to a plurality of parameters required for parameter tuning of the target application; The at least one application includes the target application, the first neural network model learns to obtain parameter value ranges of the application parameters under different operating environments according to the data information of the historical parameter tuning, and the processing unit includes: A second determining subunit is used to determine a target operating environment from the operating environments that have been learned by the first neural network model according to the environmental information, wherein the similarity between the environmental information of the target operating environment and the environmental information of the current operating environment is greater than a preset threshold; The third determination subunit is used to determine the parameter value ranges of the multiple parameters required for tuning the target application parameters in the current operating environment according to the parameter value ranges of the multiple parameters required for tuning the target application parameters in the target operating environment.

13. An electronic device, It is characterized in that include: A memory storing an executable program; A processor is used to run the program, wherein the program, when running, executes the method for determining parameter tuning information of any one of claims 1 to 11.

14. A computer program product, It is characterized in that The method comprises a computer program or an instruction, which, when executed by a processor, implements the method for determining parameter tuning information according to any one of claims 1 to 11.