Model installation method, device, electronic device, storage medium and product

Abstract

The present disclosure provides a model installation method, apparatus, electronic device, storage medium, and product, relating to the field of artificial intelligence, particularly cloud computing, deep learning, and computer vision. A specific implementation scheme comprises: in response to a user device sending a service acquisition request, obtaining target requirement information carried in the service acquisition request; determining a target model that matches the target requirement information; generating target installation information corresponding to the target model; and sending the target installation information to the user device, wherein the target installation information is used to instruct the user device to install the target model. The technical solution of the present disclosure improves the efficiency of model installation and maintenance.

Description

Technical Field

[0001] The present disclosure relates to the fields of cloud computing, deep learning, and computer vision in the field of artificial intelligence, and in particular to a model installation method, device, electronic device, storage medium, and product. Background Art

[0002] With the effective improvement of intelligent technology, mathematical models such as deep learning models and neural network models can be applied to modern industrial technology. For example, a defect detection production line built on an image detection model can detect defects on the surface of the object being inspected. However, in the existing model industrialization process, operation and maintenance personnel are required to train the model in the background to obtain a trained model. When users have industrial application needs, the trained business model is added to the business program on the user's electronic device based on the user's business needs, and the program can be used after debugging. However, this existing method requires operation and maintenance personnel to go to the user side to configure the model and program, and the efficiency of model configuration and replacement is relatively low. Summary of the Invention

[0003] The present disclosure provides a model installation method, device, electronic device, storage medium and product for improving model installation efficiency.

[0004] According to a first aspect of the present disclosure, there is provided a model installation method, comprising:

[0005] In response to a service acquisition request sent by a user equipment, acquiring target demand information carried in the service acquisition request;

[0006] Determining a target model that matches the target demand information;

[0007] generating target installation information corresponding to the target model;

[0008] The target installation information is sent to the user equipment, where the target installation information is used to instruct the user equipment to install the target model.

[0009] According to a second aspect of the present disclosure, a model installation method is provided, comprising:

[0010] Generate service acquisition request based on target demand information provided by target users;

[0011] Sending the service acquisition request to the electronic device; the service acquisition request instructs the electronic device to generate a corresponding target model according to the target demand information;

[0012] receiving target installation information corresponding to the target model sent by the electronic device;

[0013] Execute the target installation information to install the target model.

[0014] According to a third aspect of the present disclosure, there is provided a model installation system, comprising: an interface layer, a service layer connected to the interface layer;

[0015] The interface layer is used to: receive a service acquisition request sent by a user device;

[0016] The service layer is configured to: in response to the service acquisition request, obtain target demand information of a target user corresponding to the user device; determine a target model that matches the target demand information; and generate target installation information corresponding to the target model;

[0017] The interface layer is further configured to send the target installation information to the user equipment, wherein the target installation information controls the user equipment to install the target model.

[0018] According to a fourth aspect of the present disclosure, there is provided a model mounting device, comprising:

[0019] A first responding unit, configured to respond to a service acquisition request sent by a user equipment and acquire target demand information carried in the service acquisition request;

[0020] A model determination unit, configured to determine a target model that matches the target demand information;

[0021] an information generating unit, configured to generate target installation information corresponding to the target model;

[0022] A first sending unit is configured to send the target installation information to the user equipment, where the target installation information instructs the user equipment to install the target model.

[0023] According to a fifth aspect of the present disclosure, there is provided a model mounting device, comprising:

[0024] A request generation unit, configured to generate a service acquisition request based on target demand information provided by a target user;

[0025] a request sending unit, configured to send the service acquisition request to the electronic device; the service acquisition request instructs the electronic device to generate a corresponding target model according to the target requirement information;

[0026] a first receiving unit, configured to receive target installation information corresponding to the target model sent by the electronic device;

[0027] The first installation unit is configured to execute the target installation information to install the target model.

[0028] According to a sixth aspect of the present disclosure, there is provided an electronic device, including:

[0029] at least one processor; and

[0030] a memory communicatively connected to the at least one processor; wherein,

[0031] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to perform the method according to the first aspect or the second aspect of claim 1.

[0032] According to an eighth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method according to the first aspect or the second aspect.

[0033] According to a ninth aspect of the present disclosure, a computer program product is provided, comprising: a computer program, wherein the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program so that the electronic device executes the method described in the first aspect or the second aspect.

[0034] The technology disclosed in the present invention solves the problem of low efficiency in model configuration and replacement. The electronic device can obtain target demand information from the service acquisition request sent by the user device. Then, the target model that matches the target demand information is determined. The target model is automatically acquired to generate target installation information corresponding to the target model. The target installation information can be sent to the user device, and the user device can run the target installation information to complete the installation of the target model. By using the target demand information to automatically acquire the target model and automatically generate the target installation information, the installation efficiency of the target model can be improved, and the efficiency of using intelligent industrialization technology can be effectively improved.

[0035] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0037] Figure 1 This is a schematic diagram of an application scenario applied to the model installation method provided according to the first embodiment of the present disclosure;

[0038] Figure 2is a flowchart of a model installation method provided according to the second embodiment of the present disclosure;

[0039] Figure 3 is a flowchart of a model installation method provided according to the third embodiment of the present disclosure;

[0040] Figure 4 is a flowchart of a model installation method provided according to the fourth embodiment of the present disclosure;

[0041] Figure 5 is a flowchart of a model installation method provided according to the fifth embodiment of the present disclosure;

[0042] Figure 6 is a structural diagram of a model installation system provided according to a sixth embodiment of the present disclosure;

[0043] Figure 7 is a structural diagram of a model installation system provided according to a seventh embodiment of the present disclosure;

[0044] Figure 8 is a structural diagram of a model installation system provided according to an eighth embodiment of the present disclosure;

[0045] Figure 9 is a structural schematic diagram of a model installation device provided according to a ninth embodiment of the present disclosure;

[0046] Figure 10 is a structural schematic diagram of a model installation device provided according to a tenth embodiment of the present disclosure;

[0047] Figure 11 It is a block diagram of an electronic device used to implement the model installation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0048] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0049] The technical solution disclosed in the present invention can be applied to intelligent industrialization scenarios. By obtaining the demand information provided by users according to their business scenarios, the target model is automatically matched to automatically generate corresponding target installation information, and the automatic establishment of installation information according to user needs is realized, thereby improving the efficiency of model establishment and installation, and facilitating the rapid promotion of intelligent industrial technology.

[0050] In related technologies, mathematical models such as visual models and deep learning models can be applied in intelligent industrialization scenarios. However, in specific applications, operations and maintenance personnel are required to train the models offline. Once trained, the models can be used in industrial technology scenarios. However, in industrial technology applications, operations and maintenance personnel must visit users offline and deploy the models online on their devices. This process requires model installation and debugging, resulting in poor model deployment efficiency.

[0051] The disclosed technical solution, designed to automate model rollout, can send target requirement information provided by users based on their business scenarios to backend electronic devices. These devices can then directly orchestrate and train models based on these target requirement information, generating installation information. This automates the entire process from information acquisition to model deployment, improving model installation efficiency.

[0052] The present disclosure provides a model installation method, system, device, electronic device, storage medium and product, which are applied to cloud computing, deep learning and computer vision in the field of artificial intelligence, so as to improve the efficiency of model installation in industrial intelligence scenarios and realize automatic model installation.

[0053] For ease of understanding, the technical solution of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0054] like Figure 1 As shown in FIG. 1 , a schematic diagram of an application scenario of the model installation method provided by the first embodiment of the present disclosure is provided. The application scenario takes the computer vision scenario as an example, and the computer vision model can be configured in the defect detection production line. The user device 1 can establish a wired or wireless communication connection with the electronic device 2. The user device 1 is, for example, Figure 1 Computers, electronic devices such as Figure 1 Before the user device 1 is used for defect detection, a service acquisition request may be generated based on target demand information provided by the target user, and the service acquisition request may be sent to the electronic device 2.

[0055] Electronic device 2 can obtain target demand information from the service acquisition request and obtain the target model from its corresponding model installation module. The target model can be obtained through model scheduling and model training. Afterwards, electronic device 2 can generate target installation information corresponding to the target model. The target installation information can be generated after the target model is registered with the service and the target resources are acquired. Electronic device 2 can send the target installation information to user device 1. User device 1 can install the corresponding target model based on the received target installation information. The target model can be a computer vision model. It is assumed that the computer vision model can specifically include several models such as a feature extraction model, a defect area recognition model, a local feature extraction model, and a defect classification model.

[0056] After the target model is installed, user device 1 can obtain the image to be inspected, captured by camera 4 of object 5 on the inspection frame, and use the target model to perform defect detection on the image to be inspected, obtaining a defect detection result. User device 2 can also display the image with the defect or the product number corresponding to the image.

[0057] It should be noted that in addition to being applied to industrial technology application scenarios of computer vision, the technical solution of the present disclosure can also be applied to other types of technical application scenarios, such as financial scenarios, intelligent teaching, intelligent navigation, aerospace, intelligent driving cockpits, vehicle networks, electronic maps, self-driving vehicles, smart home appliances, wearable devices, smart appliances and many other technical fields. Its main application principle is to provide users with the installation information of the corresponding target program according to user needs in accordance with the technical principle provided by the present disclosure, realize the intelligent and automated application of mathematical models, and improve the efficiency of model installation and maintenance.

[0058] like Figure 2 FIG. 1 is a flowchart of a model installation method provided by the second embodiment of the present disclosure, which may include the following steps:

[0059] 201: In response to a service acquisition request sent by a user equipment, obtain target requirement information carried in the service acquisition request.

[0060] The service acquisition request may include target requirement information, which may refer to requirement information set by the target user according to the industrial use target of the mathematical model.

[0061] The service acquisition request may be generated based on the target demand information provided by the target user.

[0062] 202: Determine a target model that matches the target demand information.

[0063] The target model can include a deep learning model, a neural network model, or a computer vision model.

[0064] 203: Generate target installation information corresponding to the target model.

[0065] The target installation information may correspond to the target model.

[0066] 204: Send target installation information to the user equipment, where the target installation information is used to instruct the user equipment to install the target model.

[0067] The target installation information may include the target model's installation package, installation link, and other information.

[0068] In an embodiment of the present disclosure, an electronic device can obtain target demand information from a service acquisition request sent by a user device. A target model that matches the target demand information can then be determined. Automated acquisition of the target model is achieved to generate target installation information corresponding to the target model. The target installation information can be sent to the user device, which can execute the target installation information to complete the installation of the target model. Automatically acquiring the target model and automatically generating the target installation information using the target demand information can improve the installation efficiency of the target model and effectively improve the efficiency of using intelligent industrialization technology.

[0069] In order to enable readers to have a deeper understanding of the implementation principle of this disclosure, the following Figure 3-Figure 4 right Figure 2 The illustrated embodiment is further refined.

[0070] In order to obtain an accurate target model, such as Figure 3 FIG. 2 is a flowchart of a model installation method provided by the third embodiment of the present disclosure. Step 202: determining a target model that matches the target requirement information may include the following steps:

[0071] 301: Perform demand analysis on target demand information to obtain target model information including at least one demand processing node.

[0072] The demand analysis of target demand information can refer to the analysis of the demand side to accurately control the usage needs of the target users, avoid missing needs, and achieve accurate analysis of user needs.

[0073] Optionally, target demand information can be collected in a predefined information format. According to the information format of the target demand information, the target demand information is subjected to demand analysis to obtain at least one demand processing node that matches the target demand information. The demand processing nodes can be divided according to the target demand information and can be used to implement a processing function. Different demand processing nodes have a connection relationship. At least one demand processing node is connected according to their respective processing orders, with the first demand processing node corresponding to input data and the last demand processing node corresponding to output data. Taking a computer vision scenario as an example, assume that the scenario is defect detection on a collected image to be processed. The target demand information can be defined as input content: image and processing target: defect detection. At least one demand processing node can be obtained by demand analysis for defect detection on data input as an image, and can include: an image preprocessing node, a feature extraction node, a target area detection node, a classification node, etc. Each node corresponds to a corresponding processing order, for example, the image preprocessing node is located before the feature extraction node, and the classification node is located last.

[0074] 302: Determine a matching target model according to the target model information.

[0075] In addition to at least one required processing node, the target model information may also include an input data type and an output data type. The target model definition and training may be completed based on the at least one required processing node, the input data type, and the output data type in the target model information to obtain the final target model.

[0076] In an embodiment of the present disclosure, by performing demand analysis on the target demand information, target model information of at least one demand processing node can be obtained, and the target model can be determined according to at least one demand processing node in the target model information, so that the target model matches at least one demand processing node in the target demand information, thereby improving the acquisition efficiency and accuracy of the target model.

[0077] In order to obtain the target model, the target model can be trained. Step 302: Determine the matching target model based on the target model information, which may include the following steps:

[0078] If it is determined that the model update conditions are met, the corresponding model to be trained is constructed according to the target model information;

[0079] Get the training data for the model to be trained;

[0080] The training data is used to train the training model to obtain a matching target model.

[0081] The model to be trained can be a comprehensive mathematical model with unknown parameters and can be composed of one or more models. The training data for the model to be trained can be determined based on the input and output data types in the target model information. The training data can include data and labels. The data can be determined based on the input data type, and the label can be the true value of the data, which can be obtained through annotation. Specific data annotation methods can be referenced in related art and will not be elaborated here.

[0082] Using the training data to train the model to be trained and obtaining a matching target model may include: setting model parameters for the model to be trained, taking the data prediction results of the training data corresponding to the model parameters of the trained model to be the same as their corresponding labels as the training goal, and obtaining the target model corresponding to the model parameters that meet the training goals. The specific training method of the model to be trained can refer to the model training method in the relevant technology, which will not be repeated here.

[0083] In the disclosed embodiments, the determination of model update conditions can be used to make training decisions for the target model, avoiding overtraining and improving model utilization. When the model update conditions are met, the corresponding to-be-trained model is constructed. The to-be-trained model is trained by acquiring its model data, resulting in an accurate target model.

[0084] In one possible design, after training the target model using the training data, the following steps are also included:

[0085] Generate model version information for the target model obtained through training;

[0086] The target model and model version information are stored in association.

[0087] After training and obtaining the target model, you can generate model version information for the target model to manage its version. For example, you can regularly update the model and replace the updated target model with the latest one. You can also query the model to perform historical queries on the target model, improving model management efficiency.

[0088] Optionally, when generating model version information for the target model obtained through training, the identification generation principle of the blockchain can be used to generate unique model version information for the target model to avoid duplication of model versions and improve the security of target model version management.

[0089] In this public embodiment, after obtaining the target model, model version information can be generated for the target model obtained through training, so that the target model and the model version information are associated and stored, and version management of the target model is achieved through the model version information, thereby improving the management efficiency of the target model.

[0090] As an optional implementation, building a corresponding model to be trained based on the target model information may include the following steps:

[0091] Read at least one demand processing node in the target model information;

[0092] Determine at least one candidate mathematical model; the candidate mathematical model corresponds to a processing function;

[0093] Determining a target mathematical model whose processing function matches the required processing node from at least one candidate mathematical model to obtain a target mathematical model corresponding to at least one required processing node;

[0094] The target mathematical models corresponding to at least one demand processing node are combined to obtain a model to be trained.

[0095] The processing function of the candidate mathematical model is known. For example, model description information can be set for the candidate mathematical model, and the model description information can be used to describe the processing function of the model. The matching of the processing function of the candidate model with the required processing node can refer to the candidate model. The model description information can be used to calculate information similarity with the requirement information of the required processing node. If certain information similarity conditions are met, it is determined that the processing function of the corresponding candidate mathematical model matches the required processing node.

[0096] Optionally, if the target mathematical model corresponding to any demand processing node includes one or more, the method may further include: sending at least one target mathematical model of the demand processing node to the user device, and obtaining the target mathematical model determined by the target user for the target demand processing node sent by the user device.

[0097] In the disclosed embodiments, the required processing nodes in the target model information are obtained, and candidate mathematical models with processing capabilities are retrieved according to each required processing node, achieving accurate acquisition of each required processing node. By combining the target mathematical models corresponding to at least one required processing node, a training model can be obtained. The obtained training model matches the model requirements and achieves higher accuracy.

[0098] In another possible design, obtaining training data for the model to be trained includes:

[0099] According to the target model information, determine the input data type and output data type corresponding to the target model;

[0100] According to the input data type and the output data type, the training data of the model to be trained is determined; the training data includes target training data corresponding to the input data type and data labels corresponding to the output data type.

[0101] The data type of the target training data is the input data type. The data labels for the target training data are obtained through annotation, and their data type is consistent with the output data type. For example, in an industrialized computer vision scenario, the input data type can be an image type, and the output data type can be an integer type between 0 and 1. 0 represents the presence of a defect, and 1 represents the absence of a defect.

[0102] In the embodiment of the present disclosure, the input data type and output data type of the target model information can be determined to determine the training data for the model to be trained according to the input data type and the output data type. This can ensure that the training data includes target training data corresponding to the input data type and data labels corresponding to the output data type, thereby enabling accurate acquisition of the training data.

[0103] In some embodiments, combining target mathematical models corresponding to at least one demand processing node to obtain a model to be trained may include:

[0104] Determining target hyperparameters of a target mathematical model corresponding to at least one demand processing node based on a hyperparameter adjustment algorithm;

[0105] According to target hyperparameters corresponding to the at least one target mathematical model, the at least one target mathematical model is combined according to the association relationship of the at least one demand processing node to obtain a model to be trained.

[0106] A hyperparameter tuning algorithm may be an algorithm for adjusting parameters, and the parameter tuning may be used to accurately estimate the data.

[0107] In the embodiment of the present disclosure, the target hyperparameters of the target mathematical model are determined by a hyperparameter adjustment algorithm, so that the target hyperparameters are used to establish the model to be trained, thereby achieving accurate establishment of the model.

[0108] Before training the target model, the model update conditions may be determined. As an embodiment, before training the target model, the method may further include:

[0109] If it is determined that the model update condition is not met, the target model version information corresponding to the target model information is obtained to read the target model corresponding to the target model version information; the target model version information is pre-associated with the corresponding target model and stored.

[0110] If the model update conditions are not met, the model may not be trained. In this case, to provide accurate feedback on the target model, the corresponding historical version can be queried from the historical versions of the target model. The target model version information can be determined from the historical versions, and the target model with the target model version information can be used for model feedback. The target model version information can be model version information of any historical version. The model version information may include, for example, a version number. The historical version corresponding to the latest version number can be used as the target model version information.

[0111] In the embodiment of the present disclosure, when the model update condition is not met, the target model version information corresponding to the target model information can be obtained to read the target model corresponding to the target model version information. The method of directly obtaining the target model using the target model version information is a method of directly reading the target model, which can quickly and accurately obtain the target model.

[0112] In order to accurately determine the model update condition, as an embodiment, the method may further include:

[0113] Get the model update frequency of the model to be trained and the end time of the most recent training;

[0114] Determine the update start time of the model to be trained based on the model update frequency and end time;

[0115] When the current time reaches the update start time, it is determined that the model update condition is met.

[0116] In the disclosed embodiments, the model update frequency and the end time of the most recent training run of the model to be trained can be obtained. By obtaining the update start time based on the model update frequency and end time, the model update condition is determined to be satisfied when the current time reaches the update start time. By detecting the update start time, the update start time can be accurately detected, improving the timeliness of determining the model update condition and achieving effective judgment.

[0117] like Figure 4 FIG. 2 is a flowchart of a model installation method provided by the third embodiment of the present disclosure. 203: generating target installation information corresponding to the target model may include the following steps:

[0118] 401: Determine the registration information and target resource of the target model.

[0119] 402: Generate a target application of the target model based on the registration information and target resources of the target model and the target model; the registration information is used to instruct the target application to register for service in the user device; the target resources are used for the user device to call when using the target model.

[0120] Generating a target application for the target model based on the target model's registration information and target resources in combination with the target model may include encapsulating the target model, registration information, target resources, and other information to obtain the target application. In certain embodiments, a model interface for the target model may be determined based on an interface generation strategy, and the model interface, target model, registration information, and target resources may be encapsulated to obtain the target application. Examples of program types for the target application include a software development kit (SDK), an Android application package (APK), and the like.

[0121] The target model interface is determined based on the target generation strategy. For example, corresponding input controls can be generated based on the target model's data input interface, and corresponding output controls can be generated based on the target model's data output interface. The model interface is generated using the input controls and output controls. Furthermore, the model interface can also include other controls or pages or sub-pages related to the target model. The generation strategy for specific pages can be pre-set. For example, a model interface generation script can be preset. When a model page needs to be generated, the interface generation script can be called, the target model can be input into the interface generation script, and the script can be run to automatically generate the model interface.

[0122] 403: Determine target installation information corresponding to the target model according to the target application.

[0123] The registration information may include a registration instruction generated based on the relevant information required for registration of the target model. The target resource may include resources composed of various types of hardware, including at least one of a hard disk, a server, a central processing unit, and a graphics processing unit.

[0124] Optionally, determining target installation information corresponding to the target model based on the target application may include generating a link or download address for the target application that can be used to obtain the target installation information for the target application. The user device may download the target application based on the target installation information and complete the installation of the target model, thereby using the target model.

[0125] In the disclosed embodiments, when generating target installation information, the target model's registration information and target resources can be used in conjunction with the target model to generate a corresponding target application. The registration information can instruct the target application to register for services on the user device, and the target resources can be used by the user device to access the target model. By utilizing the target resources and registration information, the generated target application can be accurately indicated, improving the target application's registration efficiency and operational stability, ensuring the target application's normal operation.

[0126] As an embodiment, the target requirement information includes resource requirements; determining the registration information of the target model and the target resources includes:

[0127] Schedule target resources for the target model according to resource requirements;

[0128] According to the target resource, the target model is registered as a service to obtain the registration information of the target model.

[0129] Resource requirements can be defined by processing metrics, such as the total amount processed per second, the query rate per second (QPS) of each node, etc.

[0130] Registering the target model for service according to the target resource may include generating a registration instruction based on the target resource, the registration instruction being generated according to a certain instruction format, running a registration execution, registering the target model for service, and obtaining registration information of the target model.

[0131] In the disclosed embodiments, target resources can be scheduled for a target model based on resource demand, achieving resource-based scheduling, improving accurate scheduling of target resources, and avoiding invalid, reduced, or reduced scheduling. The target model can be registered with the target resource to obtain registration information. By registering the target model, automatic registration of the target resource can be achieved, improving initial testing efficiency.

[0132] In order to obtain accurate target resources, as an optional implementation method, target resources are scheduled for the target model according to resource requirements, including:

[0133] Obtaining at least one resource node and resource usage information of at least one resource node;

[0134] Determining, based on the resource usage information, a candidate resource node in an idle state from at least one resource node;

[0135] Select a target resource node from candidate resource nodes according to resource demand;

[0136] A usage cycle of a target model is established for a target resource node to obtain a target resource corresponding to the target model.

[0137] Hardware resources can be divided into at least one resource node to facilitate efficient management of the hardware resources. The resource amount of each resource node is known. The resource usage information can be node identification information of an occupied resource node among the at least one resource node. Based on the resource usage information, the occupied resource node can be determined from the at least one resource node, and unoccupied resource nodes can be obtained. In other words, resource nodes in an idle state are selected as candidate resource nodes.

[0138] The step of selecting a target resource node from the candidate resources according to resource demand may include: selecting the target resource node according to the resource demand of the candidate resources. Specifically, the number of candidate resource nodes may be calculated based on the resource demand and the resource quantity, and the corresponding target resource node may be selected according to the number of nodes.

[0139] In the disclosed embodiments, when scheduling a target resource, an idle candidate resource node can be identified from at least one resource node. A target resource node can be selected from the candidate resource nodes based on resource demand, thereby achieving quantity-based selection of the target resource node. By establishing a target model's usage cycle for the target resource node, the target resource corresponding to the target model can be obtained, enabling accurate and efficient acquisition of the target resource.

[0140] In practical applications, the target model can be updated in real time. In the above embodiment, after generating the target installation information and sending the target installation information to the user device, the method may further include:

[0141] Receive a model optimization request sent by a user device.

[0142] In response to the model optimization request, new target requirement information corresponding to the model optimization request is obtained, and the process returns to determine the target model that matches the target requirement information and continues to execute, thereby obtaining the latest target installation information.

[0143] Send the latest target installation information to the user device. The latest target installation information may instruct the user device to install the latest target model.

[0144] A model optimization request can be obtained by a model update control triggered by a user device detecting a target user. The model optimization request can include new target requirement information. After obtaining the new target requirement information, execution can return to determining a target model that matches the target requirement information and continue. The target requirement information can be new target requirement information. The new target requirement information replaces the original target requirement information, and the target model acquisition step is re-executed.

[0145] In embodiments of the present disclosure, upon receiving a model optimization request from a user device, the system can respond to the model optimization request to obtain new target requirement information corresponding to the model optimization request, thereby updating the target model according to the target requirement information and obtaining the latest target installation information. By responding to the model optimization request, the target installation information can be updated, achieving efficient and accurate updating of the target model.

[0146] In some embodiments, the method may further include:

[0147] Detect the login request sent by the user's device and obtain the login information provided by the target user;

[0148] Authenticate the target user based on the login information and obtain the verification result;

[0149] If the verification result is determined to be passed, a service acquisition request sent by the user equipment is received.

[0150] The login request may include the user's login information. The login information may include user name, password, and verification code. The user name and password may be obtained through registration.

[0151] In the disclosed embodiments, a login request sent by a user device can be detected, login information provided by the target user can be obtained, and the target user's identity can be authenticated based on the login information to obtain a corresponding verification result. The verification result is the result of the target user's identity authentication. If the verification passes, the user's safe use can be ensured. By verifying the user's identity, the security of the model system can be ensured.

[0152] As an embodiment, after generating the target installation information corresponding to the target model, the method may further include:

[0153] Establishing an installation version number for the target installation information, and storing the installation version number in association with the target installation information;

[0154] Obtain demand information provided by other user devices;

[0155] If it is determined that the requirement information provided by the other user devices matches the target requirement information, the target installation information corresponding to the latest installation version number is sent to the other user devices.

[0156] The demand information provided by other user equipment may be obtained and defined in the same manner as the target demand information, and will not be described in detail here.

[0157] The requirement information provided by other user equipment matches the target requirement information, which may include: the requirement information provided by other user equipment is identical to the target requirement information, or the similarity is higher than a certain similarity threshold.

[0158] Optionally, when generating an installation version number for the target installation information obtained through training, the identification generation principle of the blockchain can be used to generate a unique installation version number for the target installation information to avoid duplication of installation versions and improve the management efficiency and security of the installation information.

[0159] In the disclosed embodiment, an installation version number can be established for the target installation information. The installation version number is associated with the target installation information and stored. When obtaining the demand information provided by other devices, if it is determined that the demand information provided by other user devices matches the target demand information, the target installation information corresponding to the latest installation version number can be sent to the other user devices. The target installation information can be effectively managed through the installation version number. When there is demand information from other user devices, the latest version of the target installation information can be directly fed back to the user, realizing the reuse of the target installation information, improving the utilization rate of the target installation information, and effectively improving the user's installation efficiency.

[0160] like Figure 5 FIG. 1 is a flowchart of a model installation method according to a fifth embodiment of the present disclosure, which may include the following steps:

[0161] 501: Generate a service acquisition request based on the target demand information provided by the target user.

[0162] 502: Send a service acquisition request to the electronic device; the service acquisition request instructs the electronic device to generate a corresponding target model according to the target requirement information.

[0163] 503: Receive target installation information corresponding to the target model sent by the electronic device.

[0164] 504: Run the target installation information to install the target model.

[0165] Optionally, executing the target installation information to install the target model may include: detecting an installation instruction triggered by the target user, and executing the target installation information to install the target model in response to the installation instruction.

[0166] In an embodiment of the present disclosure, a user device may generate a service acquisition request based on the target demand information provided by the target user, thereby sending a service acquisition request to the electronic device. The service acquisition request may instruct the electronic device to generate a corresponding target model according to the target demand information, and the target model generates corresponding target installation information. The user device receives the target installation information sent by the electronic device, that is, it may run the target installation information and install the target model. Automatic acquisition of the target installation information is achieved without excessive user participation, which can improve the acquisition efficiency and installation efficiency of the target model.

[0167] As an embodiment, after executing the target installation information to install the target model, the following steps are further included:

[0168] Detect model optimization requests initiated by target users;

[0169] Sending a model optimization request to the electronic device, where the model optimization request controls the electronic device to update the target model;

[0170] receiving the latest target installation information sent by the electronic device;

[0171] Run the latest target installation information to install the latest target model.

[0172] In the disclosed embodiment, after the target model is installed, a model optimization request initiated by the target user may be detected and sent to the electronic device. The model optimization request may instruct the electronic device to update the target model. The latest target installation information is fed back to the user device. The user device may execute the received latest target installation information to update the target model. By initiating a model optimization request by the user, the corresponding target installation information may be automatically obtained, thereby improving the efficiency of obtaining target installation information.

[0173] like Figure 6 FIG. 6 is a schematic diagram of a structure of a model installation system provided by the sixth embodiment of the present disclosure. The model installation system may include an interface layer 601 and a service layer 602 connected to the interface layer 601 .

[0174] The interface layer 601 may be used to: receive a service acquisition request sent by a user equipment;

[0175] The service layer 602 may be configured to: obtain target demand information of a target user corresponding to a user device in response to a service acquisition request; determine a target model that matches the target demand information; and generate target installation information corresponding to the target model;

[0176] The interface layer 601 may also be used to send target installation information to the user equipment, where the target installation information controls the user equipment to install the target model.

[0177] In the disclosed embodiment, the interface layer receives service acquisition requests, responds to them through the service layer, and obtains corresponding target installation information. The interface layer can also send the target installation information to the user device. This layered collaboration between the interface layer and the service layer improves information processing efficiency and effectively enhances the generation and installation of target models.

[0178] In one possible design, Figure 7 As shown, the model installation system may include, in addition to the interface layer 701 and the service layer 702 connected to the interface layer 701 , a resource layer 703 connected to the service layer.

[0179] The resource layer 703 can be used to: determine the registration information and target resources of the target model;

[0180] The service layer 702 can also be used to: generate a target application of the target model based on the registration information and target resources of the target model, combined with the target model; the registration information is used to control the target application to register for service in the user device; the target resource is used for the user device to call when using the target model; according to the target application, determine the target installation information corresponding to the target model.

[0181] The resource layer 703 can provide users with various resources, such as data storage resources 7031. Data storage resources can include cache, persistent storage, structured storage, etc. Of course, the resource layer can also provide other processing resources, such as server resources, hard disk resources, etc., which will not be described in detail here.

[0182] In the embodiment of the present disclosure, the resource layer can be used to determine the registration information and target resources of the target model. The target model can be registered and resources can be allocated through the resource layer. Through the setting of the resource layer, the resources within the system can be effectively managed to improve resource management efficiency.

[0183] In order to enable readers to have a deeper understanding of the implementation principle of this disclosure, the following Figure 8 right Figure 6 and Figure 7 The illustrated embodiment is further refined.

[0184] Furthermore, in one possible design, if Figure 8 As shown, the model installation system includes, in addition to an interface layer 801, a service layer 802 connected to the interface layer 801, and a resource layer 803 connected to the service layer, an access layer 804. The access layer 804 can be a network layer for user devices to interact with the interface. The resource layer 803 can include data storage resources 8031.

[0185] like Figure 8 As shown, the access layer 804 may include user devices accessed via a web network. These user devices using the web network may configure at least one of a service call interface 8041, a remote call interface 8042, and a network call interface 8043. The access layer 804 may also include user devices accessed via a client. The user devices accessed via a client may include a software development kit 8044. The access layer may also include a request module 8045 for optimizing or testing the model.

[0186] When a user device interacts with an electronic device via a web network, the electronic device can provide a service call interface, a remote call interface, or a network call interface to the user device. A service acquisition request can be initiated by the user device through the service call interface, remote call interface, or network call interface provided by the electronic device.

[0187] When a user device interacts with an electronic device through a client, the electronic device can provide a service call to the user device through the client packaged by the software development kit. The service acquisition request can be initiated by the user device through the client. The client controls the user device to send the service request to the electronic device.

[0188] Of course, in actual applications, in order to facilitate the management of each module and improve management efficiency, the service layer 802 can be divided into different modules or units. Figure 8 , the service layer 802 may include: a model determination unit 8021. The model determination unit may include a demand analysis module 805 and a model orchestration module 806. Among them, the model orchestration module 806 may include a model orchestration sub-module 8061, a training acquisition sub-module 8062 and a specific model training module 8063. Of course, on this basis, the model orchestration sub-module 8061 may include modules such as hyperparameter adjustment, model management, model transmission, and model pipeline. The training acquisition sub-module 8062 may include modules such as training data preprocessing, data labeling, and sample data selection. In addition, it may also include modules such as model self-iteration, model storage, version management, termination training, model tuning, and accelerated training to achieve accurate training of the model. For the specific contents executed by each module and sub-module, please refer to the embodiment of the device in the figure below, which will not be repeated here.

[0189] In addition, in actual applications, the service layer 802 can also be used to determine a data transmission protocol with the user device, specifically selecting from candidate data transmission protocols. When the electronic device and the user device transmit data or information, the data is encapsulated according to the data transmission protocol determined by the service layer to achieve accurate data transmission.

[0190] The model installation system provided by the present disclosure can execute the above-mentioned model installation method through various levels. For the specific content of each step executed at each level, please refer to the above description of the model installation method, which will not be cited one by one here.

[0191] like Figure 9 FIG. 9 is a schematic diagram of a model installation device according to a seventh embodiment of the present disclosure. The model installation device 900 may include:

[0192] The first responding unit 901 is configured to respond to a service acquisition request sent by a user equipment and acquire target requirement information carried in the service acquisition request.

[0193] Model determination unit 902: used to determine a target model that matches the target requirement information.

[0194] Information generating unit 903: used to generate target installation information corresponding to the target model.

[0195] The first sending unit 904 is configured to send target installation information to the user equipment, where the target installation information instructs the user equipment to install the target model.

[0196] As an embodiment, the model determination unit includes:

[0197] The demand analysis module is used to perform demand analysis on target demand information and obtain target model information including at least one demand processing node.

[0198] The model arrangement module is used to determine the matching target model based on the target model information.

[0199] In some embodiments, the model orchestration module includes:

[0200] The model arrangement submodule is used to build the corresponding to-be-trained model according to the target model information if it is determined that the model update conditions are met.

[0201] The training acquisition submodule is used to obtain training data for the model to be trained.

[0202] The model training submodule is used to train the to-be-trained model using the training data to obtain a matching target model.

[0203] As a possible implementation method, it also includes:

[0204] The version acquisition unit is used to generate model version information for the target model obtained through training.

[0205] The first storage unit is used to store the target model and the model version information in association with each other.

[0206] In some embodiments, the model arrangement submodule is specifically configured to:

[0207] At least one demand processing node in the target model information is read.

[0208] At least one candidate mathematical model is determined; the candidate mathematical model corresponds to a processing function.

[0209] Determining a target mathematical model whose processing function matches the required processing node from at least one candidate mathematical model to obtain a target mathematical model corresponding to at least one required processing node;

[0210] The target mathematical models corresponding to at least one demand processing node are combined to obtain a model to be trained.

[0211] As a possible implementation method, the model orchestration submodule is specifically used to:

[0212] Determining target hyperparameters of a target mathematical model corresponding to at least one demand processing node based on a hyperparameter adjustment algorithm;

[0213] According to target hyperparameters corresponding to the at least one target mathematical model, the at least one target mathematical model is combined according to the association relationship of the at least one demand processing node to obtain a model to be trained.

[0214] As an optional implementation, the training acquisition submodule is specifically used to:

[0215] According to the target model information, determine the input data type and output data type corresponding to the target model;

[0216] According to the input data type and the output data type, the training data of the model to be trained is determined; the training data includes target training data corresponding to the input data type and data labels corresponding to the output data type.

[0217] In some embodiments, further comprising:

[0218] The execution unit is returned to obtain the target model version information corresponding to the target model information if it is determined that the model update condition is not met, so as to read the target model corresponding to the target model version information; the target model version information is pre-associated with the corresponding target model and stored.

[0219] As yet another embodiment, further comprising:

[0220] The time acquisition unit is used to obtain the model update frequency of the model to be trained and the end time of the most recent training;

[0221] A time update unit, used to determine the update start time of the model to be trained based on the model update frequency and end time;

[0222] The time detection unit is used to detect when the current time reaches the start time and determine whether the model update condition is met.

[0223] As yet another embodiment, the information generating unit includes:

[0224] Information acquisition module, used to determine the registration information and target resources of the target model;

[0225] The program generation module is used to generate a target application of the target model based on the registration information and target resources of the target model and the target model; the registration information is used to instruct the target application to register the service in the user device; the target resources are used to be called when the user device uses the target model;

[0226] The information determination module is used to determine the target installation information corresponding to the target model according to the target application.

[0227] In some embodiments, the target demand information includes resource demand; the information acquisition module includes:

[0228] The resource acquisition submodule is used to schedule target resources for the target model according to resource demand;

[0229] The model registration submodule is used to register the target model service according to the target resource and obtain the registration information of the target model.

[0230] In some embodiments, the resource acquisition submodule is specifically configured to:

[0231] Obtaining at least one resource node and resource usage information of at least one resource node;

[0232] Determining, based on the resource usage information, a candidate resource node in an idle state from at least one resource node;

[0233] Select a target resource node from candidate resource nodes according to resource demand;

[0234] A usage cycle of a target model is established for a target resource node to obtain a target resource corresponding to the target model.

[0235] As yet another embodiment, further comprising:

[0236] An optimization receiving unit, configured to receive a model optimization request sent by a user device;

[0237] The request optimization unit is used to respond to the model optimization request, obtain new target requirement information corresponding to the model optimization request, and return to the target model determined to match the target requirement information to continue execution and obtain the latest target installation information;

[0238] The second sending unit is configured to send the latest target installation information to the user equipment, where the latest target installation information instructs the user equipment to install the latest target model.

[0239] As yet another embodiment, further comprising:

[0240] A login detection unit, configured to detect a login request sent by a user device and obtain login information provided by a target user;

[0241] The authentication unit is used to authenticate the target user based on the login information and obtain the authentication result;

[0242] The request passing unit is configured to receive a service acquisition request sent by the user equipment if the verification result is determined to be verification passed.

[0243] In some embodiments, further comprising:

[0244] A version establishing unit, configured to establish an installation version number for the target installation information, and store the installation version number in association with the target installation information;

[0245] A demand acquisition unit, configured to acquire demand information provided by other user devices;

[0246] The requirement matching unit is configured to send the target installation information corresponding to the latest installation version number to the other user equipment if it is determined that the requirement information provided by the other user equipment matches the target requirement information.

[0247] like Figure 10 FIG. 1 is a schematic diagram of the structure of a model installation device provided by the ninth embodiment of the present disclosure. The model installation device 1000 may include the following units:

[0248] Request generating unit 1001: used to generate a service acquisition request based on target demand information provided by a target user;

[0249] Request sending unit 1002: used to send a service acquisition request to the electronic device; the service acquisition request instructs the electronic device to generate a corresponding target model according to the target requirement information;

[0250] The first receiving unit 1003 is configured to receive target installation information corresponding to a target model sent by an electronic device;

[0251] The first installation unit 1004 is used to execute the target installation information to install the target model.

[0252] As an embodiment, it also includes:

[0253] A request detection unit, used to detect model optimization requests initiated by target users;

[0254] a request sending unit, configured to send a model optimization request to the electronic device, wherein the model optimization request instructs the electronic device to update the target model;

[0255] a second receiving unit, configured to receive the latest target installation information sent by the electronic device;

[0256] The second installation unit is configured to run the latest target installation information to install the latest target model.

[0257] The model installation device provided by the present disclosure can execute the model installation method of the above embodiment, and the specific content of the model installation method will not be repeated.

[0258] It should be noted that the head model in this embodiment is not a head model for a specific user and cannot reflect the personal information of a specific user. It should be noted that the two-dimensional face image in this embodiment comes from a public data set.

[0259] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0260] In addition, it should be noted that the user equipment involved in this disclosure also belongs to electronic equipment and is applicable to the following Figure 11 The structural diagram of the electronic device shown does not disclose and does not impose too many limitations on the specific type of the electronic device.

[0261] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0262] According to an embodiment of the present disclosure, the present disclosure also provides a computer program product, which includes: a computer program, the computer program is stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor executes the computer program so that the electronic device executes the solution provided by any of the above embodiments.

[0263] Figure 11 A schematic block diagram of an example electronic device 1100 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0264] like Figure 11 As shown, device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. Various programs and data required for the operation of device 1100 may also be stored in RAM 1103. Computing unit 1101, ROM 1102, and RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to bus 1104.

[0265] Various components in device 1100 are connected to I / O interface 1105, including an input unit 1106, such as a keyboard and mouse; an output unit 1107, such as various types of displays and speakers; a storage unit 1108, such as a magnetic disk and optical disk; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0266] The computing unit 1101 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 101 performs the various methods and processes described above, such as the model installation method. For example, in some embodiments, the model installation method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the computing unit 1101, one or more steps of the model installation method described above can be performed. Alternatively, in other embodiments, the computing unit 1101 can be configured to perform the model installation method by any other appropriate means (e.g., by means of firmware).

[0267] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0268] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0269] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0270] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0271] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0272] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact via a communication network. This client-server relationship is established by computer programs running on the respective computers, establishing a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host, a host product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosts and VPS services ("Virtual Private Servers" or simply "VPS"). The server may also be a server in a distributed system or a server integrated with blockchain.

[0273] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0274] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A model installation method, comprising: In response to a service acquisition request sent by a user device, obtaining target demand information carried in the service acquisition request, wherein the target demand information refers to demand information set by a target user according to the target user's industrial use target for the mathematical model; Performing demand analysis on the target demand information to obtain target model information including at least one demand processing node, wherein the demand processing node is a node divided according to the target demand information and used to implement a processing function, and different demand processing nodes have a connection relationship; If it is determined that the model update condition is met, then a corresponding to-be-trained model is constructed according to the target model information, including: reading at least one demand processing node in the target model information; determining at least one candidate mathematical model, wherein the candidate mathematical model corresponds to a processing function; determining a target mathematical model whose processing function matches the demand processing node from the at least one candidate mathematical model, so as to obtain a target mathematical model corresponding to at least one demand processing node; and combining the target mathematical models corresponding to the at least one demand processing node to obtain the to-be-trained model; Obtaining training data for the model to be trained; Using the training data to train the model to be trained to obtain a matching target model; generating target installation information corresponding to the target model; The target installation information is sent to the user equipment, where the target installation information is used to instruct the user equipment to install the target model.

2. The method according to claim 1, wherein After the training model to be trained is trained using the training data to obtain a target model, the method further includes: Generating model version information for the target model obtained through training; The target model and the model version information are associated and stored.

3. The method according to claim 1, wherein Combining the target mathematical models corresponding to at least one of the demand processing nodes to obtain the model to be trained includes: Determining target hyperparameters of a target mathematical model corresponding to at least one of the demand processing nodes based on a hyperparameter adjustment algorithm; According to the target hyperparameters respectively corresponding to at least one of the target mathematical models, at least one of the target mathematical models is combined according to the association relationship of at least one of the demand processing nodes to obtain the model to be trained.

4. The method according to any one of claims 1 to 3, wherein: The obtaining of training data for the model to be trained includes: Determining the input data type and output data type corresponding to the target model according to the target model information; According to the input data type and the output data type, the training data of the model to be trained is determined; the training data includes target training data corresponding to the input data type and a data label corresponding to the output data type.

5. The method according to any one of claims 1 to 3, further comprising: If it is determined that the model update condition is not met, the target model version information corresponding to the target model information is obtained to read the target model corresponding to the target model version information; the target model version information is pre-associated with the corresponding target model and stored.

6. The method according to any one of claims 1 to 3, further comprising: Obtain the model update frequency of the model to be trained and the end time of the most recent training; Determining the update start time of the model to be trained according to the model update frequency and the end time; When it is detected that the current time reaches the update start time, it is determined that the model update condition is met.

7. The method according to any one of claims 1 to 3, wherein: The generating target installation information corresponding to the target model includes: Determining registration information and target resources of the target model; generating a target application of the target model according to the registration information and target resources of the target model in combination with the target model; the registration information is used to instruct the target application to register for service in the user equipment; the target resources are used to be called by the user equipment when using the target model; According to the target application, target installation information corresponding to the target model is determined.

8. The method according to claim 7, wherein: The target demand information includes resource demand; the determination of the registration information and target resources of the target model includes: Scheduling the target resources for the target model according to the resource demand; According to the target resource, the target model is registered for service to obtain registration information of the target model.

9. The method according to claim 8, wherein Scheduling the target resources for the target model according to the resource demand includes: Acquiring at least one resource node and resource usage information of at least one of the resource nodes; Determining, according to the resource usage information, a candidate resource node in an idle state from at least one of the resource nodes; Selecting a target resource node from the candidate resource nodes according to the resource demand; A usage cycle of the target model is established for the target resource node to obtain a target resource corresponding to the target model.

10. The method according to any one of claims 1-3, 8-9, wherein After sending the target installation information to the user equipment, the method further includes: receiving a model optimization request sent by the user equipment; In response to the model optimization request, obtaining new target requirement information corresponding to the model optimization request, and returning to the step of performing demand analysis on the target requirement information to obtain the latest target installation information; The latest target installation information is sent to the user equipment, where the latest target installation information instructs the user equipment to install the latest target model.

11. The method according to any one of claims 1 to 3, 8 to 9, further comprising: Detecting a login request sent by the user device and obtaining login information provided by the target user; authenticating the target user according to the login information to obtain a verification result; If it is determined that the verification result is verification passed, the service acquisition request sent by the user equipment is received.

12. The method according to any one of claims 1-3, 8-9, wherein: After generating the target installation information corresponding to the target model, the method further includes: Establishing an installation version number for the target installation information, and storing the installation version number in association with the target installation information; Obtain demand information provided by other user devices; If it is determined that the requirement information provided by the other user equipment matches the target requirement information, the target installation information corresponding to the latest installation version number is sent to the other user equipment.

13. A model installation method comprising: Generate a service acquisition request based on target demand information provided by the target user, wherein the target demand information refers to demand information set by the target user according to the industrial use target of the mathematical model; The service acquisition request is sent to an electronic device; the service acquisition request instructs the electronic device to perform a demand analysis on the target demand information to obtain target model information including at least one demand processing node, wherein the demand processing node is a node divided according to the target demand information and used to implement a processing function, and different demand processing nodes have a connection relationship; if it is determined that the model update condition is met, a corresponding to-be-trained model is constructed according to the target model information; training data for the to-be-trained model is obtained; and the to-be-trained model is trained using the training data to generate a matching target model; receiving target installation information corresponding to the target model sent by the electronic device; Running the target installation information to install the target model; Among them, if it is determined that the model update condition is met, a corresponding model to be trained is constructed according to the target model information, including: reading at least one demand processing node in the target model information; determining at least one candidate mathematical model, and the candidate mathematical model corresponds to a processing function; from at least one of the candidate mathematical models, determining a target mathematical model whose processing function matches the demand processing node to obtain a target mathematical model corresponding to at least one of the demand processing nodes; combining the target mathematical models corresponding to at least one of the demand processing nodes to obtain the model to be trained.

14. The method according to claim 13, wherein After executing the target installation information to install the target model, the method further includes: Detecting a model optimization request initiated by the target user; Sending the model optimization request to the electronic device, wherein the model optimization request instructs the electronic device to update the target model; receiving the latest target installation information sent by the electronic device; Run the latest target installation instructions to install the latest target model.

15. A model mounting device comprising: a first responding unit, configured to respond to a service acquisition request sent by a user device and acquire target demand information carried in the service acquisition request, wherein the target demand information refers to demand information set by a target user according to the target user's industrial use target for the mathematical model; The model determination unit includes: a demand analysis module, configured to perform demand analysis on the target demand information to obtain target model information including at least one demand processing node, wherein the demand processing node is a node divided according to the target demand information and used to implement a processing function, and different demand processing nodes are connected; Model orchestration module, including: The model arrangement submodule is configured to construct a corresponding to-be-trained model according to the target model information if it is determined that the model update condition is met, including: reading at least one demand processing node in the target model information; determining at least one candidate mathematical model, wherein the candidate mathematical model corresponds to a processing function; determining a target mathematical model whose processing function matches the demand processing node from the at least one candidate mathematical model, to obtain a target mathematical model corresponding to at least one demand processing node; and combining the target mathematical models corresponding to the at least one demand processing node to obtain the to-be-trained model; A training acquisition submodule is used to obtain training data for the model to be trained; A model training submodule is used to train the model to be trained using the training data to obtain a matching target model; an information generating unit, configured to generate target installation information corresponding to the target model; A first sending unit is configured to send the target installation information to the user equipment, where the target installation information is used to instruct the user equipment to install the target model.

16. The apparatus according to claim 15, further comprising: A version acquisition unit, configured to generate model version information for the target model obtained through training; The first storage unit is configured to store the target model and the model version information in association with each other.

17. The device according to claim 15, wherein The model arrangement submodule is specifically used for: Determining target hyperparameters of a target mathematical model corresponding to at least one of the demand processing nodes based on a hyperparameter adjustment algorithm; According to the target hyperparameters respectively corresponding to at least one of the target mathematical models, at least one of the target mathematical models is combined according to the association relationship of at least one of the demand processing nodes to obtain the model to be trained.

18. The device according to any one of claims 15 to 17, wherein: The training acquisition submodule is specifically used for: Determining the input data type and output data type corresponding to the target model according to the target model information; According to the input data type and the output data type, the training data of the model to be trained is determined; the training data includes target training data corresponding to the input data type and a data label corresponding to the output data type.

19. The apparatus according to any one of claims 15 to 17, further comprising: The return execution unit is used to obtain the target model version information corresponding to the target model information if it is determined that the model update condition is not met, so as to read the target model corresponding to the target model version information; the target model version information is pre-associated with the corresponding target model and stored.

20. The apparatus according to any one of claims 15 to 17, further comprising: A time acquisition unit, used to obtain the model update frequency of the to-be-trained model and the end time of the most recent training; A time updating unit, configured to determine an update start time of the model to be trained according to the model update frequency and the end time; The time detection unit is used to detect when the current time reaches the start time and determine that the model update condition is met.

21. The device according to any one of claims 15 to 17, wherein: The information generating unit includes: An information acquisition module, configured to determine the registration information and target resources of the target model; a program generation module, configured to generate a target application of the target model based on the registration information and target resources of the target model and in combination with the target model; the registration information is used to instruct the target application to register for service in the user equipment; the target resources are used to be called by the user equipment when using the target model; The information determination module is used to determine the target installation information corresponding to the target model according to the target application.

22. The device according to claim 21, wherein The target demand information includes resource demand; The information acquisition module includes: A resource acquisition submodule, configured to schedule the target resources for the target model according to the resource demand; The model registration submodule is used to perform service registration for the target model according to the target resource and obtain registration information of the target model.

23. The device according to claim 22, wherein The resource acquisition submodule is specifically used to: Acquiring at least one resource node and resource usage information of at least one of the resource nodes; Determining, according to the resource usage information, a candidate resource node in an idle state from at least one of the resource nodes; Selecting a target resource node from the candidate resource nodes according to the resource demand; A usage cycle of the target model is established for the target resource node to obtain a target resource corresponding to the target model.

24. The apparatus according to any one of claims 15 to 17, 22 to 23, further comprising: an optimization receiving unit, configured to receive a model optimization request sent by the user equipment; a request optimization unit, configured to, in response to the model optimization request, obtain new target requirement information corresponding to the model optimization request, and return to the step of performing demand analysis on the target requirement information to obtain the latest target installation information; The second sending unit is configured to send the latest target installation information to the user equipment, where the latest target installation information instructs the user equipment to install the latest target model.

25. The apparatus according to any one of claims 15 to 17, 22 to 23, further comprising: A login detection unit, configured to detect a login request sent by the user device and obtain login information provided by the target user; An identity verification unit, configured to authenticate the target user based on the login information and obtain a verification result; The request passing unit is configured to receive the service acquisition request sent by the user equipment if it is determined that the verification result is verification passed.

26. The apparatus according to any one of claims 15 to 17, 22 to 23, further comprising: a version establishing unit, configured to establish an installation version number for the target installation information, and store the installation version number in association with the target installation information; A demand acquisition unit, configured to acquire demand information provided by other user devices; The requirement matching unit is configured to send target installation information corresponding to the latest installation version number to the other user equipment if it is determined that the requirement information provided by the other user equipment matches the target requirement information.

27. A model mounting device comprising: a request generating unit, configured to generate a service acquisition request based on target demand information provided by a target user, wherein the target demand information refers to demand information set by the target user according to the target user's industrial use target for the mathematical model; A request sending unit is configured to send the service acquisition request to the electronic device; the service acquisition request instructs the electronic device to perform a demand analysis on the target demand information to obtain target model information including at least one demand processing node, wherein the demand processing node is a node divided according to the target demand information and used to implement a processing function, and different demand processing nodes have a connection relationship; if it is determined that the model update condition is met, a corresponding to-be-trained model is constructed according to the target model information; training data for the to-be-trained model is obtained; and the to-be-trained model is trained using the training data to generate a matching target model; a first receiving unit, configured to receive target installation information corresponding to the target model sent by the electronic device; A first installation unit, configured to execute the target installation information to install the target model; Among them, if it is determined that the model update condition is met, a corresponding model to be trained is constructed according to the target model information, including: reading at least one demand processing node in the target model information; determining at least one candidate mathematical model, and the candidate mathematical model corresponds to a processing function; from at least one of the candidate mathematical models, determining a target mathematical model whose processing function matches the demand processing node to obtain a target mathematical model corresponding to at least one of the demand processing nodes; combining the target mathematical models corresponding to at least one of the demand processing nodes to obtain the model to be trained.

28. The apparatus according to claim 27, further comprising: a request detection unit, configured to detect a model optimization request initiated by the target user; a request sending unit, configured to send the model optimization request to the electronic device, wherein the model optimization request instructs the electronic device to update the target model; a second receiving unit, configured to receive the latest target installation information sent by the electronic device; The second installation unit is configured to run the latest target installation information to install the latest target model.

29. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-12 or 13-14.

30. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-12 or 13-14.

31. A computer program product comprising a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 12 or 13 to 14.

Images