Cloud model deployment processing method and device, equipment and storage medium
By obtaining configuration items from the server and application, generating the target cloud model and updating its version, the data security issues in cross-domain/cross-tenant collaboration are resolved, thereby improving data security and flexibility.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-03-24
AI Technical Summary
Existing model training methods cannot guarantee data security in scenarios such as cross-domain/cross-tenant collaboration.
By obtaining the model template selected by the server and the access permissions of the application, the target configuration items are determined, the target cloud model is generated, and version updates and deployments are performed to achieve data security and flexibility.
It improves data security for model training and deployment, and enhances the flexibility and stability of model training.
Smart Images
Figure CN121728084A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of big data processing, and particularly relates to a cloud model deployment processing method and device, equipment and a storage medium. BACKGROUND
[0002] With the rapid development of big data and artificial intelligence technology, the Internet of Things industry also accelerates evolution, and cooperation between cross-industry and cross-enterprise is increasingly close. The demand for model training cooperation between the two parties is also increasing.
[0003] At present, the model training scheme includes adjusting the learning rate for specific features after obtaining the features required by the algorithm of the model, adjusting the direction of the training model, and thus obtaining a model that meets specific needs.
[0004] However, the existing model training method cannot guarantee the security of data in cross-domain / cross-tenant cooperation scenarios. SUMMARY
[0005] The cloud model deployment processing method, device, equipment and storage medium provided by the embodiments of the present application solve the technical problem that the existing model training method cannot guarantee the security of data in cross-domain / cross-tenant cooperation scenarios.
[0006] In a first aspect, the embodiments of the present application provide a cloud model deployment processing method applied to a cloud model platform, wherein the cloud model platform is in communication connection with an application end and a service end respectively, and the method comprises the following steps:
[0007] In response to a model creation request of the service end, a model template selected by the service end and access permissions preconfigured by the service end for a plurality of application ends are obtained;
[0008] A plurality of first to-be-selected configuration items are determined according to the model template;
[0009] A first selection request of the service end for the first to-be-selected configuration items is received to obtain a plurality of first target configuration items;
[0010] In response to a model definition request of the application end, a plurality of second to-be-selected configuration items are determined according to the first to-be-selected configuration items and the access permissions;
[0011] A second selection request of the application end for the second to-be-selected configuration items is received to obtain a plurality of second target configuration items;
[0012] A target cloud model is generated according to the model template, the first target configuration items and the second target configuration items;
[0013] The target cloud model is published to deploy the target cloud model to the service end and the application end.
[0014] In a possible implementation, the publishing the target cloud model comprises: obtaining a first current cloud model of the service end and a second current cloud model of the application end; determining a first version update strategy of the target cloud model and the first current cloud model according to a preset first traffic control ratio; determining a second version update strategy of the target cloud model and the second current cloud model according to a preset second traffic control ratio; and updating the cloud models of the service end and the application end according to the first version update strategy and the second version update strategy.
[0015] In a possible implementation, the updating the cloud models of the service end and the application end according to the first version update strategy and the second version update strategy comprises: updating the first current cloud model of the service end to the target cloud model according to the first version update strategy, and storing a version number of the target cloud model; and updating the second current cloud model of the application end to the target cloud model according to the second version update strategy, and storing the version number of the target cloud model; wherein the version number is used to identify the cloud model that takes effect in the service end and the application end.
[0016] In a possible implementation, the generating the target cloud model according to the model template, the first target configuration item and the second target configuration item comprises: generating a first approval request according to the first target configuration item, so as to trigger a first approval response of the application end to the first target configuration item; generating a second approval request according to the second target configuration item, so as to trigger a second approval response of the service end to the second target configuration item; and generating the target cloud model according to the model template, the first approval response and the second approval response.
[0017] In a possible implementation, the first to-be-selected configuration item comprises multiple items in a group consisting of a feature item, a processing mode item of the feature, a learning rate, a number of iterations, a batch size, a regularization parameter, an evaluation index, a model target variable, a training set / test set division manner and whether to enable a specific function.
[0018] In a possible implementation, the receiving the first selection request of the service end to the first to-be-selected configuration item to obtain a plurality of first target configuration items comprises: sending the first to-be-selected configuration item to the service end, so as to display the first to-be-selected configuration item on the service end; and receiving a first selection request sent by the service end to obtain a plurality of first target configuration items; wherein the first selection request is generated by the service end in response to a selection operation of a user to the first to-be-selected configuration item.
[0019] In a possible implementation, the determining, according to the first to-be-selected configuration item and the access permission, of the plurality of second to-be-selected configuration items comprises: obtaining permission isolation information of the first to-be-selected configuration item; performing filtering processing in the first to-be-selected configuration item according to the permission isolation information and the access permission, to obtain the plurality of second to-be-selected configuration items.
[0020] In a possible implementation, after the publishing of the target cloud model to deploy the target cloud model to the service end and the application end, the method further comprises: monitoring real-time state information of all service ends and all application ends on which the target cloud model is deployed; wherein the real-time state information comprises a plurality of pieces of tenant state information, node state information, model version information, and link state information; and performing a model alarm strategy according to the real-time state information.
[0021] In a possible implementation, after the publishing of the target cloud model to deploy the target cloud model to the service end and the application end, the method further comprises: obtaining a containerized image generated by the service end according to the target cloud model; wherein the containerized image is used to start the containerized image to deploy an application according to the target cloud model when the application end is grabbed; and the containerized image comprises at least one of a gateway service, a model response service, and a model training service.
[0022] In a second aspect, an embodiment of the present application provides a cloud model deployment processing apparatus, comprising:
[0023] The obtaining module is configured to, in response to a model creation request of a service end, obtain a model template selected by the service end and access permissions preconfigured by the service end for a plurality of application ends;
[0024] The first determining module is configured to determine a plurality of first to-be-selected configuration items according to the model template;
[0025] The first receiving module is configured to receive a first selection request of the service end for the first to-be-selected configuration items, to obtain a plurality of first target configuration items;
[0026] The second determining module is configured to, in response to a model definition request of an application end, determine a plurality of second to-be-selected configuration items according to the first to-be-selected configuration items and the access permissions;
[0027] The second receiving module is configured to receive a second selection request of the application end for the second to-be-selected configuration items, to obtain a plurality of second target configuration items;
[0028] The generating module is configured to generate a target cloud model according to the model template, the first target configuration items, and the second target configuration items;
[0029] The publishing module is used to publish the target cloud model so as to deploy the target cloud model to the server and the application.
[0030] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;
[0031] The memory stores computer-executed instructions;
[0032] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.
[0033] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.
[0034] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.
[0035] The cloud model deployment processing method, apparatus, device, and storage medium provided in this application embodiment obtain a model template selected by the server, and in response to the selection of configuration items to be selected by the server and the application, obtain a first target configuration item and a second target configuration item. Based on the model template, the first target configuration item, and the second target configuration item, a target cloud model is generated and published to deploy the target cloud model to the server and the application. This improves the data security of model training and deployment, and also improves the flexibility of model training. Attached Figure Description
[0036] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0037] Figure 1 A schematic diagram illustrating the deployment and processing method of the cloud model provided in this application;
[0038] Figure 2 A flowchart illustrating the deployment process of the cloud model provided in this application;
[0039] Figure 3 A schematic diagram illustrating the process of cloud model deployment supported by a cloud model platform, provided as an embodiment of this application;
[0040] Figure 4An interaction architecture diagram of a cloud model deployment process provided in an embodiment of the present application is shown in the following figure;
[0041] Figure 5 A structure diagram of a cloud model deployment process apparatus provided in the present application is shown in the following figure;
[0042] Figure 6 A structure diagram of an electronic device provided in the present application is shown in the following figure.
[0043] The specific embodiments of the present application have been shown in the above figures, and will be described in more detail hereinafter. These figures and the written description are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0044] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to indicate the same or similar components. The embodiments described in the following exemplary embodiments do not represent all the embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.
[0045] In the technical solution of the present application, the collection, storage, use, processing, transmission, provision and disclosure of financial data or user data and other information comply with relevant laws and regulations and do not violate public order and good customs.
[0046] It should be noted that in the embodiments of the present application, some software, components, models and other existing solutions in the industry may be mentioned, which should be considered as exemplary, and the purpose is only to illustrate the feasibility of the implementation of the technical solution of the present application, but does not mean that the applicant has or will necessarily use the solution.
[0047] First, the terms involved in the present application are explained:
[0048] Model training: Model training is a complex and computationally intensive process that requires a large amount of data and powerful computing resources. It needs to go through: data preparation, model design, distributed parallel training, initialization parameters, model evaluation and debugging, model deployment, etc.
[0049] Containerized deployment: Containerized deployment is a method of packaging an application and all its dependencies into a self-contained container and deploying and running it in different environments. Containerized deployment provides higher portability, scalability and flexibility by isolating the application from its running environment.
[0050] Currently, after obtaining the features required by the algorithm of the model, the learning rate is adjusted according to the specific features, and the direction of the training model is adjusted, so as to obtain a model meeting specific requirements. In cross-domain / cross-tenant cooperation and other scenarios, model construction needs to be shared, and the security of data cannot be guaranteed.
[0051] The cloud model deployment processing method provided in the application responds to a model creation request of a service end, obtains a model template selected by the service end and access permissions preconfigured by the service end for a plurality of application ends, and generates a target cloud model according to the model template, first target configuration items and second target configuration items. In the application, the template selection of the service end and the application end and the selection of the configuration items improve the data security of model training and deployment and are more flexible.
[0052] Figure 1 The scene schematic diagram of the cloud model deployment processing method provided in the application is shown in FIG. 1. Figure 1 As shown in FIG. 1, the specific application scenario of the application includes a cloud model platform 101, an application end 102 and a service end 103.
[0053] The cloud model platform 101 is in communication connection with the application end 102 and the service end 103.
[0054] Specifically, the service end 103 sends a model creation request to the cloud model platform 101; the cloud model platform 101 determines a plurality of first to-be-selected configuration items according to the model creation request; the service end 103 sends a first selection request for the first to-be-selected configuration items to the cloud model platform 101; the cloud model platform 101 obtains a plurality of first target configuration items according to the first selection request; the application end 102 sends a model definition request to the cloud model platform 101; the cloud model platform 101 determines a plurality of second to-be-selected configuration items according to the model definition request; the application end 102 sends a second selection request for the second to-be-selected configuration items to the cloud model platform 101; the cloud model platform 101 obtains a plurality of second target configuration items according to the second selection request; the cloud model platform 101 generates a target cloud model according to the model creation request, the first target configuration items and the second target configuration items; and the cloud model platform 101 sends the target cloud model to the application end 102 and the service end 103.
[0055] The technical solutions of the application and how the technical solutions of the application solve the above technical problems will be described in detail in specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described in some embodiments. The embodiments of the application will be described below with reference to the drawings.
[0056] Figure 2 The flowchart of the cloud model deployment processing method provided in the application, and the execution subject of the embodiment can be Figure 1The cloud model platform 101 shown in the middle can also be other computer devices, and the present embodiment does not particularly limit the same. As shown in the figure, the cloud model platform 101 can include a model management module 102, a model training module 103, a model storage module 104, a model management database 105, a model training database 106, a model storage database 107, a model management interface 108, a model training interface 109, and a model storage interface 110. Figure 2 The method includes the following steps.
[0057] S201, in response to a model creation request of the server, obtaining a model template selected by the server and access permissions pre-configured by the server for a plurality of application terminals.
[0058] Optionally, the model creation request is obtained in response to a selection and an input operation of a user on a plurality of candidate model templates.
[0059] S202, determining a plurality of first to-be-selected configuration items according to the model template.
[0060] Optionally, the first to-be-selected configuration items include a plurality of items selected from a feature item, a processing method item of the feature, a learning rate, a number of iterations, a batch size, a regularization parameter, an evaluation index, a model target variable, a training set / test set division method, and whether to enable a specific function.
[0061] S203, receiving a first selection request of the server for the first to-be-selected configuration items to obtain a plurality of first target configuration items.
[0062] Optionally, the first to-be-selected configuration items are sent to the server to display the first to-be-selected configuration items on the server; and the first selection request sent by the server is received to obtain the plurality of first target configuration items; wherein the first selection request is generated by the server in response to a selection operation of a user on the first to-be-selected configuration items.
[0063] By displaying the first to-be-selected configuration items on the server, interaction with the user is realized, and the interactivity and configuration flexibility of the model training are improved.
[0064] S204, in response to a model definition request of the application terminal, determining a plurality of second to-be-selected configuration items according to the first to-be-selected configuration items and the access permissions.
[0065] Optionally, the permission isolation information of the first to-be-selected configuration items is obtained; and the first to-be-selected configuration items are filtered according to the permission isolation information and the access permissions to obtain the plurality of second to-be-selected configuration items.
[0066] The application terminals are isolated by the permission isolation information and the access permissions, so that the model configuration and the training process of different application terminals are isolated from each other, data cross-contamination is avoided, and the data security is improved.
[0067] S205, receiving a second selection request of the application terminal for the second to-be-selected configuration items to obtain a plurality of second target configuration items.
[0068] Optionally, the second to-be-selected configuration item is sent to the application end to display the second to-be-selected configuration item on the application end; and a second selection request sent by the application end is received to obtain a plurality of second target configuration items; wherein the second selection request is generated by the application end in response to a selection operation of a user on the second to-be-selected configuration item.
[0069] By displaying the second to-be-selected configuration item on the application end, interaction with the user is realized, and the interactivity and configuration flexibility of model training are improved.
[0070] S206, generating a target cloud model according to the model template, the first target configuration item and the second target configuration item.
[0071] Optionally, a first approval request is generated according to the first target configuration item to trigger a first approval response of the application end to the first target configuration item; a second approval request is generated according to the second target configuration item to trigger a second approval response of the service end to the second target configuration item; and a target cloud model is generated according to the model template, the first approval response and the second approval response.
[0072] By the mutual approval of the target configuration item by the application end and the service end, a basis is provided for the local training of the model by both parties, and the security of the data can be ensured.
[0073] S207, publishing the target cloud model to deploy the target cloud model to the service end and the application end.
[0074] Optionally, a first current cloud model of the service end and a second current cloud model of the application end are obtained; a first version update strategy of the target cloud model and the first current cloud model is determined according to a preset first traffic control ratio; a second version update strategy of the target cloud model and the second current cloud model is determined according to a preset second traffic control ratio; and the cloud models of the service end and the application end are updated according to the first version update strategy and the second version update strategy.
[0075] By the preset traffic control ratio, the version update strategy is determined, and the traffic is switched in the percentage manner, so that the traffic can be gradually and smoothly migrated from the old model to the new model, the risk brought by the deployment of the new model is reduced, and the performance stability is improved.
[0076] It should be noted that when the target cloud model is published, a version number is automatically generated, the model version is distinguished by the version number, and the version number cannot be the same as that of the previous model. It is necessary to ensure that there is only one model version in effect.
[0077] Optionally, after the target cloud model is published, it further includes generating and saving a model publishing record.
[0078] Optionally, when updating the cloud model of the service end and the application end according to the first version update strategy and the second version update strategy, the method comprises: updating the first current cloud model of the service end to the target cloud model according to the first version update strategy, and storing the version number of the target cloud model; updating the second current cloud model of the application end to the target cloud model according to the second version update strategy, and storing the version number of the target cloud model; wherein the version number is used to identify the cloud model that takes effect in the service end and the application end.
[0079] By storing the version number of the target cloud model when updating to the target cloud model, it can be ensured that only one model version takes effect, thereby improving the stability of model running.
[0080] The cloud model deployment processing method provided by the embodiment of the application improves the data security of model training and deployment, and improves the flexibility of model training.
[0081] Optionally, in an embodiment of the application, after step S207, the process further comprises monitoring the real-time state information of the target cloud model, which is described in detail as follows:
[0082] The real-time state information comprises multiple pieces of information including tenant state information, node state information, model version information and link state information; and the model alarm strategy is executed according to the real-time state information.
[0083] The cloud model deployment processing method provided by the embodiment of the application can track the running of the target cloud model in real time, and timely alarm, which can reduce the risk of model application and ensure the business continuity of model service.
[0084] Optionally, in an embodiment of the application, after step S207, the process further comprises the process of obtaining a containerized image, which is described in detail as follows:
[0085] The containerized image is used to start the containerized image after being grabbed by the application end, so as to deploy the application according to the target cloud model; wherein the containerized image comprises at least one of a gateway service, a model response service and a model training service.
[0086] The cloud model deployment processing method provided by the embodiment of the application can realize cross-platform and cross-system containerization deployment of the target cloud model by deploying the target cloud model through a containerized image, and improve the adaptability of model deployment.
[0087] Figure 3 A process schematic diagram of a cloud model platform supporting cloud model deployment provided by the embodiment of the application is shown in Figure 3 The cloud model platform includes a permission management unit, a model definition unit, a model auditing unit, a model publishing unit and a model monitoring unit. The cloud model platform can be deployed on the Internet, and the change of the model needs to be signed and reviewed by the application end and the service end.
[0088] The permission management unit is configured to provide the permission of the application end and the service end to jointly define a template, create an application account by the service end, create a model and set a model security parameter, and configure the permission of the corresponding model. Through the way of different tenants, it is ensured that multiple application ends can be isolated in terms of permission.
[0089] The model definition unit is configured to realize rapid definition of the model through a candidate model template of the cloud model platform. The service end can create a new model, select a model template, and then define the model by the application end and the service end: fill in the corresponding feature check box, define the feature and learning rate in the model configuration process, and other training indicators.
[0090] The model auditing unit is configured to sign and approve by the application end and the service end.
[0091] The model publishing unit is configured to publish the model after the signed and approved, support the gray release of the model, that is, switch the traffic by percentage. The version number is used to distinguish the version, the version number of the model is automatically generated during publishing, and cannot be the same as the version number of the previous model. Only one model version is effective, and the model publishing record is retained.
[0092] The model monitoring unit includes model link tracking, model deployment end monitoring and model log monitoring, is configured to upgrade the model by capturing, can monitor whether all tenants (that is, the application end) and the nodes in the model application area are alive, and can collect the corresponding link tracking, and give a warning for the non-living nodes.
[0093] It should be noted that the model application area is a core execution layer in the cloud model platform, which undertakes the output of the model publishing link and provides model services for business scenarios, is responsible for deploying the model passed through the signed and approved as a usable service, supports the isolated calling of multiple tenants (that is, multiple application ends), and guarantees the service stability and data security.
[0094] It should be noted that the nodes in the model application area can be container instances of the model application area in the cloud model platform. The container instance is a dynamic running unit that can perform specific tasks based on a containerized image.
[0095] Optionally, with reference back to Figure 3 The embodiment of the application provides another cloud model deployment processing method, which comprises the following steps:
[0096] Through the permission management unit, the server preconfigures a plurality of application end access permissions by creating an application end account; the server creates a model and sets a model security parameter.
[0097] In the model definition unit, a model is newly created in response to a model new creation request of the server, and a model template selected by the server is obtained; a first target configuration item is obtained by receiving a first selection request of the server; a second target configuration item is obtained by receiving a second selection request of the application end, and common model definition is completed.
[0098] In the model audit unit, a first approval response of the application end to the first target configuration item is obtained; a second approval response of the server to the second target configuration item is obtained, and double-approval is completed.
[0099] In the model publishing unit, if the result of the double-approval is approval, the target cloud model is published to end the cloud model deployment processing. After the target cloud model is published, the target cloud model is adjusted according to the model template selected by the server.
[0100] The cloud model deployment processing method provided by the embodiment of the application can improve data security by means of double-approval for model publishing.
[0101] Figure 4 An interaction architecture schematic diagram of a cloud model deployment processing provided by the embodiment of the application is provided. As Figure 4 shown, the architecture comprises a cloud model platform, a private line and a model application domain. The model application domain is a model application domain of the application end, comprising a local model deployment unit and an application end self-owned application.
[0102] It should be noted that the model application domain refers to a logical isolation unit divided according to business scenarios, tenants or functional requirements in the model application area, which is used to realize data isolation, permission isolation and resource isolation between different businesses and different partners.
[0103] On the basis of the above embodiment, after the cloud model platform publishes the target cloud model, the server generates a containerized image according to the target cloud model; the application end starts in batches through a container management tool after grabbing the containerized image, and completes deployment application, wherein the network mapping and path mapping outside and inside the container are configurable.
[0104] Among them, the local model deployment unit includes a gateway service, a model response service and a model training service, which respectively play different roles, and all realize transactions through stateless desensitization security response mode.
[0105] The gateway service is used for link tracking and adopts asymmetric encryption to encrypt the interaction message. The model response service and the model training service interact with the cloud model platform through the gateway service.
[0106] The model response service reports a heartbeat through the gateway service, the heartbeat contains application end basic information and the current cloud model version number, and can realize model upgrade through real-time grabbing, implement traffic switching configuration of model gray release, that is, traffic switching through percentage. The application end self-application can initiate training to the model response service.
[0107] The model training service is used for processing and packaging the local feature data of the application end to obtain application packaging data; obtaining a first feature score of the application end to the application packaging data; obtaining a second feature score of the desensitization feature data of the local feature data of the service end; and obtaining a joint training model by using a target cloud model for model training according to the first feature score and the second feature score. The model training is triggered by the model response service.
[0108] In another embodiment of the application, the purpose of the joint training model is to introduce target customers for the application end. Accordingly, the local feature data of the application end is processed and packaged to obtain application packaging data, which can be processed and packaged for transaction flow to obtain transaction flow standard deviation.
[0109] It should be noted that when obtaining the first feature score of the application end to the application packaging data, the first target configuration item of the service end is required as the basis, and the application end needs to package the corresponding local feature data.
[0110] The interaction architecture of the cloud model deployment processing provided by the embodiment of the application realizes model training through the application end local model deployment unit, further encapsulates the local feature data, and then trains the cloud model loaded locally on the cloud model platform of the application end. The service end pressure is reduced, and the data security is improved.
[0111] Figure 5 The structure diagram of the cloud model deployment processing device provided by the application is shown in Figure 5 The cloud model deployment processing device provided by the embodiment of the application includes an obtaining module 501, a first determining module 502, a first receiving module 503, a second determining module 504, a second receiving module 505, a generating module 506 and a publishing module 507.
[0112] The acquisition module 501 is configured to acquire a model template selected by the service end and access rights preconfigured by the service end for a plurality of application ends in response to a model creation request of the service end.
[0113] The first determination module 502 is configured to determine a plurality of first to-be-selected configuration items according to the model template.
[0114] The first receiving module 503 is configured to receive a first selection request of the service end for the first to-be-selected configuration items, to obtain a plurality of first target configuration items.
[0115] The second determination module 504 is configured to determine a plurality of second to-be-selected configuration items according to the first to-be-selected configuration items and the access rights in response to a model definition request of the application end.
[0116] The second receiving module 505 is configured to receive a second selection request of the application end for the second to-be-selected configuration items, to obtain a plurality of second target configuration items.
[0117] The generation module 506 is configured to generate a target cloud model according to the model template, the first target configuration items and the second target configuration items.
[0118] The publishing module 507 is configured to publish the target cloud model, to deploy the target cloud model to the service end and the application end.
[0119] In a possible implementation, the publishing module 507 specifically includes: acquiring a first current cloud model of the service end and a second current cloud model of the application end; determining a first version update strategy of the target cloud model and the first current cloud model according to a preset first traffic control ratio; determining a second version update strategy of the target cloud model and the second current cloud model according to a preset second traffic control ratio; and updating the cloud models of the service end and the application end according to the first version update strategy and the second version update strategy.
[0120] In a possible implementation, when updating the cloud models of the service end and the application end according to the first version update strategy and the second version update strategy, the publishing module 507 specifically includes: updating the first current cloud model of the service end to the target cloud model according to the first version update strategy, and storing a version number of the target cloud model; updating the second current cloud model of the application end to the target cloud model according to the second version update strategy, and storing the version number of the target cloud model; and the version number is used to identify the cloud model that takes effect in the service end and the application end.
[0121] In a possible implementation, the generating module 506 specifically includes: generating a first approval request according to the first target configuration item, to trigger a first approval response of the application end for the first target configuration item; generating a second approval request according to the second target configuration item, to trigger a second approval response of the service end for the second target configuration item; and generating the target cloud model according to the model template, the first approval response and the second approval response.
[0122] In a possible implementation, the first to-be-selected configuration item includes multiple of the feature item, the processing mode item of the feature, the learning rate, the iteration number, the batch size, the regularization parameter, the evaluation index, the model target variable, the training set / test set division mode and whether to enable a specific function.
[0123] In a possible implementation, the first receiving module 503 specifically includes: sending the first to-be-selected configuration item to the service end, to display the first to-be-selected configuration item on the service end; and receiving a first selection request sent by the service end, to obtain the multiple first target configuration items; wherein the first selection request is generated by the service end in response to a selection operation of a user on the first to-be-selected configuration item.
[0124] In a possible implementation, the second determining module 504 specifically includes: obtaining permission isolation information of the first to-be-selected configuration item; and performing filtering processing on the first to-be-selected configuration item according to the permission isolation information and the access permission, to obtain the multiple second to-be-selected configuration items.
[0125] In a possible implementation, the cloud model deployment processing apparatus further includes:
[0126] The alarm module is configured to monitor real-time state information of all service ends and all application ends to which the target cloud model is deployed, and execute a model alarm strategy according to the real-time state information; wherein the real-time state information includes multiple of tenant state information, node state information, model version information and link state information.
[0127] In a possible implementation, the obtaining module 501 is further configured to obtain a containerized image generated by the service end according to the target cloud model; wherein the containerized image is used to start the containerized image to deploy the application according to the target cloud model after the application end captures the containerized image; and the containerized image includes at least one of a gateway service, a model response service and a model training service.
[0128] The cloud model deployment processing apparatus provided in this embodiment can execute the method provided in the method embodiments, and has similar implementation principles and technical effects, which will not be described here in detail.
[0129] Figure 6 The structural schematic diagram of the electronic device provided in this application is shown in FIG. 1. Figure 6As shown, the electronic device provided in this embodiment includes at least one processor 601 and a memory 602. Optionally, the electronic device further includes a communication component 603. The processor 601, the memory 602 and the communication component 603 are connected through a bus 604.
[0130] In the implementation process, the at least one processor 601 executes the computer execution instructions stored in the memory 602, so that the at least one processor 601 performs the cloud model deployment processing method described above.
[0131] The specific implementation process of the processor 601 can refer to the method embodiments described above, which have similar implementation principles and technical effects, and will not be described here in this embodiment.
[0132] In the above embodiments, it should be understood that the processor can be a central processing unit (English: Central Processing Unit, for short: CPU), and can also be other general-purpose processors, digital signal processors (English: Digital Signal Processor, for short: DSP), application specific integrated circuits (English: Application Specific Integrated Circuit, for short: ASIC) and the like. The general-purpose processor can be a microprocessor, or the processor can also be any conventional processor, etc. The steps of the method disclosed in combination with the application can be directly embodied as execution completed by a hardware processor, or executed by a combination of hardware and software modules in the processor.
[0133] The memory can include a random access memory (Random Access Memory, RAM), and can also include a non-volatile memory (Non-volatile Memory, NVM), for example, at least one disk memory.
[0134] The bus can be an industry standard architecture (Industry Standard Architecture, ISA) bus, a peripheral component interconnect (Peripheral Component, PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the bus in the drawings of the present application does not limit only one bus or one type of bus.
[0135] The present application also provides a computer program product, including a computer program, which is executed by a processor to implement the cloud model deployment processing method described above.
[0136] The application further provides a computer readable storage medium, wherein computer execution instructions are stored in the computer readable storage medium, and when a processor executes the computer execution instructions, the cloud model deployment processing method is realized.
[0137] The readable storage medium can be implemented by any type of volatile or nonvolatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special purpose computer.
[0138] An exemplary readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in the device.
[0139] The division of units is only a logical function division, and in actual implementation, there can be another division mode, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0140] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on a plurality of network units. According to actual needs, some or all of the units can be selected to achieve the purpose of the embodiment.
[0141] In addition, the functional units in each embodiment of the application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.
[0142] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.
[0143] It can be understood by those skilled in the art that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware. The aforementioned program can be stored in a computer readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, and various media that can store program codes.
[0144] Finally, it should be noted that: those skilled in the art will easily think of other embodiments of the present application after considering the specification and practicing the application disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application that follow the general principles of the present application and include common knowledge or conventional technical means in the art that are not disclosed in the present application, and is not limited to the precise structure described above and shown in the drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present application is only limited by the appended claims.
Claims
1. A method for deploying and processing a cloud model, characterized in that, Applied to a cloud model platform, the cloud model platform establishes communication connections with both the application terminal and the server terminal; the method includes: In response to a model creation request from the server, obtain the model template selected by the server and the access permissions pre-configured by the server for multiple applications; Based on the model template, multiple first configuration items to be selected are determined; Receive the server's first selection request for the first configuration item to be selected, so as to obtain multiple first target configuration items; In response to the model definition request from the application, multiple second configuration options are determined based on the first configuration option to be selected and the access permissions. Receive a second selection request from the application for the second configuration item to be selected, so as to obtain multiple second target configuration items; A target cloud model is generated based on the model template, the first target configuration item, and the second target configuration item. Publish the target cloud model to deploy the target cloud model to the server and the application.
2. The method according to claim 1, characterized in that, The release of the target cloud model includes: Obtain the first current cloud model of the server and the second current cloud model of the application. Based on a preset first traffic control ratio, determine the first version update strategy for the target cloud model and the first current cloud model; Based on the preset second traffic control ratio, a second version update strategy for the target cloud model and the second current cloud model is determined; The cloud models of the server and the application are updated according to the first version update strategy and the second version update strategy.
3. The method according to claim 2, characterized in that, The step of updating the cloud model of the server and the application according to the first version update strategy and the second version update strategy includes: According to the first version update strategy, the first current cloud model on the server is updated to the target cloud model, and the version number of the target cloud model is stored; According to the second version update strategy, the second current cloud model of the application is updated to the target cloud model, and the version number of the target cloud model is stored; The version number is used to identify the cloud model that is in effect on the server and the application.
4. The method according to claim 1, characterized in that, The step of generating a target cloud model based on the model template, the first target configuration item, and the second target configuration item includes: Based on the first target configuration item, a first approval request is generated to trigger the application's first approval response to the first target configuration item; Based on the second target configuration item, a second approval request is generated to trigger the server's second approval response to the second target configuration item; A target cloud model is generated based on the model template, the first approval response, and the second approval response.
5. The method according to claim 1, characterized in that, The first selectable configuration items include multiple items from the following: feature items, feature processing method items, learning rate, number of iterations, batch size, regularization parameter, evaluation metric, model target variable, training set / test set partitioning method, and whether to enable specific functions.
6. The method according to any one of claims 1 to 5, characterized in that, The step of receiving the server's first selection request for the first configuration item to be selected, in order to obtain multiple first target configuration items, includes: The first configuration item to be selected is sent to the server so that the first configuration item to be selected is displayed on the server. The server receives a first selection request to obtain multiple first target configuration items; wherein the first selection request is generated by the server in response to the user's selection operation on the first configuration item to be selected.
7. The method according to any one of claims 1 to 5, characterized in that, The step of determining multiple second configuration options based on the first configuration option to be selected and the access permissions includes: Obtain the permission isolation information of the first configuration item to be selected; Based on the permission isolation information and the access permissions, a filtering process is performed on the first candidate configuration item to obtain multiple second candidate configuration items.
8. The method according to any one of claims 1 to 5, characterized in that, After publishing the target cloud model to deploy it to the server and the application, the method further includes: Monitor the real-time status information of all servers and applications deploying the target cloud model; wherein, the real-time status information includes multiple of the following: tenant status information, node status information, model version information, and link status information; Based on the real-time status information, execute the model alarm strategy.
9. The method according to any one of claims 1 to 5, characterized in that, After publishing the target cloud model to deploy it to the server and the application, the method further includes: Obtain the containerized image generated by the server based on the target cloud model; wherein, the containerized image is used to launch the containerized image after the application server crawls it, so as to deploy the application according to the target cloud model; The containerized image includes at least one of a gateway service, a model response service, and a model training service.
10. A cloud model deployment and processing device, characterized in that, include: The acquisition module is used to respond to the model creation request from the server and acquire the model template selected by the server and the access permissions pre-configured by the server for multiple application clients. The first determining module is used to determine multiple first selectable configuration items based on the model template; The first receiving module is used to receive the first selection request from the server for the first configuration item to be selected, so as to obtain multiple first target configuration items; The second determining module is used to respond to the model definition request from the application and determine a plurality of second selectable configuration items based on the first selectable configuration item and the access permission. The second receiving module is used to receive the second selection request from the application terminal for the second configuration item to be selected, so as to obtain multiple second target configuration items; The generation module is used to generate a target cloud model based on the model template, the first target configuration item, and the second target configuration item; The publishing module is used to publish the target cloud model so as to deploy the target cloud model to the server and the application.
11. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-9.
12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-9.
13. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-9.