A method, apparatus, system, and device for determining a configuration for a business model

Abstract

The application discloses a method, device, system and equipment for determining configuration of a business model, relates to the technical field of artificial intelligence, and comprises the following steps: collecting data occupied by a business model during running as computing power test data; obtaining a configuration allocation strategy according to the collected computing power test data and a self-supervision model; and finally, configuring the business model according to the configuration allocation strategy, so that the resources, services and models of the configured business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively. Since the process of allocating resources is performed by using a neural network model, the resources can be flexibly allocated according to real-time data of the business model, so that the resource utilization rate of the business model is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a method, device, system and equipment for determining configuration of a business model. BACKGROUND

[0002] With the development of science and technology, artificial intelligence technology is gradually applied to people's life. At present, the algorithm of artificial intelligence is often complex network calculation, which requires large computing power of computer and high demand for computer hardware, such as GPU (Graphics Processing Unit) computing power, GPU memory, network and CPU (Central Processing Unit) computing power and other resources. With the continuous expansion of the application range of artificial intelligence algorithm, it is usually necessary to combine the resources of multiple computers to provide services. In this case, the scheduling of the computing power resources of the business model is particularly important.

[0003] At present, the related configuration scheduling system in the machine learning platform is usually used. However, in actual application, due to the inconsistency of data or algorithm of each business model, the existing configuration scheduling system cannot flexibly adapt to each business model, and cannot efficiently configure the computing power resources of each business model, resulting in waste of computing power resources. SUMMARY

[0004] The present application provides a method, device, system and equipment for determining configuration of a business model, which can flexibly configure the computing power resources of the business model by using neural network model, thereby improving the resource utilization rate when the business model is running.

[0005] In a first aspect, the present application provides a method for determining configuration of a business model, which comprises:

[0006] Collecting computing power test data of the business model, the computing power test data being used to reflect the related data of resources, services and models when the business model runs;

[0007] Determining a configuration allocation strategy based on the computing power test data and a self-supervised model, the self-supervised model being a trained neural network model, and the training samples of the self-supervised model including a plurality of historical computing power data and a training adjustment amount corresponding to each historical computing power data in the plurality of historical computing power data;

[0008] Configuring the business model according to the configuration allocation strategy, and the resources, services and models of the configured business model corresponding to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0009] Optionally, before determining the configuration allocation strategy based on the computing power test data and the self-supervised model, the method further comprises:

[0010] determining whether the computing power test data is in a preset alert range;

[0011] determining a configuration allocation strategy based on the computing power test data and the self-supervised model, including:

[0012] If it is determined that the computing power test data is in the preset alert range, determining the configuration allocation strategy based on the computing power test data and the self-supervised model is executed.

[0013] Optionally, determining the configuration allocation strategy based on the computing power test data and the self-supervised model includes:

[0014] inputting the computing power test data into the self-supervised model to obtain an adjustment amount of each index in the computing power test data;

[0015] determining the configuration allocation strategy according to the adjustment amount of each index.

[0016] Optionally, the training process of the self-supervised model includes:

[0017] obtaining a plurality of historical computing power data of a plurality of business models;

[0018] performing mathematical calculation on the plurality of historical computing power data respectively by using a mathematical model to obtain a training adjustment amount corresponding to each historical computing power data in the plurality of historical computing power data;

[0019] training the initial model by taking the plurality of historical computing power data and the training adjustment amount corresponding to each historical computing power data in the plurality of historical computing power data as training samples to obtain the self-supervised model.

[0020] Optionally, the method further includes:

[0021] saving the computing power test data and the configuration allocation strategy into a database, and the data in the database is used for optimizing training of the self-supervised model.

[0022] Optionally, the computing power test data includes at least one of the following data: resource data occupied by the business model during runtime, service data occupied by the business model during runtime, and model data of the business model.

[0023] In a second aspect, the present application provides a system for determining configuration of a business model, which includes a collection module, a decision module and an execution module,

[0024] The collection module is configured to collect computing power test data of the business model and transmit the computing power test data to the decision module, wherein the computing power test data is used to reflect relevant data of resources, services and models during runtime of the business model.

[0025] The decision module is configured to determine a configuration allocation strategy based on the computing power test data and a self-supervised model, and send the configuration allocation strategy to the execution module, wherein the self-supervised model is a trained neural network model, and training samples of the self-supervised model include a plurality of historical computing power data and a training adjustment amount corresponding to each of the plurality of historical computing power data.

[0026] The execution module is configured to configure the service model according to the configuration allocation strategy, and the resources, services and models of the configured service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0027] In a third aspect, the present application provides a device for determining configuration of a service model, the device comprising:

[0028] The acquisition unit is configured to acquire computing power test data of the service model, wherein the computing power test data is used to reflect relevant data of resources, services and models of the service model during running;

[0029] The decision unit is configured to determine a configuration allocation strategy based on the computing power test data and a self-supervised model, wherein the self-supervised model is a trained neural network model, and training samples of the self-supervised model include a plurality of historical computing power data and a training adjustment amount corresponding to each of the plurality of historical computing power data.

[0030] The processing unit is configured to configure the service model according to the configuration allocation strategy, and the resources, services and models of the configured service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0031] In a fourth aspect, the present application provides an electronic device comprising a memory and a processor:

[0032] The memory is configured to store a computer program.

[0033] The processor is configured to execute the method provided in the first aspect according to the computer program.

[0034] In a fifth aspect, the present application further provides a computer readable storage medium for storing a computer program, wherein the computer program is used to execute the method provided in the first aspect.

[0035] Therefore, the present application has the following advantages:

[0036] The application provides a method for determining configuration of a business model, first collecting data occupied by the business model during running as computing power test data, then obtaining a configuration allocation strategy according to the collected computing power test data and a self-supervised model, and finally configuring the business model according to the configuration allocation strategy, wherein the resources, services and models of the configured business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively. Since the process of allocating resources is performed by using a neural network model, the allocation can be flexible according to real-time data of the business model, and for the same business model, the configuration allocation strategy of the model can be continuously optimized, thereby improving the resource utilization rate of the business model and also improving the running efficiency of the business model to a certain extent. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art according to these drawings.

[0038] Figure 1 A flowchart of a method for determining configuration of a business model in an embodiment of the present application;

[0039] Figure 2 A structure diagram of a system 200 for determining configuration of a business model in an embodiment of the present application;

[0040] Figure 3 A flowchart of an embodiment of a method for determining configuration of a business model in an embodiment of the present application;

[0041] Figure 4 A structure diagram of an apparatus 400 for determining configuration of a business model in an embodiment of the present application;

[0042] Figure 5 A structure diagram of an electronic device 500 in an embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0044] The "first" in the names such as "first computing power test data", "first training sample" mentioned in the embodiments of the present application is only used for name identification and does not represent the first in order. The same rule applies to "second", "third", and the like.

[0045] At present, the system for determining the configuration of the business model is the configuration system of the business model platform. Since the internal configuration of the configuration system is not allowed to be changed by the user according to the business model, the configuration system cannot be reasonably configured for different business models, and the business model cannot be reasonably configured, resulting in insufficient or wasted computing power resources when the business model is running.

[0046] In the embodiments of the present application, the computing power test data of the business model is collected, and then a more reasonable configuration allocation strategy is determined based on the computing power test data and the self-supervised model, so that it is possible to configure the business model. Specifically, the method may, for example, include: first, collecting the computing power test data of the business model during running, then determining the configuration allocation strategy based on the computing power test data and the self-supervised model, and finally configuring the business model according to the configuration allocation strategy. The resources, services and models of the configured business model correspond to the resource configuration parameters, service configuration parameters and model configuration parameters respectively. It can be seen that the method provided by the present application can process the computing power test data of the business model with the trained self-supervised model, which can make the configuration allocation process for the business model more flexible, and can configure the computing power resources for the business model in a targeted and reasonable manner, thereby improving the utilization rate of computing power resources.

[0047] In order to facilitate understanding of the specific implementation of the method for determining the configuration of the business model provided by the embodiments of the present application, the following will be described in conjunction with the accompanying drawings.

[0048] It should be noted that the subject implementing the method for determining the configuration of the business model can be the system for determining the configuration of the business model provided by the embodiments of the present application, or the device for determining the configuration of the business model provided by the embodiments of the present application. The device for determining the configuration of the business model can be carried in an electronic device or a functional module of an electronic device. The electronic device in the embodiments of the present application can be any device capable of implementing the method for determining the configuration of the business model in the embodiments of the present application, for example, an Internet of Things (IoT) device.

[0049] Figure 1 A flowchart of a method for determining the configuration of a business model provided by the embodiments of the present application. The method can be applied to a system for determining the configuration of a business model, which may, for example, be Figure 2The system 200 for determining the configuration of the business model is shown; the method can also be applied to a device for determining the configuration of the business model, which can be, for example, as Figure 4 The device 400 for determining the configuration of the business model is shown, or the device for determining the configuration of the business model can also be integrated into Figure 5 The functional modules in the electronic device 500 are shown.

[0050] As shown in Figure 1 The method includes the following S101-S103:

[0051] S101: Collecting computing power test data of a business model, the computing power test data being used to embody relevant data of resources, services and models of the business model at runtime.

[0052] The business model is an artificial intelligence model used by a computer for business services. In order to rationalize the configuration allocation strategy of the business model, the computing power test data of the business model needs to be collected first, then the configuration allocation strategy of the business model is determined based on the computing power test data and a self-supervised model, and finally the business model is configured according to the configuration allocation strategy. Therefore, the present application embodiment collects the computing power test data of the business model through S101, thereby providing a precondition for obtaining the configuration allocation strategy.

[0053] When the business model is running, it will occupy resource data and service data of the computer, and the business model itself will also generate model data. Therefore, the computing power test data in S101 includes at least one of the above data. The resource data can include three parts: CPU data, memory data and GPU data. The service data can include two parts: resource data of stateless services and process data of stateless services.

[0054] The CPU data can include, for example, CPU load data and CPU occupancy rate data. The memory data can include, for example, total memory data, free memory data, remaining available memory data and occupancy rate data of the memory. The GPU data can include, for example, power data, temperature data, total memory data, free memory data, occupancy rate data and process data of the GPU.

[0055] The resource data of stateless services can include, for example, memory occupancy data and GPU occupancy data of the service. The process data of stateless services can include, for example, service call quantity data, service call time (computing time and network time) data and service call network traffic data.

[0056] The model data of the business model can include floating-point operations (FLOPs) of the business model, GPU memory occupation data of the business model at runtime, calculation time consumption data of the business model in a past period of time, and failure rate data of the business model.

[0057] S102: determining a configuration allocation strategy based on the computing power test data and a self-supervised model, the self-supervised model being a trained neural network model, and training samples of the self-supervised model including a plurality of historical computing power data and a training adjustment amount corresponding to each of the plurality of historical computing power data.

[0058] As an example, S102 can be preceded by determining whether the computing power test data is within a preset alert range, and if it is determined that the computing power test data is within the preset alert range, performing S102.

[0059] The computing power test data has a corresponding preset alert range. If the computing power test data is within the alert range when the business model is running, it indicates that the computing power resources used by the business model when not intervened by the embodiments of the present application are not reasonable, and the business model needs to be reconfigured by the method provided by the embodiments of the present application, so that the computing power resources allocated to the business model after intervention are more reasonable (i.e., the computing power test data of the business model after intervention is within the preset alert range). In this way, when the business model is subsequently run, the utilization rate of computing power resources by the business model is improved. In addition, if the computing power test data is not within the preset alert range, it indicates that the computing power resources used by the business model when not intervened by the embodiments of the present application are relatively reasonable, and the business model can not be intervened by the method provided by the embodiments of the present application, i.e., S102 can not be performed on the business model.

[0060] As an example, S102 can include inputting the computing power test data into the self-supervised model, obtaining an adjustment amount of each indicator in the computing power test data, and determining the configuration allocation strategy according to the adjustment amount of each indicator.

[0061] The training process of the self-supervised model includes obtaining a plurality of historical computing power data of a plurality of business models, performing mathematical calculations on the plurality of historical computing power data using a mathematical model to obtain a training adjustment amount corresponding to each of the plurality of historical computing power data, and training an initial model using the plurality of historical computing power data and the training adjustment amount corresponding to each of the plurality of historical computing power data as training samples to obtain the self-supervised model. The historical computing power data can be computing power test data of each business model at runtime saved in a database.

[0062] The specific process of training the initial model based on the training samples to obtain the self-supervised model may include, for example, inputting the historical computing power data 1 in the training sample 1 into the initial model, then comparing the output adjustment amount 1 of the initial model with the training adjustment amount 1 of the training sample 1 to obtain a comparison result 0, adjusting the initial model based on the comparison result 0 to obtain a self-supervised model 1; inputting the historical computing power data 2 in the training sample 2 into the self-supervised model 1, comparing the output adjustment amount 2 of the self-supervised model 1 with the training adjustment amount 2 of the training sample 2 to obtain a comparison result 1, adjusting the self-supervised model 1 based on the comparison result 1 to obtain a self-supervised model 2; and so on, until the obtained self-supervised model meets a preset condition, and then the self-supervised model when the preset condition is met is recorded as a trained self-supervised model (i.e., the self-supervised model in S102).

[0063] The algorithm used in the process of performing data calculation on the plurality of historical computing power data by using the mathematical model may include a back propagation algorithm.

[0064] S103: configuring the service model according to the configuration allocation strategy, and the resources, services and models of the configured service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0065] It should be noted that the computing power test data and the configuration allocation strategy are saved into the database. According to actual needs, the configuration allocation strategy can be periodically or other preset optimized, the self-supervised model is optimized and trained based on the data in the database, so as to update the configuration allocation strategy of the self-supervised model for different service models, so that the configuration of the service model can be better determined.

[0066] The computing power test data saved in the database is historical data, and the configuration allocation strategy corresponding to the computing power test data is a historical configuration allocation strategy. The historical configuration allocation strategy can be divided into different historical configuration allocation strategies according to different periods, so as to construct historical configuration allocation strategies of different time dimensions, so that the analysis of the historical configuration allocation strategy can be more intuitive.

[0067] It can be seen that the method of the embodiment of the application first collects computing power test data of a service model running, then determines a configuration allocation strategy based on the computing power test data and a self-supervised model, and finally configures the service model according to the configuration allocation strategy, so that the resources, services and models of the configured service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively, thereby realizing flexible determination of the configuration of the service model according to the data of the service model and performing corresponding operations, and improving the utilization rate of computing power resources of the service model running.

[0068] The system in the embodiment of the application may, for example, refer to Figure 2The system 200 for determining the configuration of the business model shown may include, for example:

[0069] The collection module 201 is configured to collect the computing power test data of the business model and transmit the computing power test data to the decision module 202.

[0070] The decision module 202 is configured to determine a configuration allocation strategy based on the computing power test data and the self-supervised model, and send a message containing the configuration allocation strategy to the execution module 203.

[0071] The execution module 203 is configured to configure the business model according to the configuration allocation strategy, and the resources, services and models of the configured business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0072] Optionally, the decision module 202 further includes:

[0073] The self-supervised model 2021 is configured to output the configuration allocation strategy based on the computing power test data.

[0074] Optionally, the decision module 202 can be further configured to:

[0075] Save the computing power test data and the configuration allocation strategy to a database, and the data in the database is used for optimizing training of the self-supervised model.

[0076] Classify the data in the database according to time dimension to obtain the classified configuration allocation strategy.

[0077] In order to make the method provided by the embodiments of the present application more clear and easy to understand, the following will be combined with Figure 3 The method is applied to Figure 2 A specific example of the system 200 for determining the configuration of the business model is shown.

[0078] As Figure 3 The embodiment may include, for example:

[0079] S301: The collection module 201 collects the first computing power test data of the business model and transmits it to the decision module 202, and the first computing power data is used to reflect the related data of the resources, services and models of the business model when running.

[0080] S302: The decision module 202 determines that when the first computing power test data is within a preset warning value range, S303 is performed.

[0081] The decision module receives the first computing power test data sent by the collection module, first judges whether the first computing power test data is within the preset warning value range, if not, it means that the configuration allocation of the business model belongs to a reasonable configuration allocation strategy, and there is no need to reconfigure; if yes, it means that the business model needs to be reconfigured, so that the subsequent configuration allocation reaches a reasonable configuration allocation strategy.

[0082] S303: The decision module 202 inputs the first computing power test data into the trained self-supervised model to obtain a first output adjustment amount, and determines a first configuration allocation strategy based on the first output adjustment amount.

[0083] In the training sample for training the self-supervised model, the historical computing power data is mathematically calculated by using a mathematical model to obtain a training adjustment amount corresponding to the historical computing power data. The mathematical model may be, for example, a plurality of linear models.

[0084] As an example, the training sample of the self-supervised model can include a plurality of historical computing power data collected by the collection module and a training adjustment amount corresponding to each of the plurality of historical computing power data. The plurality of output adjustment amounts obtained from the initial model are compared with the plurality of training adjustment amounts, and the initial model is repeatedly adjusted until the initial model meets a preset condition, and the initial model that meets the preset condition is recorded as the trained self-supervised model. The preset condition can include that the difference between the output adjustment amount and the training adjustment amount is less than 10%, or the training sample is used for one training.

[0085] For the trained self-supervised model, the first output adjustment amount corresponding to the first computing power test data can be approximately equal to the training adjustment amount obtained by mathematically calculating the first computing power test data using a mathematical model, which is a more reasonable basis for configuring and allocating the computing power of the business model.

[0086] S304: The decision module 202 saves the first computing power test data and the first configuration allocation strategy to the database.

[0087] In order to optimize the training of the self-supervised model, the computing power test data and the configuration allocation strategy need to be saved to the database. In addition, the configuration allocation strategies in the database can also be classified according to the time dimension to construct configuration allocation strategies of different time categories, so that the data can be intuitively analyzed.

[0088] S305: The decision module 202 transmits the first configuration allocation strategy to the execution module 203.

[0089] It should be noted that the execution order of S304 and S305 in the embodiments of the present application is not limited, which can be executed in S304 first and then in S305, or in S305 first and then in S304, or simultaneously in S304 and S305.

[0090] S306: The execution module 203 configures the service model according to the first configuration allocation strategy. After the configuration, the resources, services and models of the service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0091] The embodiments provide a method for determining configuration of a service model. First, the collection module 201 collects computing power test data of the service model and transmits the first computing power test data to the decision module 202. If the first computing power test data is within a preset warning value, the decision module 202 inputs the first computing power test data into the trained self-supervised model to obtain a first output adjustment amount and determine a first configuration allocation strategy based on the first output adjustment amount. The decision module 202 saves the first computing power test data and the first configuration allocation strategy to the database, so that the self-supervised model can be optimized and trained. The decision module 202 transmits the first configuration allocation strategy to the execution module 203, and the execution module 203 configures the service model according to the first configuration allocation strategy.

[0092] In this way, the embodiments of the present application collect computing power test data of the service model through the collection module 201, and then perform deep learning and adjustment training on the computing power test data through the self-supervised model in the decision module 202 to obtain a configuration allocation strategy, realize centralized processing and statistics of data, make the process of configuring and allocating the service model more flexible, reasonably configure computing power resources, improve the utilization rate of computing power resources, and after obtaining the configuration allocation strategy, trigger the corresponding operation work, automatically add service instances and register online, thereby realizing automatic expansion and reduction of corresponding resources.

[0093] Referring to Figure 4 The embodiments of the present application provide a device 400 for determining configuration of a service model, which comprises:

[0094] The collection unit 401 is configured to collect computing power test data of the service model, and the computing power test data is used to reflect the related data of the resources, services and models of the service model during running;

[0095] The decision unit 402 is configured to determine a configuration allocation strategy based on the computing power test data and a self-supervised model. The self-supervised model is a trained neural network model, and the training samples of the self-supervised model include a plurality of historical computing power data and a training adjustment amount corresponding to each historical computing power data in the plurality of historical computing power data.

[0096] The first processing unit 403 is configured to configure the service model according to the configuration allocation strategy, and the resources, services and model of the configured service model correspond to resource configuration parameters, service configuration parameters and model configuration parameters respectively.

[0097] Optionally, the apparatus 400 further includes:

[0098] The second processing unit is configured to determine whether the computing power test data is in a preset alert range, and if it is determined that the computing power test data is in the preset alert range, to execute the configuration allocation strategy based on the computing power test data and the self-supervised model.

[0099] Optionally, the decision unit 402 is specifically configured to:

[0100] The computing power test data is input into the self-supervised model to obtain an adjustment amount of each index in the computing power test data, and the configuration allocation strategy is determined according to the adjustment amount of each index.

[0101] Optionally, the apparatus 400 further includes:

[0102] The saving unit is configured to save the computing power test data and the configuration allocation strategy into a database, and the data in the database is used for optimizing training of the self-supervised model.

[0103] It should be noted that the specific implementation manner of the apparatus 400 and the effects achieved can be referred to the related descriptions in the method provided by the present application above, and will not be described here in detail. Figure 1 or Figure 3 The method provided by the present application.

[0104] The present application also provides an electronic device 500, as shown in the accompanying drawings, the device 500 includes a memory 501 and a processor 502: Figure 5

[0105] The memory 501 is configured to store a computer program.

[0106] The processor 502 is configured to execute the method provided by the present application according to the computer program. Figure 1 or Figure 3 The method provided by the present application.

[0107] In addition, the present application also provides a computer readable storage medium, which is used for storing a computer program, and the computer program is used for executing the method provided by the present application. Figure 1 or Figure 3 The method provided by the present application.

[0108] ​Those skilled in the art can clearly understand that all or part of the steps of the above-mentioned method in the embodiments can be implemented by means of software and a general hardware platform. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment or some parts of the embodiments of the present application.

[0109] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, they are described more simply, and the relevant parts can be referred to the part of the description of the method embodiments. The device embodiments described above are only schematic, and the modules shown as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the goal of the embodiments according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0110] The above-mentioned only is the exemplary embodiment of the present application, and is not used to limit the protection scope of the present application.

Claims

1. A method for determining configuration for a business model, characterized in that, The method includes: Collect computing power test data for the business model, which is used to reflect the resources, services and model-related data during the operation of the business model; If the computing power test data is determined to be within the preset warning range, then, based on the computing power test data and the self-supervised model, a configuration allocation strategy is determined. The self-supervised model is a trained neural network model, and the training samples of the self-supervised model include multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data in the multiple historical computing power data. The training process of the self-supervised model includes: obtaining multiple historical computing power data for multiple business models; performing mathematical calculations on the multiple historical computing power data using a mathematical model to obtain the training adjustment amount corresponding to each historical computing power data; and using the multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data as training samples to train the initial model and obtain the self-supervised model. The business model is configured according to the configuration allocation strategy. After configuration, the resources, services and models of the business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters, respectively.

2. The method according to claim 1, characterized in that, The step of determining the configuration allocation strategy based on the computing power test data and the self-supervised model includes: The computing power test data is input into the self-supervised model to obtain the adjustment amount of each indicator in the computing power test data; The configuration allocation strategy is determined based on the adjustment amount of each indicator.

3. The method according to claim 1 or 2, characterized in that, The method further includes: The computing power test data and the configuration allocation strategy are saved to a database, and the data in the database is used to optimize and train the self-supervised model.

4. The method according to claim 1 or 2, characterized in that, The computing power test data includes at least one of the following: resource data occupied by the business model during runtime, service data occupied by the business model during runtime, and model data of the business model.

5. A system for determining configurations for a business model, characterized in that, The system includes: a data acquisition module, a decision-making module, and an execution module. The acquisition module is used to collect computing power test data of the business model and transmit the computing power test data to the decision module. The computing power test data is used to reflect the relevant data of resources, services and models during the operation of the business model. The decision module is used to determine a configuration allocation strategy based on the computing power test data and the self-supervised model if it is determined that the computing power test data is within a preset warning range, and to send the configuration allocation strategy to the execution module. The self-supervised model is a trained neural network model, and the training samples of the self-supervised model include multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data in the multiple historical computing power data. The training process of the self-supervised model includes: obtaining multiple historical computing power data for multiple business models; performing mathematical calculations on the multiple historical computing power data using a mathematical model to obtain the training adjustment amount corresponding to each historical computing power data; and using the multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data as training samples to train the initial model and obtain the self-supervised model. The execution module is used to configure the business model according to the configuration allocation strategy. After configuration, the resources, services and model of the business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters, respectively.

6. An apparatus for determining configuration for a business model, characterized in that, The device includes: The data acquisition unit is used to collect computing power test data of the business model. The computing power test data is used to reflect the resources, services and model-related data during the operation of the business model. The decision unit is used to determine a configuration allocation strategy based on the computing power test data and the self-supervised model if it is determined that the computing power test data is within a preset warning range. The self-supervised model is a trained neural network model, and the training samples of the self-supervised model include multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data in the multiple historical computing power data. The training process of the self-supervised model includes: obtaining multiple historical computing power data for multiple business models; performing mathematical calculations on the multiple historical computing power data using a mathematical model to obtain the training adjustment amount corresponding to each historical computing power data; and using the multiple historical computing power data and the training adjustment amount corresponding to each historical computing power data as training samples to train the initial model and obtain the self-supervised model. The processing unit is used to configure the business model according to the configuration allocation strategy. After configuration, the resources, services and models of the business model correspond to resource configuration parameters, service configuration parameters and model configuration parameters, respectively.

7. An electronic device, characterized in that, The device includes a memory and a processor, the processor being configured to execute a program stored in the memory, performing the method as described in any one of claims 1-4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method according to any one of claims 1-4.

Images