Model subscription method, network function and storage medium

Abstract

The embodiment of the invention provides a model subscription method, a network function and a storage medium, and when the method is applied to a first network function, the method comprises the following steps: receiving a model subscription request message sent by a second network function, the model subscription request message comprising a model identifier, the model identifier is used for indicating the same model name of a group of models requested to be subscribed; a model subscription service of the set of models is established for the second network function. According to the embodiment of the invention, by defining the model identifier borne in the model change subscription request information, the flexible model subscription requirement is met.

Description

Technical Field

[0001] This application relates to technical fields such as communication, Artificial Intelligence (AI), and Machine Learning (ML), and particularly relates to a method for subscribing to a model, a network function, and a storage medium. Background Art

[0002] Current communication systems, including wireless communication networks, optical fiber communication networks, etc., are increasingly relying on Artificial Intelligence / Machine Learning models (AI / ML models) for optimization and management. These AI / ML models can be used to predict network traffic, detect network faults, optimize resource allocation, etc.

[0003] In the standards of the Open Radio Access Network (O-RAN) Alliance, to help run and uniformly manage these AI / ML models, AI / ML workflow services have been formulated in the Non-Real-Time Radio Intelligent Controller (Non-RT RIC).

[0004] Currently, although O-RAN defines a description of the AI / ML model change subscription function, its content is still vacant. For the AI / ML model change subscription function, there are various requirements in actual use, and the model subscription methods of the existing technologies are less flexible. Therefore, there is an urgent need for a flexible model subscription method to meet the requirements for the AI / ML model change subscription function in the O-RAN architecture. Summary of the Invention

[0005] At least one embodiment of this application provides a method for subscribing to a model, a network function, and a storage medium, which are used to solve the problem that the model subscription methods of the existing technologies are less flexible.

[0006] To solve the above technical problems, this application is implemented as follows:

[0007] In a first aspect, an embodiment of this application provides a method for subscribing to a model, which is applied to a first network function and includes:

[0008] Receiving a model subscription request message sent by a second network function, where the model subscription request message includes partial or all content of a model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches the partial or all content of the model identifier;

[0009] Establishing a model subscription.

[0010] Optionally, it further includes:

[0011] Send a model subscription response message to the second network function.

[0012] Optionally, the model identifier includes at least one of the following: model name, model version, build version.

[0013] Optionally, the model version or build version in the model subscription request message is represented by a first representation or a second representation, where the first representation is represented by the upper and lower bounds of a version number range, and the second representation is represented by the version number of each version.

[0014] Optionally, the model subscription request message further includes:

[0015] A first indication message for indicating whether the version number range of the model version or build version adopts the first representation or the second representation.

[0016] Optionally, the build version in the model subscription request message includes at least one of the following fields:

[0017] A first field for indicating the source of the training set for the build version;

[0018] A second field for indicating the scenario in which the training set for the build version is obtained;

[0019] A third field for indicating the time when the training set for the build version is obtained;

[0020] A fourth field for uniquely identifying the build version.

[0021] Optionally, the model version and / or build version in the model subscription request message further includes at least one of the following fields:

[0022] A fifth field for indicating the lifecycle stage of the model;

[0023] A sixth field for indicating the training task that generates the model.

[0024] Optionally, when the model identifier only includes the model name, at least one model that matches part or all of the model identifier is: a model with the model name in the model subscription request message;

[0025] When the model identifier only includes the model name and model version, at least one model that matches part or all of the model identifier is: a first type of model;

[0026] When the model identifier only includes the model name and the build version, at least one model that matches part or all of the content of the model identifier is: a second type of model;

[0027] When the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0028] Among them, the first type of model is a model that has the model name in the model identifier and matches the model version in the model identifier; the second type of model is a model that has the model name in the model identifier and matches the build version in the model identifier.

[0029] Optionally, the model subscription request message further includes:

[0030] Second indication information, which is used to indicate that when the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is the union of the first type of model and the second type of model, or the intersection of the first type of model and the second type of model.

[0031] Optionally, the model identifier further includes the following content: the behavior of model change;

[0032] When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier;

[0033] When the model identifier only includes the model name, the model version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a third type of model;

[0034] When the model identifier only includes the model name, the build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a fourth type of model;

[0035] When the model identifier includes the model name, the model version, the build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0036] Among them, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

[0037] Optionally, the behavior of model change includes at least one of the following: registering to a target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

[0038] Optionally, the model identifier further includes:

[0039] The third indication information is used to indicate that when the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is the union or intersection of the third type of model and the fourth type of model.

[0040] In a second aspect, an embodiment of the present application provides a method for subscribing to a model, which is applied to a second network function and includes:

[0041] Sending a model subscription request message to a first network function, where the model subscription request message includes part or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches part or all of the content of the model identifier.

[0042] Optionally, it further includes:

[0043] Receiving a model subscription response message sent by the second network function.

[0044] Optionally, the model identifier includes at least one of the following contents: model name, model version, build version.

[0045] Optionally, the model version or build version in the model subscription request message is represented by a first representation method or a second representation method, where the first representation method is represented by the upper and lower bounds of the version number range, and the second representation method is represented by the version number of each version.

[0046] Optionally, the model subscription request message further includes:

[0047] The first indication information is used to indicate that the version number range of the model version or build version adopts the first representation method or the second representation method.

[0048] Optionally, the build version in the model subscription request message includes at least one of the following fields:

[0049] The first field, which is used to indicate the source of obtaining the training set of the build version;

[0050] The second field, which is used to indicate the scenario of obtaining the training set of the build version;

[0051] The third field, which is used to indicate the time of obtaining the training set of the build version;

[0052] The fourth field, which is used to uniquely identify the build version.

[0053] Optionally, the model version and / or the build version in the model subscription request message further includes at least one of the following fields:

[0054] The fifth field, which is used to indicate the life cycle stage of the model;

[0055] The sixth field, which is used to indicate the training task that generates the model.

[0056] Optionally, when the model identifier only includes the model name, at least one model that matches part or all of the content of the model identifier is: the model with the model name in the model subscription request message;

[0057] When the model identifier only includes the model name and the model version, at least one model that matches part or all of the content of the model identifier is: the first type of model;

[0058] When the model identifier only includes the model name and the build version, at least one model that matches part or all of the content of the model identifier is: the second type of model;

[0059] When the model identifier includes the model name, the model version and the build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0060] Wherein, the first type of model is the model that has the model name in the model identifier and matches the model version in the model identifier; the second type of model is the model that has the model name in the model identifier and matches the build version in the model identifier.

[0061] Optionally, the model subscription request message further includes:

[0062] Second indication information, which is used to indicate that when the model identifier includes the model name, model version, and build version, at least one model that matches some or all of the content of the model identifier is the union or intersection of the first type of model and the second type of model.

[0063] Optionally, the model identifier further includes the following: the behavior of model change;

[0064] When the model identifier only includes the model name and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier;

[0065] When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: a third type of model;

[0066] When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: a fourth type of model;

[0067] When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0068] Wherein, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

[0069] Optionally, the behavior of model change includes at least one of the following: registering to a target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

[0070] Optionally, the model identifier further includes:

[0071] Third indication information, which is used to indicate that when the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is the union or intersection of the third type of model and the fourth type of model.

[0072] In a third aspect, an embodiment of the present application provides a first network function, including a transceiver and a processor, wherein,

[0073] The transceiver is configured to receive a model subscription request message sent by a second network function. The model subscription request message includes partial or all content of a model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches the partial or all content of the model identifier;

[0074] The processor is configured to establish a model subscription.

[0075] In a fourth aspect, an embodiment of the present application provides a first network function, including: a processor, a memory, and a program stored on the memory and executable on the processor. When the program is executed by the processor, it implements the steps of the method described in the first aspect.

[0076] In a fifth aspect, an embodiment of the present application provides a second network function, including a transceiver and a processor, wherein,

[0077] The transceiver is configured to send a model subscription request message to a first network function. The model subscription request message includes partial or all content of a model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches the partial or all content of the model identifier.

[0078] In a sixth aspect, an embodiment of the present application provides a second network function, including: a processor, a memory, and a program stored on the memory and executable on the processor. When the program is executed by the processor, it implements the steps of the method described in the second aspect.

[0079] In a seventh aspect, an embodiment of the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements the steps of the method described above.

[0080] Compared with the prior art, the model subscription method, network function, and storage medium provided by the embodiments of the present application meet the flexible model subscription requirements by defining the model identifier carried in the model change subscription request information. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of illustrating the preferred embodiments and are not considered to be a limitation of the present application. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0082] Figure 1 A flowchart of a model subscription method according to an embodiment of the present application;

[0083] Figure 2 Another flowchart of a model subscription method according to an embodiment of the present application;

[0084] Figure 3 An example diagram of an interaction process of a model subscription method according to an embodiment of the present application;

[0085] Figure 4 Another example diagram of an interaction process of a model subscription method according to an embodiment of the present application;

[0086] Figure 5 Yet another example diagram of an interaction process of a model subscription method according to an embodiment of the present application;

[0087] Figure 6 A schematic structural diagram of a first network function according to an embodiment of the present application;

[0088] Figure 7 A schematic structural diagram of a second network function according to an embodiment of the present application;

[0089] Figure 8 A schematic structural diagram of a first network function according to another embodiment of the present application;

[0090] Figure 9 A schematic structural diagram of a second network function according to another embodiment of the present application. Detailed implementation manners

[0091] Hereinafter, exemplary embodiments of the present application will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be completely conveyed to those skilled in the art.

[0092] The terms "first", "second", etc. in the description and claims of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented, for example, in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices. "And / or" in the description and claims means at least one of the connected objects.

[0093] The following description provides examples and is not intended to limit the scope, applicability, or configuration set forth in the claims. Changes may be made to the functions and arrangements of the elements discussed without departing from the spirit and scope of the disclosure. Various examples may appropriately omit, substitute, or add various procedures or components. For example, the methods described may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0094] The related-art AI / ML workflow services include:

[0095] - AI / ML training services;

[0096] - AI / ML model Management and Exposure services (MME);

[0097] - AI / ML model performance monitoring services.

[0098] Among them, the AI / ML model Management and Exposure services (MME) further include:

[0099] - AI / ML model registration service;

[0100] - AI / ML model storage service;

[0101] - AI / ML model discovery services;

[0102] - AI / ML model change subscription service;

[0103] - AI / ML model training capability registration service.

[0104] The role of AI / ML model management and exposure services is to help AI / ML models be used by different network optimization applications (rApps) in the Non-RT RIC. This is because developing duplicate AI / ML models for the same purpose consumes a large amount of human and material resources. The model registration-discovery-acquisition process provided by the platform facilitates the use of the same AI / ML model among different applications. That is, a certain rApp can register its own AI / ML model on the MME, and other rApps can discover the registered AI / ML model through certain conditions and download it for use or request AI / ML training services to train the discovered model.

[0105] There is also an important service among the above services, namely the AI / ML model change subscription service. This subscription service allows rApps to subscribe to changes in AI / ML model registration information. That is, an rApp sends an AI / ML model identifier to the MME. If this model is registered or deregistered, or the model information changes, or it is stored to or deleted from the MME, then the MME will notify the subscribed rApps of this information. The specific AI / ML model change subscription service defines three operations:

[0106] - Subscribe to AI / ML model changes;

[0107] - Unsubscribe from AI / ML model changes;

[0108] - Notify AI / ML model changes.

[0109] The AI / ML model identifier consists of three parts:

[0110] - Model name: Usually determined by the manufacturer of the AI / ML model, indicating the purpose of the model.

[0111] - Model version: Usually determined by the manufacturer of the AI / ML model, indicating the version number of the model.

[0112] - Artifact version: Usually determined by the MME. The same model trained with different training data sets will generate different build version numbers to distinguish different training data sets.

[0113] However, the existing method of defining the AI / ML model change information to be subscribed based on the AI / ML model identifier has its limitations, which are specifically analyzed below:

[0114] In the actual application of AI / ML models, the models are usually trained on a batch of historical data and then used to predict or optimize the future network state. However, due to various reasons, including changes in the network environment, changes in user behavior patterns, access of new devices, new service requirements, etc., the distribution of data may change, causing the performance of the trained model to decline over time and unable to adapt to environmental changes.

[0115] To solve the problem of model performance degradation, one approach is to retrain the model periodically. This usually involves retraining the model on a new data set. The relevant service is defined on the Non-RT RIC, namely the AI / ML model training service. The registered model can be passed to the model trainer (the Non-RT RIC platform or the AI / ML model training service provider), and then the model trained with new data is registered on the MME. The MME will assign a new artifact version to the new AI / ML model while keeping the model name and model version unchanged, indicating that the new AI / ML model is trained from a new training set.

[0116] As described in the background technology, O-RAN defines the description of the AI / ML model change subscription function, but its content is still vacant. For the AI / ML model change subscription function, there are various requirements in actual use, such as the subscription method based on the AI / ML model identifier (e.g., Model Name subscription, Model Version subscription), the subscription method based on the AI / ML model training scenario (e.g., the subscription method for AI / ML models generated in different steps such as model distribution, local training, model upload, model aggregation, and model deployment in federated learning), and the subscription method based on the AI / ML model change content (e.g., new model registration, modification of registration information, deregistration, change of storage status, etc.), or subscription methods considering other factors. Therefore, a flexible model subscription method is needed to meet the requirements of the AI / ML model change subscription function in the O-RAN architecture.

[0117] To solve at least one of the above problems, an embodiment of the present application provides a method for subscribing to a model, which can improve the flexibility of model subscription and meet different subscription requirements. Please refer to Figure 1 An embodiment of the present application provides a method for subscribing to a model, which is applied to the first network function side. The first network function may be an AI / ML model management and exposure service (MME) function, or other network functions, and the present application does not make specific limitations thereto. As Figure 1 shown, the method includes:

[0118] Step 11, receiving a model subscription request message sent by a second network function. The model subscription request message includes part or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches part or all of the content of the model identifier.

[0119] Here, the model identifier includes at least one of the following: model name, model version, build version. Among them, the model name can be defined by the model manufacturer. Optionally, the purpose of the model can be represented by the model name for the convenience of users. The model version can be defined by the model manufacturer and is used to represent the version number of the model. The artifact version can be defined by the provider of the training dataset or the first network function (such as MME). For example, different build version numbers are defined for the same model trained with different training datasets to distinguish different training datasets.

[0120] In an embodiment of the present application, the model subscription request message may be a model change subscription request. At this time, the second network function requests to subscribe to the change of the set of model registration information. The change may specifically include at least one of the following: the model is registered, deregistered, the model information changes, is stored in the target platform (such as the first network function or other platforms), is deleted from the target platform, etc.

[0121] Optionally, the model change subscription request message may further include the identification information of the second network function to indicate the second network function that sends the model subscription request message.

[0122] Step 12, establishing a model subscription.

[0123] Here, when the first network function allows to provide the model subscription service requested by the model subscription request message for the second network function, it will establish a model subscription for the second network function according to the model subscription request message.

[0124] Specifically, the first network function determines at least one model that matches part or all of the content of the model identifier in the model identifier, and then creates a model subscription for the at least one model for the second network function.

[0125] Through the above steps, the embodiments of the present application can create a model subscription for at least one model that matches part or all of the content of the model identifier in the model identifier, thereby improving the flexibility of the model subscription and meeting the subscription requirements for the AI / ML model change subscription function in the O-RAN architecture.

[0126] After the above step 12, the first network function may send a model subscription response message to the second network function to indicate whether the model subscription is successfully created.

[0127] After establishing the subscription service, the first network function may send a subscription notification to the second network function when it detects a model that meets the model subscription requested by the model subscription request message. For example, when the model subscription request message may be a model change subscription request, if the first network function creates a model change subscription service for the second network function, then when any model in the models requested to be subscribed by the model subscription request message changes, the first network function may send a model change notification message to the second network function, and the model change notification message is used to indicate information such as the changed model and / the changed content.

[0128] In addition, in the embodiments of the present application, the first network function may also generate a subscription identifier, and the subscription identifier is used to indicate the model subscription service of the group of models. Subsequently, when the first network function sends the model subscription response message and the subscription notification message, the subscription identifier may be carried in the above messages to indicate a specific model subscription.

[0129] In the embodiments of the present application, the model version or build version in the model subscription request message may be represented by a first representation method or a second representation method, where the first representation method is represented by the upper and lower bounds of the version number range, and the second representation method is represented by the version number of each version.

[0130] Specifically, it may be default or pre-agreed that the range of the model version or build version is represented by the upper and lower bounds of the version number. Of course, it may also be default or pre-agreed that the range of the model version or build version is represented by the version number of each version.

[0131] Optionally, the model subscription request message may further include: a first indication information for indicating that the version number range of the model version or build version adopts the first representation or the second representation. In this way, the first network function can determine the first representation or the second representation through the first indication information, and then, according to the determined representation, determine the model version or build version indicated by the model version or build version in the model subscription request message.

[0132] For example, the first indication information may be 1 bit. When the value is 1, it indicates that the version number range of the model version or build version adopts the first representation; when the value is 0, it indicates that the version number range of the model version or build version adopts the second representation. The first indication information may also be 2 bits. When the value is 00, it indicates that the version number ranges of both the model version and the build version adopt the first representation; when the value is 10, it indicates that the model version adopts the second representation and the version number range of the build version adopts the first representation; when the value is 01, it indicates that the model version adopts the first representation and the version number range of the build version adopts the second representation; when the value is 11, it indicates that the version number ranges of both the model version and the build version adopt the second representation.

[0133] In the embodiments of the present application, the build version in the model subscription request message may include at least one of the following fields:

[0134] A first field for indicating the source of obtaining the training set of the build version;

[0135] A second field for indicating the scenario of obtaining the training set of the build version;

[0136] A third field for indicating the time of obtaining the training set of the build version;

[0137] A fourth field for uniquely identifying the build version.

[0138] Through the above fields, the embodiments of the present application can implement the designation of the data set of the build version of the model requested to be subscribed, and achieve a more refined model subscription.

[0139] Further, the build version in the model subscription request message may further include at least one of the following fields:

[0140] A fifth field for indicating the life cycle stage of the model;

[0141] A sixth field for indicating the training task that generates the model.

[0142] The above-mentioned fifth field and sixth field can also be carried in the model version in the model subscription request message, or carried in both the model version and the build version in the model subscription request message.

[0143] In the embodiments of the present application, the first network function determines the model requested to be subscribed to in the model subscription request message according to the content included in the model identifier, that is, determines at least one model that matches part or all of the content of the model identifier. Specifically:

[0144] (1) When the model identifier only includes the model name, at least one model that matches part or all of the content of the model identifier is: the model with the model name in the model subscription request message;

[0145] (2) When the model identifier only includes the model name and the model version, at least one model that matches part or all of the content of the model identifier is: the first type of model;

[0146] (3) When the model identifier only includes the model name and the build version, at least one model that matches part or all of the content of the model identifier is: the second type of model;

[0147] (4) When the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0148] Among them, the first type of model is the model that has the model name in the model identifier and matches the model version in the model identifier; the second type of model is the model that has the model name in the model identifier and matches the build version in the model identifier.

[0149] Specifically, when the model identifier includes the model name, the model version, and the build version, it can be default or pre-agreed that at least one model that matches part or all of the content of the model identifier is the union of the first type of model and the second type of model. Of course, it can also be default or pre-agreed that at least one model that matches part or all of the content of the model identifier is the intersection of the first type of model and the second type of model.

[0150] Optionally, in the embodiment of the present application, the model subscription request message may further include second indication information, which is used to indicate that when the model identifier includes the model name, model version, and build version, at least one model that matches part or all of the content of the model identifier is the union or intersection of the first type of model and the second type of model. In this way, according to the second indication information, the first network function can determine whether to use the union or intersection to determine the model requested to be subscribed.

[0151] In another embodiment of the present application, the model identifier further includes the following content: the behavior of model change. Specifically, the behavior of model change includes at least one of the following: registering to the target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform. The target platform may be the first network function, other network functions or entities, and the embodiment of the present application does not make specific limitations thereto.

[0152] In the case where the model identifier further includes the behavior of model change, the first network function determines the model requested to be subscribed by the model subscription request message according to the content included in the model identifier, specifically including:

[0153] (1) When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier;

[0154] (2) When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the third type of model;

[0155] (3) When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the fourth type of model;

[0156] (4) When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0157] Among them, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

[0158] Similarly, when the model identifier includes the model name, model version, build version, and behavior of model change, it can be default or pre-agreed that at least one model that matches part or all of the content of the model identifier is the union of the third type of model and the fourth type of model. Of course, it can also be default or pre-agreed that at least one model that matches part or all of the content of the model identifier is the intersection of the third type of model and the fourth type of model.

[0159] Optionally, an embodiment of the present application can also include third indication information in the model subscription request message. The third indication information is used to indicate that when the model identifier includes the model name, model version, build version, and behavior of model change, at least one model that matches part or all of the content of the model identifier is the union of the third type of model and the fourth type of model, or is the intersection of the third type of model and the fourth type of model. In this way, according to the third indication information, the first network function can determine whether to use the union or the intersection to determine the model for which the subscription is requested.

[0160] It should be noted that the above "first field", "second field",..., "first indication information", "second indication information", etc. are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. That is to say, the above "first field", "second field",..., "first indication information", "second indication information", etc. are not used to limit the order or sequence of these fields or indication information.

[0161] Several specific examples of the model identifier are provided below.

[0162] A typical application scenario of AI / ML model change subscription is as follows:

[0163] 1. The rApp discovers the AI / ML model it needs through the MME;

[0164] 2. The rApp deploys the discovered AI / ML model to run inside it.

[0165] 3. The rApp hopes to continuously monitor the AI / ML model it is using. At this time, a new version of this model appears, and the platform notifies the rApp.

[0166] 4. The rApp decides whether to replace the currently running model with the new AI / ML model.

[0167] In each of the following examples, taking the first network function as the MME and the second network function as the rAPP, the rAPP sends an AI / ML model change subscription request message (corresponding to the model subscription request message above) to the MME, and the AI / ML model change subscription request message includes an AI / ML model identifier (corresponding to the model identifier above).

[0168] As an example, as shown in Table 1, the AI / ML model identifier includes the same Model Name (corresponding to the model name above) of a group of models for which subscription is requested, and may also include: Model Version (corresponding to the model version above), and / or, Artifact Version (corresponding to the build version above). Among them, the Model Version is used to indicate the range of the model versions of the group of models for which subscription is requested, and the Artifact Version is used to indicate the range of the build versions of the group of models for which subscription is requested.

[0169] Table 1

[0170]

[0171] After receiving the above AI / ML model identifier, the MME confirms the scope of the subscribed AI / ML model according to the content in the AI / ML model identifier.

[0172] 1) When the AI / ML model identifier only contains the Model Name and does not contain the Model Version and Artifact Version, the MME considers that the scope of the AI / ML models subscribed by the rApp is all AI / ML models with this Model Name and different Model Versions and different Artifact Versions. Thus, when any AI / ML model with this Model Name is newly registered, deregistered, its registration information is changed, stored in the non-real-time radio intelligent controller platform (Non-RT RIC Platform, hereinafter referred to as the platform) using the AI / ML model storage service or deleted from the platform, then this information will be notified to the rApp that requests the subscription.

[0173] 2) When the AI / ML model identifier only contains Model Name and Model Version and does not contain ArtifactVersion, the MME believes that the scope of the AI / ML models subscribed by the rApp is all AI / ML models with this Model Name and this Model Version and different Artifact Names. When any AI / ML model with this Model Name and Model Version is newly registered, deregistered, has its registration information changed, is stored using the AI / ML model storage service to the Non-RT RIC Platform (hereinafter referred to as the platform) or deleted from the platform, then this information will be notified to the rApp that requested the subscription.

[0174] 3) When the AI / ML model identifier contains all of Model Name, Model Version, and ArtifactVersion, the MME believes that the AI / ML model subscribed by the rApp is only the AI / ML model with this Model Name, this Model Version, and this Artifact Name. When this AI / ML model is newly registered, deregistered, has its registration information changed, is stored using the AI / ML model storage service to the Non-RT RIC Platform (hereinafter referred to as the platform) or deleted from the platform, then this information will be notified to the rApp that requested the subscription.

[0175] Furthermore, this example can also define the scope of the model version or build version to be subscribed based on the range expressions of the model version and build version in the AI / ML model identifier, and thus send it as part of the subscription request message to the MME, as shown in Table 2:

[0176] Table 2

[0177]

[0178]

[0179] In this way, the Model Name part is still mandatory and must be the same as the Model Name of the AI / ML model registered in the MME to be subscribed. For the latter two parts of the AI / ML model identifier, there are multiple ways to express:

[0180] For example, when the presentation mode is "1", the expression only includes the lower bound lowerBound and the upper bound upperBound of the version (corresponding to the first representation method above). Any ModelVersion or Artifact Version within the upper and lower bounds is the scope of subscribing to the AI / ML model. Among them, Model Version and ArtifactVersion define two sets of AI / ML models. What is actually subscribed is the intersection or union of these two model sets, which can be represented by another symbol.

[0181] For another example, when the presentation mode is "2", the expression is in the form of enumeration (corresponding to the second representation method above), that is, any enumerated Model Version or Artifact Version is the scope of subscribing to the AI / ML model. Among them, Model Version and Artifact Version define two sets of AI / ML models. What is actually subscribed is the intersection or union of the two sets, which can be indicated by another piece of information (such as the intersection and union information above).

[0182] It can be seen that by introducing optional model versions and / or build versions in the model identifier in the above examples, the specific model version and / or build version of the model requested to be subscribed can be indicated. That is to say, the model subscribed to by the model subscription request message is determined according to the model name, model version and / or build version carried in the model identifier.

[0183] As another example, the model identifier can further include the content of the model change behavior on the basis of the model version and the build version to indicate the model change behavior of the model requested to be subscribed, so as to achieve more refined model subscription. As shown in Table 3, this example further adds the model change behavior on the basis of the model version and / or component version. As shown by "modelBehavior" in Table 3, it can indicate the model behavior that the model version and / or component version requested to be subscribed needs to meet.

[0184] Table 3

[0185]

[0186] For example, the specific model change behaviors include, but are not limited to, one or more of the following behaviors:

[0187] - "newReg": The subscription scope is only for models with new registrations

[0188] - "Updreg": The subscription scope is only for models with changed registration information

[0189] - "deReg": The subscription scope is only for the model where deregistration information appears.

[0190] - "newStore": The subscription scope is only for the model where new storage to the platform appears.

[0191] - "delStore": The subscription scope is only for the model where new storage deletion from the platform appears.

[0192] -...

[0193] As another example, this example can also introduce fields such as the first field, the second field, the third field, and the fourth field in the build version mentioned above to indicate information such as the build version of the model, its training set, and life cycle, so as to achieve more refined model subscription. For example, the build version (Artifact version) of the model contains one or more of the following fields: aa.bb.cc.dd.

[0194] Among them, aa represents the identifier of the source of the training set acquisition. This identifier can be used to indicate that the data set is obtained from a single unit / module, or it can also be obtained from the combination of several units / modules. The specific representation method is not listed in detail.

[0195] bb represents the identifier of the scenario for training set acquisition. This identifier can indicate a specific scenario or a combination of several scenarios. For example, as shown in Table 4:

[0196] Table 4

[0197] bb meaning 1 Urban Microcell (Umi) 2 Urban Macrocell (Uma) 3 Suburban Macrocell (Sma) 4 Indoor scenario 5 Railway scenario 6 Ultra-dense networking scenario

[0198] cc represents the identifier of the time for training set acquisition. This identifier can indicate the specific time of data collection or the time span of data collection. One implementation method is as shown in Table 5. Among them, cc also includes multiple sub-fields, such as m, n, x, y, and z, etc. The meaning of each sub-field is as follows:

[0199] Table 5

[0200]

[0201]

[0202] dd represents the unique identifier (number / character) of the model. This identifier does not represent any meaning. Each trained model has a different identifier. Even if other fields are the same, the dd identifier can distinguish different models.

[0203] As another example, this example can also use the fifth and sixth fields above to indicate the lifecycle stage of the model for which subscription is requested and the training task that generates the model.

[0204] For example, allow the rApp to define the scope of the model it wants to subscribe to based on different links of a specific model training method. Send it to the MME as part of the subscription request signaling. For example, in distributed training, the collaboration methods between the Non-RT RIC and the Near-RT RIC include federated learning and training task migration. During the process of federated learning, different subscription methods for different models can be provided by combining with other indication information above for different links:

[0205] For example, in the federated learning process of the O-RAN architecture, the generation of the final model goes through multiple stages:

[0206] - Generation / Registration of the base AI / ML model

[0207] - The base model is sent to the Near-RT Radio Intelligent Controller (Near-RT RIC)

[0208] - The base model is locally trained in the Near-RT RIC

[0209] - The model completed local training is sent back from the Near-RT RIC to the Non-RT RIC

[0210] - The Non-RT RIC collects multiple locally trained models and performs Aggregation

[0211] - The Non-RT RIC distributes / deploys the aggregated model to the Near-RT RIC

[0212] During this process, the fields in the model identifier can be used to identify different lifecycle stages of federated learning. In this way, the rApp can narrow the scope of model subscription by giving the identifier of a specific stage.

[0213] In addition, in the training task migration in distributed model training, the Non-RT RIC may obtain the model to be trained from the Near-RT RIC and return it to the Near-RT RIC after training is completed in the Non-RT RIC.

[0214] In this example, a feasible method is to add a field in the artifact version field of the model identifier to represent the lifecycle stage of the model, for example:

[0215] Artifact version: aa.bb.cc.dd.ee

[0216] Among them, aa.bb.cc.dd has the same meaning as expressed above, and the meaning of ee is shown in Table 6 as follows:

[0217] Table 6

[0218] ee meaning 1 Model not participating in distributed training 2 Base model participating in federated learning 3 Model locally trained by Near-RT RIC in federated learning 4 Model aggregated by Non-RT RIC in federated learning 5 Model sent by Near-RT RIC for training 6 Model requested by Near-RT RIC and trained by Non-RT RIC

[0219] Meanwhile, in order to help the rApp distinguish which training task the model belongs to, the task identifier (corresponding to the sixth field above) can also be added to the Model Version or Artifact Version to further limit the model changes that occur only during the subscription of a specific training task. That is to say, in the embodiments of the present application, the model version or build version further includes a task identifier (i.e., the sixth field above), and the task identifier indicates a specific training task, so that it can be matched with the training task indicated by the ninth indication information in the model subscription request message. The task identifier can be implemented in the model build version or model version. Specifically:

[0220] Implementation method 1:

[0221] Artifact version: aa.bb.cc.dd.ee.ff

[0222] Among them, ff is the training task identifier, representing the model generated in the target training task corresponding to this identifier.

[0223] Implementation method 2:

[0224] Add a field to the Model version, which is the training task identifier or its abbreviation, representing the model generated in the target training task corresponding to this identifier.

[0225] Please refer to Figure 2 , a model subscription method provided by the embodiments of the present application is applied to the second network function side. The second network function can be functions such as rAPP, and the present application does not make specific limitations thereon. As Figure 2 shown, this method includes:

[0226] Step 21, send a model subscription request message to the first network function. The model subscription request message includes part or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches part or all of the content of the model identifier.

[0227] Here, the model identifier includes at least one of the following: model name, model version, build version.

[0228] Through the above steps, the second network function in the embodiment of the present application can request to subscribe to at least one model that matches some or all of the content of the model identifier through the model identifier in the model subscription request message, thereby improving the flexibility of model subscription and meeting the subscription requirements for the AI / ML model change subscription function in the O-RAN architecture.

[0229] Optionally, after the above step 21, the second network function can also receive the model subscription response message sent by the second network function, so as to determine whether the model subscription is successful according to this response message.

[0230] As an implementation, in the embodiment of the present application, the model identifier includes at least one of the following: model name, model version, build version.

[0231] Wherein, the model version or build version in the model subscription request message is represented by a first representation method or a second representation method. The first representation method is represented by the upper and lower bounds of the version number range, and the second representation method is represented by the version number of each version.

[0232] Optionally, the model subscription request message may further include:

[0233] The first indication information is used to indicate whether the version number range of the model version or build version adopts the first representation method or the second representation method.

[0234] Optionally, the build version in the model subscription request message includes at least one of the following fields:

[0235] The first field is used to indicate the source of the training set for the build version;

[0236] The second field is used to indicate the scenario for obtaining the training set of the build version;

[0237] The third field is used to indicate the time for obtaining the training set of the build version;

[0238] The fourth field is used to uniquely identify the build version.

[0239] Optionally, the model version and / or build version in the model subscription request message further includes at least one of the following fields:

[0240] The fifth field is used to indicate the life cycle stage of the model;

[0241] The sixth field is used to indicate the training task for generating the model.

[0242] In the embodiments of the present application, according to the different contents in the model identifier, at least one model that matches part or all of the contents of the model identifier is different. Specifically:

[0243] (1) When the model identifier only includes the model name, at least one model that matches part or all of the contents of the model identifier is: the model with the model name in the model subscription request message;

[0244] (2) When the model identifier only includes the model name and the model version, at least one model that matches part or all of the contents of the model identifier is: the first type of model;

[0245] (3) When the model identifier only includes the model name and the build version, at least one model that matches part or all of the contents of the model identifier is: the second type of model;

[0246] (4) When the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the contents of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0247] Among them, the first type of model is the model that has the model name in the model identifier and matches the model version in the model identifier; the second type of model is the model that has the model name in the model identifier and matches the build version in the model identifier.

[0248] Optionally, the model subscription request message further includes:

[0249] The second indication information is used to indicate that when the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the contents of the model identifier is the union of the first type of model and the second type of model, or the intersection of the first type of model and the second type of model.

[0250] Optionally, in the embodiments of the present application, the model identifier further includes the following content: the behavior of model change. At this time, according to the different contents in the model identifier, at least one model that matches part or all of the contents of the model identifier is specifically:

[0251] (1) When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier;

[0252] (2) When the model identifier only includes the model name, model version and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the third type of model;

[0253] (3) When the model identifier only includes the model name, build version and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the fourth type of model;

[0254] (4) When the model identifier includes the model name, model version, build version and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0255] Among them, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

[0256] Here, the behavior of model change includes at least one of the following: registering to the target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

[0257] Optionally, the model identifier further includes:

[0258] The third indication information is used to indicate that when the model identifier includes the model name, model version, build version and the behavior of model change, at least one model that matches part or all of the content of the model identifier is the union of the third type of model and the fourth type of model, or the intersection of the third type of model and the fourth type of model.

[0259] Several interaction examples of model subscription are provided below.

[0260] Interaction example 1:

[0261] In Non-RT RIC, the subscription service of the aim model includes three sub-operations, namely subscription, notification, and unsubscription. Among them, the process of subscribing to the change of the AI / ML model can be decomposed as follows Figure 3The following steps shown:

[0262] Step 31: The rApp sends a request to the AI / ML workflow functions to subscribe to an AI / ML model. The request includes the rAppId and a model scope definition based on the model identifier. The model scope definition IE is determined by one or more of the 5 extension methods given above.

[0263] Step 32: The AI / ML workflow functions establish a subscription service

[0264] Step 33: The AI / ML workflow functions return a success or failure response to the rApp. Optionally, the AI / ML workflow functions generate a subscription identifier and send it in the response to identify this subscription service.

[0265] Among them, the process of notifying the change of the AI / ML model can be decomposed into the following steps as Figure 4 shown:

[0266] Step 41: When the AI / ML workflow functions detect that a model within the subscription scope exhibits behavior that meets the subscription request, they send a notification to the subscribing rApp. The notification includes the AI / ML model identifier, change details, etc. Optionally, the response can include the subscription identifier to locate the specific subscription service.

[0267] Among them, the process of unsubscribing from the change of the AI / ML model can be decomposed into the following steps as Figure 5 shown:

[0268] Step 51: The rApp sends a request to the AI / ML workflow functions to unsubscribe from an AI / ML model. The request includes the rAppId and the AI / ML model identifier. Optionally, the request can include the subscription identifier to locate the specific subscription service.

[0269] Step 52: The AI / ML workflow functions cancel the subscription service

[0270] Step 53: The AI / ML workflow functions return a success or failure response to the rApp.

[0271] As can be seen from the above examples, in the embodiments of the present application, by defining the model identifier carried in the model change subscription request information, the flexible model subscription requirements are met. By stipulating the training data source, training set scenario, training set collection time, etc. in the model change subscription request, the subscription scope of the model is further customized. By allowing expressions of different model identifier parts transmitted in the model change subscription request, the subscription scope of the model is further customized. By means of the model training phase identifier in the model change subscription request, the model subscription scope for a specific training phase is further allowed.

[0272] Compared with the traditional model subscription method, the embodiments of the present application have at least the following advantages:

[0273] To address the demand for model change subscriptions in the network, the traditional subscription method only subscribes to changes in the model specified by the identifier by transmitting the model identifier. However, with the provision of the online model training service of the network platform itself and the emergence of collaborative model training among multiple platforms, the traditional method can no longer adapt to the complex model generation methods generated by different data set training and different training process stages. The embodiments of the present application can implement a more flexible and customized subscription model scope based on the model generation method. Compared with the prior art, the embodiments of the present application can, without changing the core process of model subscription, meet different subscription requirements by changing the definition method of IE and various extension methods.

[0274] The various methods of the embodiments of the present application are introduced above. Next, an apparatus for implementing the above methods will be further provided.

[0275] Please refer to Figure 6 , the embodiments of the present application further provide a first network function 600, including: a transceiver 601 and a processor 602;

[0276] The transceiver 601 is configured to receive a model subscription request message sent by a second network function, where the model subscription request message includes partial or all content of a model identifier, and the model subscription request message is used to request subscribing to at least one model that matches the partial or all content of the model identifier;

[0277] The processor 602 is configured to establish a model subscription.

[0278] Optionally, the transceiver is further configured to send a model subscription response message to the second network function.

[0279] Optionally, the model identifier includes at least one of the following: model name, model version, build version.

[0280] Optionally, the model version or build version in the model subscription request message is represented by a first representation or a second representation, where the first representation is represented by the upper and lower bounds of a version number range, and the second representation is represented by the version number of each version.

[0281] Optionally, the model subscription request message further includes:

[0282] A first indication information for indicating that the version number range of the model version or build version adopts the first representation or the second representation.

[0283] Optionally, the build version in the model subscription request message includes at least one of the following fields:

[0284] A first field for indicating the acquisition source of the training set of the build version;

[0285] A second field for indicating the scenario of obtaining the training set of the build version;

[0286] A third field for indicating the time of obtaining the training set of the build version;

[0287] A fourth field for uniquely identifying the build version.

[0288] Optionally, the model version and / or build version in the model subscription request message further includes at least one of the following fields:

[0289] A fifth field for indicating the lifecycle stage of the model;

[0290] A sixth field for indicating the training task that generates the model.

[0291] Optionally, when the model identifier only includes the model name, at least one model that matches part or all of the content of the model identifier is: a model with the model name in the model subscription request message;

[0292] When the model identifier only includes the model name and model version, at least one model that matches part or all of the content of the model identifier is: the first type of model;

[0293] When the model identifier only includes the model name and build version, at least one model that matches part or all of the content of the model identifier is: the second type of model;

[0294] When the model identifier includes the model name, model version, and build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0295] Wherein, the first type of model is a model having the model name in the model identifier and matching the model version in the model identifier; the second type of model is a model having the model name in the model identifier and matching the build version in the model identifier.

[0296] Optionally, the model subscription request message further includes:

[0297] Second indication information, which is used to indicate that when the model identifier includes the model name, model version, and build version, at least one model that matches part or all of the content of the model identifier is the union or intersection of the first type of model and the second type of model.

[0298] Optionally, the model identifier further includes the following: the behavior of model change;

[0299] When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model having the model name in the model identifier and matching the behavior of model change in the model identifier;

[0300] When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a third type of model;

[0301] When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a fourth type of model;

[0302] When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0303] Wherein, the third type of model is a model having the model name in the model identifier and matching both the model version and the behavior of model change in the model identifier; the fourth type of model is a model having the model name in the model identifier and matching both the build version and the behavior of model change in the model identifier.

[0304] Optionally, the behavior of model change includes at least one of the following: registering to a target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

[0305] Optionally, the model identifier further includes:

[0306] The third indication information is used to indicate that when the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is the union of the third type of model and the fourth type of model, or the intersection of the third type of model and the fourth type of model.

[0307] It should be noted that the device in this embodiment is the device corresponding to the method applied to the first network function above. The implementation manners in the above embodiments are all applicable to the embodiments of this device and can also achieve the same technical effects. The above device provided in the embodiments of the present application can implement all the method steps implemented in the above method embodiments and can achieve the same technical effects. Therefore, the same parts and beneficial effects as those in the method embodiments will not be specifically described in this embodiment.

[0308] Please refer to Figure 7 , the embodiments of the present application further provide a second network function 700, including: a transceiver 701 and a processor 702;

[0309] The transceiver 701 is configured to send a model subscription request message to the first network function, where the model subscription request message includes part or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches part or all of the content of the model identifier.

[0310] Optionally, the transceiver is further configured to receive a model subscription response message sent by the second network function.

[0311] Optionally, the model identifier includes at least one of the following contents: model name, model version, build version.

[0312] Optionally, the model version or build version in the model subscription request message is represented by a first representation method or a second representation method, where the first representation method is represented by the upper and lower bounds of the version number range, and the second representation method is represented by the version number of each version.

[0313] Optionally, the model subscription request message further includes:

[0314] The first indication information is used to indicate that the version number range of the model version or build version adopts the first representation method or the second representation method.

[0315] Optionally, the build version in the model subscription request message includes at least one of the following fields:

[0316] The first field is used to indicate the source for obtaining the training set of the build version;

[0317] The second field is used to indicate the scenario for obtaining the training set of the build version;

[0318] The third field is used to indicate the time for obtaining the training set of the build version;

[0319] The fourth field is used to uniquely identify the build version.

[0320] Optionally, the model version and / or the build version in the model subscription request message further includes at least one of the following fields:

[0321] The fifth field is used to indicate the lifecycle stage of the model;

[0322] The sixth field is used to indicate the training task that generates the model.

[0323] Optionally, when the model identifier only includes the model name, at least one model that matches part or all of the content of the model identifier is: a model with the model name in the model subscription request message;

[0324] When the model identifier only includes the model name and the model version, at least one model that matches part or all of the content of the model identifier is: the first type of model;

[0325] When the model identifier only includes the model name and the build version, at least one model that matches part or all of the content of the model identifier is: the second type of model;

[0326] When the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model;

[0327] Among them, the first type of model is a model that has the model name in the model identifier and matches the model version in the model identifier; the second type of model is a model that has the model name in the model identifier and matches the build version in the model identifier.

[0328] Optionally, the model subscription request message further includes:

[0329] The second indication information is used to indicate that when the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is the union of the first type of model and the second type of model, or the intersection of the first type of model and the second type of model.

[0330] Optionally, the model identifier further includes the following: the behavior of model change;

[0331] When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier;

[0332] When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a third type of model;

[0333] When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a fourth type of model;

[0334] When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model;

[0335] Wherein, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

[0336] Optionally, the behavior of model change includes at least one of the following: registering to a target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

[0337] Optionally, the model identifier further includes:

[0338] Third indication information, which is used to indicate that when the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is the union of the third type of model and the fourth type of model, or is the intersection of the third type of model and the fourth type of model.

[0339] It should be noted that the device in this embodiment corresponds to the method applied to the second network function above. The implementation manners in the above embodiments are all applicable to the embodiments of this device and can achieve the same technical effects. The device provided in the embodiments of this application can implement all the method steps implemented by the above method embodiments and can achieve the same technical effects. Therefore, the same parts and beneficial effects as those in the method embodiments will not be specifically described herein again.

[0340] Please refer to Figure 8 , an embodiment of this application also provides a terminal 800, including a processor 801, a memory 802, and a computer program stored on the memory 802 and executable on the processor 801. When the computer program is executed by the processor 801, it implements each process of the above-described method embodiment for subscribing to a model executed by the first network function and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0341] Please refer to Figure 9 , an embodiment of this application also provides a network device 900, including a processor 901, a memory 902, and a computer program stored on the memory 902 and executable on the processor 901. When the computer program is executed by the processor 901, it implements each process of the above-described method embodiment for subscribing to a model executed by the second network function and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0342] An embodiment of this application also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements each process of the above-described method embodiment for subscribing to a model and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium may be, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0343] It should be noted that in this article, the term "including", "containing" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element.

[0344] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0345] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

Claims

1. A subscription method for a model, applied to a first network function, characterized in that, including: receiving a model subscription request message sent by a second network function, where the model subscription request message includes partial or all content of a model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches the partial or all content of the model identifier; establishing a model subscription.

2. The method according to claim 1, characterized in that, It further includes: sending a model subscription response message to the second network function.

3. The method according to claim 1, characterized in that The model identifier includes at least one of the following contents: model name, model version, build version.

4. The method according to claim 3, wherein the model version or build version in the model subscription request message is represented by a first representation method or a second representation method, where the first representation method is represented by the upper and lower bounds of a version number range, and the second representation method is represented by the version number of each version.

5. The method according to claim 4, characterized in that, The model subscription request message further includes: a first indication information for indicating whether the version number range of the model version or build version adopts the first representation method or the second representation method.

6. The method according to claim 3, wherein The build version in the model subscription request message includes at least one of the following fields: a first field for indicating the source of the training set for the build version; a second field for indicating the scenario of obtaining the training set for the build version; a third field for indicating the time of obtaining the training set for the build version; a fourth field for uniquely identifying the build version.

7. The method according to claim 6, characterized in that The model version and / or build version in the model subscription request message further includes at least one of the following fields: a fifth field for indicating the lifecycle stage of the model; a sixth field for indicating the training task that generates the model.

8. The method according to any one of claims 3 to 7, wherein when the model identifier only includes the model name, at least one model that matches the partial or all content of the model identifier is: a model having the model name in the model subscription request message; when the model identifier only includes the model name and model version, at least one model that matches the partial or all content of the model identifier is: a first type of model; when the model identifier only includes the model name and build version, at least one model that matches the partial or all content of the model identifier is: a second type of model; when the model identifier includes the model name, model version, and build version, at least one model that matches the partial or all content of the model identifier is: the union or intersection of the first type of model and the second type of model; wherein, the first type of model is a model having the model name in the model identifier and matching the model version in the model identifier; the second type of model is a model having the model name in the model identifier and matching the build version in the model identifier.

9. The method according to claim 8, wherein The model subscription request message further includes: Second indication information, which is used to indicate that when the model identifier includes the model name, model version, and build version, at least one model that matches part or all of the content of the model identifier is the union or intersection of the first type of model and the second type of model.

10. The method according to any one of claims 3 to 7, characterized in that The model identifier further includes the following: the behavior of model change; When the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model that has the model name in the model identifier and matches the behavior of model change in the model identifier; When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a third type of model; When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a fourth type of model; When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model; Wherein, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

11. The method according to claim 10, wherein The behavior of model change includes at least one of the following: registering to a target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

12. The method according to claim 10, wherein The model identifier further includes: Third indication information, which is used to indicate that when the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches part or all of the content of the model identifier is the union or intersection of the third type of model and the fourth type of model.

13. A subscription method for a model, applied to a second network function, characterized in that Including: Sending a model subscription request message to a first network function, where the model subscription request message includes part or all of the content of the model identifier, and the model subscription request message is used to request a subscription to at least one model that matches part or all of the content of the model identifier.

14. The method according to claim 13, characterized in that Further including: Receiving a model subscription response message sent by the second network function.

15. The method according to claim 13, wherein The model identifier includes at least one of the following: model name, model version, build version.

16. The method according to claim 15, wherein The model version or build version in the model subscription request message is represented by a first representation method or a second representation method, where the first representation method is represented by the upper and lower bounds of a version number range, and the second representation method is represented by the version number of each version.

17. The method according to claim 16, wherein The model subscription request message further includes: First indication information for indicating whether the version number range of the model version or build version adopts the first representation method or the second representation method.

18. The method according to claim 15, wherein The build version in the model subscription request message includes at least one of the following fields: A first field for indicating the source of obtaining the training set of the build version; A second field for indicating the scenario of obtaining the training set of the build version; A third field for indicating the time of obtaining the training set of the build version; A fourth field for uniquely identifying the build version.

19. The method according to claim 18, wherein The model version and / or build version in the model subscription request message further includes at least one of the following fields: A fifth field for indicating the life cycle stage of the model; A sixth field for indicating the training task that generates the model.

20. The method according to any one of claims 15 to 19, wherein when the model identifier only includes the model name, at least one model that matches part or all of the content of the model identifier is: a model having the model name in the model subscription request message; when the model identifier only includes the model name and the model version, at least one model that matches part or all of the content of the model identifier is: a first type of model; when the model identifier only includes the model name and the build version, at least one model that matches part or all of the content of the model identifier is: a second type of model; when the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is: the union or intersection of the first type of model and the second type of model; wherein the first type of model is a model having the model name in the model identifier and matching the model version in the model identifier; the second type of model is a model having the model name in the model identifier and matching the build version in the model identifier.

21. The method according to claim 20, characterized in that, The model subscription request message further includes: Second indication information for indicating that when the model identifier includes the model name, the model version, and the build version, at least one model that matches part or all of the content of the model identifier is the union of the first type of model and the second type of model, or the intersection of the first type of model and the second type of model.

22. The method according to any one of claims 15 to 19, characterized in that, The model identifier further includes the following content: the behavior of model change; when the model identifier only includes the model name and the behavior of model change, at least one model that matches part or all of the content of the model identifier is: a model having the model name in the model identifier and matching the behavior of model change in the model identifier; When the model identifier only includes the model name, model version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: the third type of model; When the model identifier only includes the model name, build version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: the fourth type of model; When the model identifier includes the model name, model version, build version, and the behavior of model change, at least one model that matches some or all of the content of the model identifier is: the union or intersection of the third type of model and the fourth type of model; Among them, the third type of model is a model that has the model name in the model identifier and matches both the model version and the behavior of model change in the model identifier; the fourth type of model is a model that has the model name in the model identifier and matches both the build version and the behavior of model change in the model identifier.

23. The method according to claim 22, characterized in that, The behavior of model change includes at least one of the following: registering to the target platform; the registration information changes; deregistering from the target platform; storing to the target platform; deleting from the target platform.

24. The method according to claim 22, wherein The model identifier further includes: The third indication information, when the model identifier includes the model name, model version, build version, and the behavior of model change, indicating that at least one model that matches some or all of the content of the model identifier is the union of the third type of model and the fourth type of model, or the intersection of the third type of model and the fourth type of model.

25. A first network function, characterized in that, Comprising a transceiver and a processor, wherein, The transceiver is configured to receive a model subscription request message sent by a second network function, the model subscription request message includes some or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches some or all of the content of the model identifier; The processor is configured to establish a model subscription.

26. A first network function, characterized in that, Comprising: A processor, a memory, and a program stored on the memory and executable on the processor, and when the program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 12.

27. A second network function, characterized in that, Comprising a transceiver and a processor, wherein, The transceiver is configured to send a model subscription request message to a first network function, the model subscription request message includes some or all of the content of the model identifier, and the model subscription request message is used to request to subscribe to at least one model that matches some or all of the content of the model identifier.

28. A second network function, characterized in that, Comprising: A processor, a memory, and a program stored on the memory and executable on the processor, and when the program is executed by the processor, it implements the steps of the method according to any one of claims 13 to 24.

29. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 12, or implements the steps of the method according to any one of claims 13 to 24.

Images