Data processing system for optimizing model operation

Abstract

The invention discloses a data processing system for optimizing model operation, and the system carries out the analysis and decision processing of first model information and operation environment information, and obtains target operation result information.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more specifically to a data processing system for optimizing model operation. Background Technology

[0002] With the rapid development of artificial intelligence (AI) technology, more and more AI models are running on terminals, and scenarios requiring the simultaneous operation of multiple AI models are also gradually increasing. However, the AI ​​computing power and system resources on terminals are limited, and the inference time of AI models fluctuates significantly depending on the scarcity of system resources. This results in poor actual performance of AI algorithms when running multiple AI algorithms simultaneously on a terminal. Summary of the Invention

[0003] This application provides a data processing system for optimizing model operation.

[0004] In a first aspect, this application provides a method comprising:

[0005] Obtain the first model information and runtime environment information;

[0006] The information from the first model and the operating environment is analyzed and processed to obtain the target operating result information.

[0007] Secondly, this application provides a system comprising:

[0008] The information acquisition module is used to acquire information about the first model and the operating environment.

[0009] The analysis and processing module is used to analyze and make decisions on the first model information and the operating environment information to obtain the target operating result information.

[0010] Thirdly, this application also provides a computer device, which includes:

[0011] One or more processors;

[0012] Memory; and

[0013] One or more applications, wherein the applications are stored in memory and configured to be executed by a processor to implement the methods of any one of the first aspects.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to perform the steps of the method in any of the first aspects. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of a data processing system provided in an embodiment of the present invention;

[0017] Figure 2 This is a flowchart of one embodiment of the data processing method provided by the present invention;

[0018] Figure 3 This is a flowchart illustrating a specific embodiment of the present invention for analyzing and making decisions based on first model information and operating environment information.

[0019] Figure 4 This is a flowchart of another embodiment of the data processing method provided in this invention;

[0020] Figure 5 This is a schematic block diagram of the data processing system provided in the embodiments of the present invention;

[0021] Figure 6 This is a schematic diagram of an embodiment of the computer device provided in this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this application. Furthermore, the terms "first," "second," and "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," "third," etc., may explicitly or implicitly include one or more features. In the description of this application, "several" means at least one, and "multiple" means two or more, unless otherwise explicitly specified.

[0024] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0025] It should be noted that since the method in this application embodiment is executed in a computer device, the processing objects of each computer device exist in the form of data or information, such as time, which is essentially time information. It is understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, they are all corresponding data that exist so that the computer device can process them. Specific details will not be elaborated here.

[0026] This application provides a data processing method, system, device, and storage medium for optimizing model operation, which will be described in detail below.

[0027] Please see Figure 1 , Figure 1 This is a schematic diagram of a data processing system provided in an embodiment of this application. The data processing system may include a computer device 100, which integrates the data processing system, such as... Figure 1 Computer equipment in the country.

[0028] In this embodiment, the computer device 100 is mainly used to acquire first model information and operating environment information; to analyze and make decisions on the first model information and operating environment information to obtain target operating result information. While ensuring the operating effect of the AI ​​model, the device can enable the terminal to support the simultaneous operation of more AI models with limited resources.

[0029] In this embodiment, the computer device 100 can be a standalone server, a server network, or a server cluster. For example, the computer device 100 described in this embodiment includes, but is not limited to, a computer, a network host, a single network server, a set of multiple network servers, or a cloud server composed of multiple servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0030] It is understood that the computer device 100 used in the embodiments of this application can be a device that includes both receiving and transmitting hardware, that is, a device having receiving and transmitting hardware capable of performing bidirectional communication on a bidirectional communication link. Such a device may include: cellular or other communication devices having a single-line display, a multi-line display, or a cellular or other communication device without a multi-line display. Specifically, the computer device 100 may be a desktop terminal or a mobile terminal, and may also be one of a mobile phone, tablet computer, laptop computer, etc.

[0031] Those skilled in the art will understand that Figure 1 The application environment shown is merely one application scenario of the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include those that are more specific to this application. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one computer device is shown in the diagram. It is understood that the data processing system may also include one or more other services, which are not limited here.

[0032] In addition, such as Figure 1 As shown, the data processing system may also include a memory 200 for storing data, such as running characteristic information, such as first running characteristic information, second running characteristic information, third running characteristic information, etc., and target running result information, such as first running result information, second running result information, etc.

[0033] It should be noted that, Figure 1The schematic diagram of the data processing system shown is merely an example. The data processing system and scenario described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of data processing systems and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0034] like Figure 2 The diagram shown is a flowchart of an embodiment of the data processing method in this application. The data processing method may include the following steps S201 to S202, as detailed below:

[0035] S201. Obtain the first model information and the operating environment information.

[0036] In this embodiment, the first model information is the model information of the second target model currently requested by the user. The first model information may include one or more of the following: the resource usage information of the second target model during runtime, the inference time of the second target model, and the number of times the second target model runs within a first time period. This embodiment does not limit the information.

[0037] Furthermore, the operating environment information includes various environmental conditions and data related to the operation of the second target model. The operating environment information can include not only physical environment information, but also software environment information, network environment information, etc. For example, if the second target model runs on the target device, the operating environment information includes the temperature and humidity of the target device, the number of models currently running on the target device, the system resource usage of the target device, the network latency and bandwidth of the target device, etc. This embodiment does not limit this.

[0038] Optionally, the computer device can obtain the first model information and runtime environment information in various ways. For example, if the second target model runs on the computer device, the computer device can directly obtain the first model information and runtime environment information. Alternatively, if the second target model runs on other devices, the computer device can obtain the first model information and runtime environment information from other devices via networks, Bluetooth, etc. This embodiment does not impose limitations. For instance, in this embodiment, the data processing method is executed via a smartphone, and the second target model also runs on the smartphone. When the smartphone receives a user's request to run the second target model, it can directly obtain the first model information and runtime environment information. Alternatively, in this embodiment, the data processing method is executed via a server, and the second target model runs on the smartphone. When the smartphone receives a user's request to run the second target model, it can send the first model information and runtime environment information to the server via networks, Bluetooth, etc., so that the server can perform data processing based on the first model information and runtime environment information.

[0039] It should be noted that the target model in this embodiment refers to an AI algorithm composed of one or more AI models. That is, the target model includes at least one AI model. Artificial Intelligence (AI) model refers to a system that can simulate human intelligent behavior through computer algorithms and data training. Through artificial intelligence models, computers can complete a series of tasks including image recognition, speech recognition, natural language processing, machine translation, etc.

[0040] S202. Analyze and process the information of the first model and the operating environment to obtain the target operating result information.

[0041] In this embodiment, the target running result information is the running result information of the second target model. Based on the target running result information, it can be determined whether to run the second target model. Optionally, the target running result information includes first running result information, and / or second running result information, and / or third running result information, and / or fourth running result information. The first running result information indicates that the second target model is allowed to run; the second running result information indicates that all currently running first target models are shut down, and the second target model is allowed to run; the third running result information indicates that some currently running first target models are shut down, and the second target model is allowed to run; the fourth running result information indicates that the second target model is not allowed to run. This embodiment analyzes and processes the first model information and the running environment information to obtain the target running result information. It can combine the first model information and the running environment information to schedule model operation, ensuring the model's running effect while enabling the edge to support the simultaneous operation of more models with limited resources.

[0042] In a specific implementation method, refer to Figure 3 As shown, the analysis and decision-making process of the first model information and the operating environment information in step S202 above to obtain the target operating result information may include steps S301 to S302, as follows:

[0043] S301. Extract features from the operating environment information to obtain the first operating feature information.

[0044] In this embodiment of the application, the first running feature information represents the number of several first target models currently running. The second target model runs on the target device, and the several first target models are the models currently running forward on the target device. For example, if the second target model runs on a computer device, and the number of models currently running on the computer device is 3, then the first running feature information is 3. If the number of models currently running on the computer device is 0, then the first running feature information is 0.

[0045] In one specific embodiment, the operating environment information includes the number of models currently running on the target device. By extracting features from the operating environment information, the first operating environment information can be obtained.

[0046] S302. Based on the first model information and the first operational characteristic information, determine the target operational result information.

[0047] In one specific embodiment, the step of determining the target running result information based on the first model information and the first running feature information specifically includes: if the feature value corresponding to the first running feature information is less than or equal to the first threshold, the target running result information is determined to be the first running result information; and / or, if the feature value corresponding to the first running feature information is greater than the first threshold, the target running result information is determined based on the first model information.

[0048] In this embodiment, the first threshold is a pre-set threshold used to measure the number of currently running first target models. The first threshold can be set according to actual needs; for example, the first threshold can be set to 0, 1, 2, etc. The first running result information indicates that the second target model is allowed to run. If the feature value corresponding to the first running feature information is less than or equal to the first threshold, it indicates that the number of currently running first target models is small, and the target running result information is determined as the first running result information, that is, the first target model is allowed to run directly. Conversely, if the feature value corresponding to the first running feature information is greater than the first threshold, it indicates that the number of currently running first target models is large, and the remaining resources of the target device may not be sufficient to support the running of the second target model. Therefore, the target running result information is determined based on the first model information.

[0049] In one specific embodiment, the first model information includes resource usage information and second operational feature information. The resource usage information characterizes the resource usage of the second target model during runtime. The second operational feature information includes several first feature information and second feature information corresponding to the second target model. The second target model includes several first models. The first feature information characterizes the inference time of each first model, and the second feature information characterizes the number of times the second target model runs within a first time period. For example, the second target model is implemented through AI model 1, AI model 2, ..., AI model N. The second operational feature information includes the inference time of AI model 1 to AI model N and the minimum number of runs of the second target model within one second.

[0050] Optionally, the step of determining the target operation result information based on the first model information specifically includes: if the resource occupancy value corresponding to the resource occupancy information is greater than or equal to the second threshold, determining the target operation result information as the second operation result information; and / or, if the resource occupancy value corresponding to the resource occupancy information is less than the second threshold, determining the target operation result information based on the first target feature information corresponding to the second operation feature information.

[0051] In this embodiment, the second threshold is a pre-set threshold used to measure whether the second target model needs to exclusively occupy system resources during runtime. The second threshold can be set according to actual needs; for example, the second threshold can be set to 100%, 90%, 80%, etc. The second running result information represents shutting down all currently running first target models and allowing the second target model to run. If the resource occupation value corresponding to the resource occupation information is greater than or equal to the second threshold, it indicates that the second target model needs to exclusively occupy system resources during runtime, and the target running result information is determined as the second running result information, that is, shutting down all currently running first target models and allowing the second target model to run; conversely, if the resource occupation value corresponding to the resource occupation information is less than the second threshold, it indicates that the second target model does not need to exclusively occupy system resources during runtime, and the target running result information is determined based on the first target feature information corresponding to the second running feature information. Further, the first target feature information represents the theoretical time consumption of the second target model. The first target feature information is determined in the following way: a weighted summation of several first feature information is performed to obtain the third feature information; the third feature information and the second feature information are multiplied to obtain the first target feature information.

[0052] In this embodiment, the third feature information represents the total inference time of several first models. In a specific embodiment, the process of determining the third feature information can be expressed as follows: S1 represents the third feature information, t i r represents the first feature information of the i-th first model. i Indicates t i The corresponding weight values, where n represents the number of the first model, and r i It can be set as needed, 0 <r i ≤1, for example, r1, r2, ..., r i If all equal to 1, then

[0053] In one specific embodiment, the step of determining the target running result information based on the first target feature information corresponding to the second running feature information specifically includes: determining the second target feature information based on the first target feature information and the third running feature information corresponding to the second running feature information; and determining the target running result information based on the second target feature information. In this embodiment, determining the second target feature information based on the first target feature information and the third running feature information, and determining the target running result information based on the second target feature information, allows multiple models to be scheduled to run in a certain order according to the actual inference speed of the AI ​​model. When multiple models run simultaneously, the running effect of multiple models can be guaranteed.

[0054] Optionally, the step of determining the second target feature information based on the first target feature information and the third operating feature information corresponding to the second operating feature information specifically includes: determining the third target feature information based on the third operating feature information; and subtracting the first target feature information corresponding to the third target feature information and the second operating feature information to obtain the second target feature information. In this embodiment, the third target feature information can characterize the available idle time of the target device, and the first target feature information can characterize the theoretical time consumption of the second target model. Subtracting the third target feature information from the first target feature information can determine whether the remaining idle resources of the target device are sufficient to support the operation of the second target model.

[0055] In one specific embodiment, the process of subtracting the third target feature information from the first target feature information can be expressed as: S = S3 - S2, where S represents the second target feature information, S2 represents the first target feature information, and S3 represents the third target feature information.

[0056] In one specific embodiment, the third operational feature information includes a fourth feature information and a fifth feature information corresponding to each first target model. The fourth feature information represents the time required for each first target model to run once, and the fifth feature information represents the number of times each first target model runs within a first time period. For example, a plurality of first target models include first target model 1, first target model 2, ..., first target model N, and the third operational feature information includes the time required for each of the first target models 1, first target model 2, ..., first target model N to run once, and the actual number of times each first target model runs within one second.

[0057] Optionally, the steps for determining the third target feature information based on the third operational feature information specifically include: multiplying the fourth and fifth feature information corresponding to each first target model to obtain the sixth feature information corresponding to each first target model; adding the sixth feature information corresponding to several first target models to obtain the seventh feature information; and subtracting the first and seventh feature information to obtain the third target feature information.

[0058] In one specific embodiment, the first time is 1 second, the sixth feature information represents the actual time consumed per second of each first target model, and the seventh feature information represents the total time consumed per second of several first target models. Subtracting the first time and the seventh feature information means subtracting the total time consumed per second of several first target models from 1 second, thereby obtaining the available idle time of the target device.

[0059] In one specific embodiment, the fifth feature information corresponding to each first target model is determined based on the following method: obtaining the eighth feature information of each first target model; dividing the eighth feature information by the second time to obtain the fifth feature information corresponding to each first target model.

[0060] In this embodiment of the application, the eighth feature information represents the number of times each first target model runs in the second time period. The second time period can be set as needed. For example, the second time period can be set to 1 second, 2 seconds, 5 seconds, 10 seconds, etc. By dividing the eighth feature information and the second time period, the actual number of times each first target model runs in one second can be obtained.

[0061] In one specific embodiment, the step of determining the target running result information based on the second target feature information specifically includes: if the target feature value corresponding to the second target feature information is greater than the third threshold, determining the target running result information as the first running result information; and / or, if the target feature value is less than or equal to the third threshold, determining the target running result information based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models.

[0062] In this embodiment, the third threshold is a pre-set threshold used to measure whether the remaining computing resources of the target device are sufficient. If the target feature value corresponding to the second target feature information is greater than the third threshold, it is determined that the remaining computing resources of the target device are sufficient and can support the operation of the second target model. In this case, the target operation result information is determined as the first operation result information, that is, the second target model is directly allowed to run. Conversely, if the target feature value is less than or equal to the third threshold, it indicates that the remaining computing resources of the target device are insufficient. Then, the target operation result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models. In this embodiment, when the remaining computing resources cannot support the operation of the second target model, the target operation result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models. This ensures that high-priority models run first, while guaranteeing the model's operating effect.

[0063] Optionally, the first priority information is information related to the priority of the second target model, which can characterize the priority of the second target model. The second priority information is information related to the priority of each of the several first target models, which can characterize the priority of each first target model. In a specific embodiment, the step of determining the target running result information based on the first priority information corresponding to the second target model and the second priority information corresponding to the several first target models specifically includes: determining whether there is a model with a lower priority than the second target model among the several first target models based on the first priority information corresponding to the second target model and the second priority information corresponding to the several first target models; when there is a model with a lower priority than the second target model among the several first target models, determining the target running result information as the third running result information; and / or, when there is no model with a lower priority than the second target model among the several first target models, determining the target running result information as the fourth running result information.

[0064] In this embodiment, when the terminal used to run the second target model does not have sufficient computing resources to run the second target model, it is determined whether there is a model with a lower priority than the second target model among the currently running first target models. If there is a model with a lower priority than the second target model among the currently running first target models, the target running result information is determined as the third running result information, that is, some models among the first target models are closed and the second target model is allowed to run; otherwise, if there is no model with a lower priority than the second target model among the currently running first target models, the target running result information is determined as the fourth running result information, that is, the second target model is not allowed to run. This setting can ensure that high-priority models run first while guaranteeing the running effect of the models.

[0065] In one specific embodiment, since the inference time of the AI ​​model can fluctuate significantly depending on the scarcity of system resources, in order to determine whether several currently running models can actually run simultaneously, refer to... Figure 4 As shown, after analyzing and making decisions on the first model information and the operating environment information in step S202 to obtain the target operating result information, steps S401 to S402 may be included, as follows:

[0066] S401. Obtain the fourth operational feature information and the fifth operational feature information of several currently running third target models.

[0067] In this embodiment, the plurality of third target models are the models currently running on the target device after running the second target model. The fourth running feature information represents the actual number of times each third target model runs within the first time period, and the fifth running feature information represents the minimum number of times each third target model runs within the first time period.

[0068] S402. Based on the fourth and fifth operational feature information, optimize the operational environment of several third target models.

[0069] In one specific embodiment, the step of optimizing the operating environment of several third target models based on fourth and fifth operating feature information specifically includes: for any third target model among several third target models, comparing the fourth operating feature information of the third target model with the fifth operating feature information of the third target model; if the number of runs corresponding to the fourth operating feature information of the third target model is less than the number of runs corresponding to the fifth operating feature information of the third target model, determining the fourth target model from among several third target models based on the third priority information of several third target models, and stopping the operation of the fourth target model.

[0070] In this embodiment, the third priority information is information related to the priorities of several third target models, and the third priority information can characterize the priorities of several third target models. The fourth target model is the model whose priority among several third target models satisfies the first condition. The model whose priority satisfies the first condition can be the model with the lowest priority, or it can be a model whose priority is within the priority range. This embodiment does not impose any limitations.

[0071] Optionally, if the number of runs corresponding to the fourth operational feature information of the third target model is greater than or equal to the number of runs corresponding to the fifth operational feature information of the third target model, it indicates that the computing resources can satisfy the operation of several third target models; if the number of runs corresponding to the fourth operational feature information of the third target model is less than the number of runs corresponding to the fifth operational feature information of the third target model, it indicates that the computing resources cannot satisfy the operation of several third target models, then the fourth target model is stopped from running, thereby ensuring that high-priority models run first while guaranteeing the model's performance.

[0072] In summary, the data processing method provided in this implementation plan, by acquiring first model information and operating environment information, analyzes and processes this information to obtain target operating result information. It can schedule AI model operation based on the first model information and operating environment information, ensuring the effectiveness of AI model operation while enabling the edge to support the simultaneous operation of more AI models with limited resources. Furthermore, if the target feature value is less than or equal to a third threshold, the target operating result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models. Model operation can then be scheduled based on model priority. While prioritizing the execution of high-priority models, the effectiveness of the models can be guaranteed. Furthermore, for any third target model among several third target models, the fourth operational feature information of the third target model is compared with the fifth operational feature information of the third target model. If the number of executions corresponding to the fourth operational feature information of the third target model is less than the number of executions corresponding to the fifth operational feature information of the third target model, the fourth target model is determined from the several third target models based on the third priority information of several third target models, and the execution of the fourth target model is stopped. This ensures that high-priority models run first even when computing resources do not support the simultaneous execution of multiple models.

[0073] To better implement the data processing method in the embodiments of this application, a data processing system is also provided in the embodiments of this application, such as... Figure 5 As shown, the data processing system 600 includes:

[0074] Information acquisition module 610 is used to acquire first model information and operating environment information;

[0075] The analysis and processing module 620 is used to analyze and make decisions on the first model information and the operating environment information to obtain the target operating result information.

[0076] In this embodiment, by acquiring first model information and operating environment information, and analyzing and making decisions on the first model information and operating environment information to obtain target operating result information, the AI ​​model can be scheduled to run in combination with the first model information and operating environment information. While ensuring the operating effect of the AI ​​model, the edge can support the simultaneous operation of more AI models with limited resources.

[0077] In some embodiments of this application, the analysis and processing module 620 performs analysis and decision processing on the first model information and the operating environment information to obtain target operating result information, including:

[0078] Feature extraction is performed on the runtime environment information to obtain the first runtime feature information; the first runtime feature information represents the number of several first target models currently running;

[0079] Based on the first model information and the first operational characteristic information, the target operational result information is determined.

[0080] In some embodiments of this application, the analysis and processing module 620 determines the target running result information based on the first model information and the first running feature information, including:

[0081] If the feature value corresponding to the first running feature information is less than or equal to the first threshold, the target running result information is determined to be the first running result information; and / or,

[0082] If the feature value corresponding to the first running feature information is greater than the first threshold, the target running result information is determined based on the first model information.

[0083] In some embodiments of this application, the first model information includes resource usage information and second operational characteristic information. The resource usage information characterizes the resource usage of the second target model during runtime. Based on the first model information, the analysis and processing module 620 determines the target operational result information, including:

[0084] If the resource usage value corresponding to the resource usage information is greater than or equal to the second threshold, the target running result information is determined as the second running result information; and / or,

[0085] If the resource occupancy value corresponding to the resource occupancy information is less than the second threshold, the target operation result information is determined based on the first target feature information corresponding to the second operation feature information.

[0086] In some embodiments of this application, the analysis and processing module 620 determines the target running result information based on the first target feature information corresponding to the second running feature information, including:

[0087] Based on the first target feature information and the third operational feature information corresponding to the second operational feature information, the second target feature information is determined;

[0088] Based on the second target feature information, the target operation result information is determined.

[0089] In some embodiments of this application, the second running feature information includes a plurality of first feature information and second feature information corresponding to the second target model. The second target model includes a plurality of first models. The first feature information represents the inference time of each first model, and the second feature information represents the number of times the second target model runs in the first time period.

[0090] The first target feature information is obtained by the analysis and processing module 620 through the following steps:

[0091] The third feature information is obtained by performing a weighted summation on several first feature information.

[0092] The third feature information and the second feature information are multiplied together to obtain the first target feature information.

[0093] In some embodiments of this application, the analysis and processing module 620 determines the second target feature information based on the first target feature information and the third running feature information corresponding to the second running feature information, including:

[0094] Based on the third operational characteristic information, the third target characteristic information is determined;

[0095] The second target feature information is obtained by subtracting the first target feature information corresponding to the third target feature information and the second running feature information.

[0096] In some embodiments of this application, the third running feature information includes a fourth feature information and a fifth feature information corresponding to each first target model. The fourth feature information represents the time required for each first target model to run once, and the fifth feature information represents the number of times each first target model runs within a first time period.

[0097] Based on the third operational feature information, the analysis and processing module 620 determines the third target feature information, including:

[0098] Multiply the fourth and fifth feature information corresponding to each first target model to obtain the sixth feature information corresponding to each first target model;

[0099] The sixth feature information corresponding to several first target models is added together to obtain the seventh feature information;

[0100] Subtract the first and seventh feature information to obtain the third target feature information.

[0101] In some embodiments of this application, the fifth feature information corresponding to each first target model is obtained by the analysis and processing module 620 through the following steps:

[0102] Obtain the eighth feature information of each first target model; the eighth feature information represents the number of times each first target model runs in the second time period;

[0103] The eighth feature information is divided by the second time to obtain the fifth feature information corresponding to each first target model.

[0104] In some embodiments of this application, the analysis and processing module 620 determines target execution result information based on the second target feature information, including:

[0105] If the target feature value corresponding to the second target feature information is greater than the third threshold, the target running result information is determined as the first running result information; and / or,

[0106] If the target feature value is less than or equal to the third threshold, the target running result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models.

[0107] In some embodiments of this application, the analysis and processing module 620 determines target execution result information based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models, including:

[0108] Based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models, determine whether there are models among the several first target models with a lower priority than the second target model;

[0109] When there is a model among several first-target models with a lower priority than the second-target model, the target execution result information is determined as the third execution result information; and / or,

[0110] When there is no model with a lower priority than the second target model among several first target models, the target running result information is determined as the fourth running result information.

[0111] In some embodiments of this application, after the analysis and processing module 620 performs analysis and decision processing on the first model information and the operating environment information to obtain the target operating result information, the analysis and processing module 620 is further used for:

[0112] Obtain the fourth operational feature information and the fifth operational feature information of several currently running third target models; the fourth operational feature information represents the actual number of times each third target model runs in the first time period, and the fifth operational feature information represents the minimum number of times each third target model runs in the first time period.

[0113] Based on the fourth and fifth operational feature information, the operational environment of several third target models is optimized.

[0114] In some embodiments of this application, the analysis and processing module 620 optimizes the operating environment of several third target models based on fourth and fifth operating feature information, including:

[0115] For any third objective model among several third objective models, compare the fourth operational feature information of the third objective model with the fifth operational feature information of the third objective model;

[0116] If the number of runs corresponding to the fourth operational feature information of the third target model is less than the number of runs corresponding to the fifth operational feature information of the third target model, a fourth target model is determined from the several third target models based on the third priority information of several third target models, and the fourth target model is stopped from running; the fourth target model is the model whose priority satisfies the first condition among the multiple third target models.

[0117] This application also provides a computer device that integrates any of the data processing systems provided in this application. The computer device includes:

[0118] One or more processors;

[0119] Memory; and

[0120] One or more applications, wherein the applications are stored in memory and configured to be executed by a processor from the steps of the data processing method in any of the embodiments described above.

[0121] This application also provides a computer device that integrates any of the data processing systems provided in this application. For example... Figure 6 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:

[0122] The computer device may include components such as a processor 801 with one or more processing cores, a memory 802 with one or more computer-readable storage media, a power supply 803, and an input unit 804. Those skilled in the art will understand that... Figure 6The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0123] The processor 801 is the control center of the computer device. It connects various parts of the computer device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 802, and by calling data stored in the memory 802, thereby providing overall monitoring of the computer device. Optionally, the processor 801 may include one or more processing cores; optionally, the processor 801 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor 801.

[0124] The memory 802 can be used to store software programs and modules. The processor 801 executes various functional applications and data processing by running the software programs and modules stored in the memory 802. The memory 802 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 802 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 802 may also include a memory controller to provide the processor 801 with access to the memory 802.

[0125] The computer device also includes a power supply 803 that supplies power to the various components. Optionally, the power supply 803 can be logically connected to the processor 801 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 803 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0126] The computer device may also include an input unit 804, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0127] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 801 in the computer device loads the executable files corresponding to the processes of one or more application programs into the memory 802 according to the following instructions, and the processor 801 runs the application programs stored in the memory 802 to realize various functions, as follows:

[0128] Obtain the first model information and runtime environment information;

[0129] The information from the first model and the operating environment is analyzed and processed to obtain the target operating result information.

[0130] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0131] Therefore, embodiments of this application provide a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc. A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in any of the data processing methods provided in embodiments of this application. For example, the computer program loaded by the processor can execute the following steps:

[0132] Obtain the first model information and runtime environment information;

[0133] The information from the first model and the operating environment is analyzed and processed to obtain the target operating result information.

[0134] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the detailed descriptions of other embodiments above, which will not be repeated here.

[0135] In practice, each of the above units or structures can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units or structures, please refer to the previous method embodiments, which will not be repeated here.

[0136] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0137] The foregoing has provided a detailed description of a data processing method, system, device, and storage medium for optimizing model operation, as provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method, characterized in that, include: Obtain the first model information and runtime environment information; The first model information and the operating environment information are analyzed and processed to obtain the target operating result information.

2. The method according to claim 1, characterized in that, The step of analyzing and making decisions on the first model information and the operating environment information to obtain target operating result information includes: Feature extraction is performed on the runtime environment information to obtain first runtime feature information; the first runtime feature information represents the number of several first target models currently running; Based on the first model information and the first operational feature information, the target operational result information is determined.

3. The method according to claim 2, characterized in that, The step of determining the target running result information based on the first model information and the first running feature information includes: If the feature value corresponding to the first running feature information is less than or equal to the first threshold, the target running result information is determined to be the first running result information; and / or, If the feature value corresponding to the first running feature information is greater than the first threshold, the target running result information is determined based on the first model information.

4. The method according to claim 3, characterized in that, The first model information includes resource usage information and second operational feature information, wherein the resource usage information characterizes the resource usage of the second target model during runtime; The step of determining the target execution result information based on the first model information includes: If the resource occupancy value corresponding to the resource occupancy information is greater than or equal to the second threshold, the target running result information is determined to be the second running result information. And / or, If the resource occupancy value corresponding to the resource occupancy information is less than the second threshold, the target operation result information is determined based on the first target feature information corresponding to the second operation feature information.

5. The method according to claim 4, characterized in that, The step of determining the target running result information based on the first target feature information corresponding to the second running feature information includes: Based on the first target feature information and the third operating feature information corresponding to the second operating feature information, the second target feature information is determined; Based on the second target feature information, the target operation result information is determined.

6. The method according to claim 4, characterized in that, The second running feature information includes several first feature information and second feature information corresponding to the second target model. The second target model includes several first models. The first feature information represents the inference time of each first model, and the second feature information represents the number of times the second target model runs in the first time period. The first target feature information is determined based on the following method: A weighted summation of several first feature information items is performed to obtain third feature information; The third feature information and the second feature information are multiplied together to obtain the first target feature information.

7. The method according to claim 5, characterized in that, The step of determining the second target feature information based on the first target feature information and the third running feature information corresponding to the second running feature information includes: Based on the third operational characteristic information, the third target characteristic information is determined; The second target feature information is obtained by subtracting the third target feature information from the first target feature information corresponding to the second running feature information.

8. The method according to claim 7, characterized in that, The third operational feature information includes a fourth feature information and a fifth feature information corresponding to each of the first target models. The fourth feature information represents the time required for each of the first target models to run once, and the fifth feature information represents the number of times each of the first target models runs within a first time period. The determination of the third target feature information based on the third operational feature information includes: The fourth and fifth feature information corresponding to each first target model are multiplied together to obtain the sixth feature information corresponding to each first target model. The sixth feature information corresponding to several first target models is added together to obtain the seventh feature information; The first time and the seventh feature information are subtracted to obtain the third target feature information.

9. The method according to claim 8, characterized in that, The fifth feature information corresponding to each of the first target models is determined based on the following method: Obtain the eighth feature information of each of the first target models; the eighth feature information represents the number of times each of the first target models runs in the second time period; The eighth feature information and the second time are divided to obtain the fifth feature information corresponding to each of the first target models.

10. The method according to claim 5, characterized in that, The determination of target execution result information based on the second target feature information includes: If the target feature value corresponding to the second target feature information is greater than the third threshold, the target running result information is determined to be the first running result information; and / or, If the target feature value is less than or equal to the third threshold, the target running result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models.

11. The method according to claim 10, characterized in that, The determination of target execution result information based on the first priority information corresponding to the second target model and several second priority information corresponding to the first target model includes: Based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models, determine whether there is a model among several first target models with a lower priority than the second target model; When there is a model among the first target models with a lower priority than the second target model, the target execution result information is determined as the third execution result information; and / or, When there is no model with a lower priority than the second target model among the first target models, the target running result information is determined to be the fourth running result information.

12. The method according to any one of claims 1 to 11, characterized in that, After analyzing and making decisions on the first model information and the operating environment information to obtain the target operating result information, the process includes: Obtain fourth operational feature information and fifth operational feature information of several currently running third target models; the fourth operational feature information represents the actual number of times each third target model runs within a first time period, and the fifth operational feature information represents the minimum number of times each third target model runs within the first time period; Based on the fourth and fifth operational feature information, the operational environment of several third target models is optimized.

13. The method according to claim 12, characterized in that, The optimization process for the operating environment of several third target models based on the fourth and fifth operating feature information includes: For any one of the plurality of third target models, the fourth operational feature information of the third target model is compared with the fifth operational feature information of the third target model; If the number of runs corresponding to the fourth running feature information of the third target model is less than the number of runs corresponding to the fifth running feature information of the third target model, a fourth target model is determined from the third target models based on the third priority information of the third target models, and the fourth target model is stopped from running; the fourth target model is the model whose priority satisfies the first condition among the multiple third target models.

14. A system, characterized in that, include: The information acquisition module is used to acquire information about the first model and the operating environment. The analysis and processing module is used to analyze and make decisions on the first model information and the operating environment information to obtain the target operating result information. Optionally, the analysis and processing module performs analysis and decision processing on the first model information and the operating environment information to obtain target operating result information, including: Feature extraction is performed on the runtime environment information to obtain first runtime feature information; the first runtime feature information represents the number of several first target models currently running; Based on the first model information and the first operational feature information, the target operational result information is determined; Optionally, the analysis and processing module determines the target execution result information based on the first model information and the first execution feature information, including: If the feature value corresponding to the first running feature information is less than or equal to the first threshold, the target running result information is determined to be the first running result information; and / or, If the feature value corresponding to the first running feature information is greater than the first threshold, the target running result information is determined based on the first model information; Optionally, the first model information includes resource usage information and second operational characteristic information. The resource usage information characterizes the resource usage during the operation of the second target model. Based on the first model information, the analysis and processing module determines the target operational result information, including: If the resource occupancy value corresponding to the resource occupancy information is greater than or equal to the second threshold, the target running result information is determined to be the second running result information; and / or, If the resource occupancy value corresponding to the resource occupancy information is less than the second threshold, the target operation result information is determined based on the first target feature information corresponding to the second operation feature information; Optionally, the analysis and processing module determines the target running result information based on the first target feature information corresponding to the second running feature information, including: Based on the first target feature information and the third operating feature information corresponding to the second operating feature information, the second target feature information is determined; Based on the second target feature information, the target operation result information is determined; Optionally, the second running feature information includes a plurality of first feature information and second feature information corresponding to the second target model. The second target model includes a plurality of first models. The first feature information represents the inference time of each first model, and the second feature information represents the number of times the second target model runs in the first time period. The first target feature information is obtained by the analysis and processing module through the following steps: A weighted summation of several first feature information items is performed to obtain third feature information; The third feature information and the second feature information are multiplied together to obtain the first target feature information; Optionally, the analysis and processing module determines the second target feature information based on the first target feature information and the third running feature information corresponding to the second running feature information, including: Based on the third operational characteristic information, the third target characteristic information is determined; Subtract the first target feature information corresponding to the third target feature information and the second running feature information to obtain the second target feature information; Optionally, the third running feature information includes a fourth feature information and a fifth feature information corresponding to each of the first target models. The fourth feature information represents the time required for each of the first target models to run once, and the fifth feature information represents the number of times each of the first target models runs within a first time period. The analysis and processing module determines the third target feature information based on the third operational feature information, including: The fourth and fifth feature information corresponding to each first target model are multiplied together to obtain the sixth feature information corresponding to each first target model. The sixth feature information corresponding to several first target models is added together to obtain the seventh feature information; Subtract the first time and the seventh feature information to obtain the third target feature information; Optionally, the fifth feature information corresponding to each of the first target models is obtained by the analysis and processing module through the following steps: Obtain the eighth feature information of each of the first target models; the eighth feature information represents the number of times each of the first target models runs in the second time period; The eighth feature information and the second time are divided to obtain the fifth feature information corresponding to each first target model; Optionally, the analysis and processing module determines the target execution result information based on the second target feature information, including: If the target feature value corresponding to the second target feature information is greater than the third threshold, the target running result information is determined to be the first running result information; and / or, If the target feature value is less than or equal to the third threshold, the target running result information is determined based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models; Optionally, the analysis and processing module determines the target execution result information based on the first priority information corresponding to the second target model and several second priority information corresponding to the first target model, including: Based on the first priority information corresponding to the second target model and the second priority information corresponding to several first target models, determine whether there is a model among several first target models with a lower priority than the second target model; When there is a model among the first target models with a lower priority than the second target model, the target execution result information is determined as the third execution result information; and / or, When there is no model with a lower priority than the second target model among the several first target models, the target running result information is determined to be the fourth running result information; Optionally, after the analysis and processing module performs analysis and decision processing on the first model information and the operating environment information to obtain the target operating result information, the analysis and processing module is further used for: Obtain fourth operational feature information and fifth operational feature information of several currently running third target models; the fourth operational feature information represents the actual number of times each third target model runs within a first time period, and the fifth operational feature information represents the minimum number of times each third target model runs within the first time period; Based on the fourth and fifth operational feature information, the operational environment of several third target models is optimized. Optionally, the analysis and processing module optimizes the operating environment of several third target models based on the fourth and fifth operating feature information, including: For any one of the plurality of third target models, the fourth operational feature information of the third target model is compared with the fifth operational feature information of the third target model; If the number of runs corresponding to the fourth running feature information of the third target model is less than the number of runs corresponding to the fifth running feature information of the third target model, a fourth target model is determined from the third target models based on the third priority information of the third target models, and the fourth target model is stopped from running; the fourth target model is the model whose priority satisfies the first condition among the multiple third target models.

15. A computer device, characterized in that, The computer device includes: One or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the method of any one of claims 1 to 13.

16. A computer-readable storage medium, characterized in that, It contains a computer program that is loaded by a processor to perform the steps of the method according to any one of claims 1 to 13.

Images