A model request method, a model feedback method, a communication entity, and a storage medium

Abstract

The application discloses a model request method, a model feedback method, a communication entity and a storage medium. The model request method is applied to a first communication entity, and the method comprises the following steps: according to a target model description file, a model request message is sent to a second communication entity, and model request information is used for requesting a target model; and a model feedback message sent by the second communication entity is received, and the model feedback message comprises information of the target model.

Description

Technical Field

[0001] This application relates to the field of communication technology, such as a model request method, a model feedback method, a communication entity, and a storage medium. Background Technology

[0002] In the TS23.288 protocol, the Analytics Logical Function (AnLF) can discover the Model Training Logical Function (MTLF) from the Network Repository Function (NRF) and send a model retrieval request to the MTLF. The MTLF then sends its trained model to the AnLF based on this request. However, the model sent by the MTLF may not satisfy the AnLF. For example, the AnLF may not support running the model, the model's inference speed may be too slow, or the model quality may be problematic. Summary of the Invention

[0003] This application provides a model request method applied to a first communication entity, the method comprising:

[0004] Based on the target model description file, a model request message is sent to the second communication entity. The model request information is used to request the target model.

[0005] Receive the model feedback message sent by the second communication entity. The model feedback message includes information about the target model.

[0006] This application provides a model feedback method applied to a second communication entity, the method comprising:

[0007] Receive a model request message sent by the first communication entity. The model request information is a message used to request the target model, which is determined based on the target model description file.

[0008] Based on the model request message, determine the information of the target model;

[0009] A model feedback message is sent to the first communication entity. The model feedback message includes information about the target model.

[0010] This application provides a communication entity, including a processor; the processor is used to implement the method of any of the above embodiments when executing a computer program.

[0011] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method of any of the above embodiments.

[0012] Further details regarding the above embodiments and other aspects of this application, as well as their implementations, are provided in the accompanying drawings, detailed description, and claims. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a core network architecture provided in one embodiment;

[0014] Figure 2 This is a schematic diagram of an architecture for a communication entity to collect data, provided in one embodiment.

[0015] Figure 3 This is a schematic diagram of another communication entity's data collection architecture provided in one embodiment;

[0016] Figure 4 This is a schematic diagram of an architecture for a communication entity to provide data, provided in one embodiment.

[0017] Figure 5 This is a schematic diagram of another communication entity providing data according to one embodiment;

[0018] Figure 6 This is a schematic diagram of a 5G system architecture including ADRF provided in one embodiment;

[0019] Figure 7 This is a flowchart illustrating a model request method provided in one embodiment;

[0020] Figure 8 This is a flowchart illustrating a model feedback method provided in one embodiment;

[0021] Figure 9 This is a flowchart illustrating a document feedback method provided in one embodiment;

[0022] Figure 10 This is an interactive flowchart provided in Example 1;

[0023] Figure 11 This is an interactive flowchart provided in Example 2;

[0024] Figure 12 This is a flowchart illustrating a method for storing model description files according to one embodiment;

[0025] Figure 13 This is a flowchart illustrating another method for storing model description files provided in one embodiment;

[0026] Figure 14 This is an interactive schematic diagram of a method for storing model description files provided in one embodiment;

[0027] Figure 15This is a schematic diagram of the structure of a model request device provided in one embodiment;

[0028] Figure 16 This is a schematic diagram of another model request device provided in one embodiment;

[0029] Figure 17 This is a schematic diagram of the structure of a model feedback device provided in one embodiment;

[0030] Figure 18 This is a schematic diagram of the structure of a communication entity provided in one embodiment. Detailed Implementation

[0031] It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0032] The model request method and model feedback method provided in this application can be applied to various wireless communication systems, such as 5th-generation (5G) systems, long-term evolution (LTE) and 5G hybrid architecture systems, 5G New Radio (NR) systems, and new communication systems emerging in future communication development, such as 6th-generation (6G) systems. Optionally, the model request method and model feedback method can be applied to the core network of the communication system.

[0033] Figure 1 This is a schematic diagram of a core network architecture provided in one embodiment. For example... Figure 1 As shown, the User Equipment (UE) connects to the core network's network functions via the Radio Access Network (RAN). The RAN manages radio resources, transmits user data received through the N3 interface to the UE, and transmits user data from the UE through the N3 interface. The RAN maps Quality of Service (QoS) traffic between Dedicated Radio Bearer (DRB) and Protocol Data Unit (PDU) sessions.

[0034] The Access and Mobility Management Function (AMF) includes the following functions: registration management, connection management, reachability management, and mobility management. This function also performs access authentication and authorization. The AMF is a Non-Access Stratum (NAS) secure terminal that forwards SM NAS, etc., between the UE and the Session Management Function (SMF).

[0035] SMF includes the following functions: session establishment, modification, and release; UE network protocol (Internet Protocol, IP) address allocation and management (including optional authorization functions); selection and control of User Plane Functions (UPFs); and downlink data notification. SMF is associated with and controls the UPF through the N4 interface. SMF provides the UPF with Packet Detection Rules (PDRs) instructing how to detect user data traffic; it also provides Forwarding Action Rules (FARs), QoS Enforcement Rules (QERs), and Usage Reporting Rules (URRs) instructing the UPF how to perform user data traffic forwarding, QoS processing, and usage reporting on user data traffic detected using PDRs.

[0036] The UPF includes the following functions: serving as an anchor point for movement within / between Radio Access Technology (RAT), packet routing and forwarding, traffic usage reporting, user plane QoS processing, downlink packet buffering, and downlink data notification triggering. The GTP-U tunnel is used for the N3 interface between the RAN and the UPF. The GTP-U tunnel operates on a per-PDU session basis. For downlink traffic, the UPF binds the downlink traffic to the QoS traffic within the PDU session's GTP-U tunnel using the FAR received from the SMF. For uplink traffic, the RAN transmits user plane traffic to the QoS stream recognized by the UE.

[0037] The Policy Control Function (PCF) provides QoS policy rules to control plane functions for enforcement. The PCF translates Application Function (AF) requests into Policy and Charging Control (PCC) rules applicable to PDU sessions.

[0038] Unified Data Management (UDM) performs tasks such as generating 3GPP Authentication and Key Agreement (AKA) credentials, authorizing access based on subscription data, managing UE service element registration (e.g., storing AMF for UEs and SMF for UE PDU sessions), and subscription management. UDM accesses the Unified Data Repository (UDR) to retrieve UE subscription data and store the UE context in the UDR. UDM and UDR can be deployed together.

[0039] The model request method provided in this application embodiment can be applied to a first communication entity, and the model feedback method can be applied to a second communication entity. The first communication entity can be a 5G core network (5GC) element containing AnLF (such as NWDAF), or a Location Management Function (LMF), etc.; the second communication entity can be a 5GC network element containing MTLF (such as NWDAF), or an LMF, Analytic Data Repository Function (ADRF), etc. Furthermore, the third communication entity mentioned in this application embodiment can be a 5GC network element with storage capabilities, such as NRF, ADRF, etc.

[0040] NWDAF is a 5GC network element located in the control plane, which performs statistical data and machine learning related tasks in the 5GC. NWDAF can interact with different communication entities for various purposes: 1. Collect data based on event subscriptions provided by AMF, SMF, UPF, PCF, UDM, Network Slice Admission Control Function (NSACF), AF (directly or via Network Exposure Function (NEF)), and Operation Administration and Maintenance (OAM); 2. (Optional) Perform analysis and data collection using the Data Collection Coordination Function (DCCF); 3. Retrieve information from data repositories, for example, retrieving user-related UDRs via UDM or PFD information via NEF (Packet Flow Descriptions Function (PFDF)); 4. Collect location information data from Location Services (LCS) systems; 5. (Optional) Store and retrieve information from ADRF; 6. (Optional) Receive information from the Messaging Framework Adapter function. 7. Analyze and collect data for network elements; 8. Retrieve information about network elements, such as retrieving network element-related information from Network Repository Functions (NRFs); 9. Provide analytics to consumers on demand; 10. Provide bulk data related to analytics identifiers (IDs); 11. Provide accuracy information for analytics IDs; 12. Provide ML model accuracy information or ML model accuracy degradation indicators for machine learning (ML) models.

[0041] A single instance or multiple instances of NWDAF can be deployed in a Public Land Mobile Network (PLMN). An NWDAF can contain the following logical functions: AnLF, a logical function within the NWDAF used to perform inference, derive analytical information (i.e., derive statistical data and / or predictions based on analytical consumer requests), and expose analytical services (i.e., Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo); and MTLF, another logical function within the NWDAF used to train ML models and expose new training services (such as providing pre-trained ML models). An NWDAF can contain one MTLF or one AnLF, or both logical functions simultaneously. DCCF is also a network element on the 5GC control plane. DCCF is responsible for coordinating the collection and distribution of data requested by network element consumers. It prevents data sources from processing multiple subscriptions to the same data and prevents the sending of multiple notifications containing the same information due to incoordination of data consumer requests. DCCF is applicable to: NWDAFs that request data from data sources (such as for computational analysis); network element consumers that request analysis from NWDAF data sources; network element consumers that request data from ADRF data sources; and ADRFs that receive data from network element data sources.

[0042] Figure 2 This is a schematic diagram of an architecture for a communication entity to collect data, provided in one embodiment. Figure 3 This is a schematic diagram of another communication entity architecture for collecting data, provided in one embodiment. Figure 2 This demonstrates that NWDAF can collect data from any NF (also known as a data source). Figure 3 This demonstrates that NWDAF can collect data from any NF via DCCF. That is, Figure 2 and Figure 3 Two data collection structures are presented to facilitate NWDAF's data analysis and model training.

[0043] Figure 4 This is a schematic diagram of an architecture for a communication entity to provide data, provided in one embodiment. Figure 5 This is a schematic diagram of an architecture for another communication entity to provide data, provided in one embodiment. Figure 4 This demonstrates that NWDAF can provide data to any service network function (NF). Figure 5 This demonstrates that NWDAF can provide data to any NF via DCCF. The data here includes at least one of the following: analysis results, statistical results, and predictive information. That is, Figure 4 and Figure 5Two possible network data analytics exposure architectures are demonstrated for use by any network element consumer that subscribes to or requests analytics.

[0044] Figure 6 This is a schematic diagram of a 5G system architecture including ADRF provided in one embodiment. Figure 6 The 5G system architecture is shown to allow ADRF to store and retrieve collected data and perform analysis, supporting the following options: ADRF exposes Nadrf services for other 5GC network functions (such as NWDAF) to store and retrieve data, and data can be accessed through Nadrf services; based on network function requests or configurations on DCCF, DCCF can identify ADRF and interact with it directly or indirectly to request or store data, and the interaction can be direct or indirect.

[0045] For direct interaction, DCCF requests data to be stored in ADRF via the Nadrf service or Ndccf_DataManagement_Notify (as when ADRF requests DCCF to notify it of data collection). Additionally, DCCF retrieves data from ADRF via the Nadrf service.

[0046] For indirect interactions, DCCF requires the message framework to store data in ADRF via the Nadrf service or Nmfaf_3daDataManagement_Configure. The message framework may contain one or more adapters for conversion between 3GPP-defined protocols.

[0047] It should be noted that the internal logic of the message frame is not within the 3GPP scope; only MFAF and the interface between MFAF and other network functions defined by 3GPP are within the 3GPP scope.

[0048] Consumer network functions can specify in the request sent to DCCF that data provided by the data source needs to be stored in ADRF. ADRF stores data received directly from Nadrf_DataManagement_StorageRequest sent by the network function, or data received from Ndccf_DataManagement_Notify / Nmfaf_3caDataManagement_Notify or Nnwdaf_DataManagement_Notify sent by DCCF, MFAF or NWDAF.

[0049] In this application embodiment, a model request method, a model feedback method, a communication entity, and a storage medium are provided. By introducing model description files, model consumers can request model source files based on one or more selected target model description files, thereby effectively solving the problems of incompatibility and poor inference performance when directly requesting black-box model files.

[0050] The following describes the model request method, model feedback method, communication entity, and their technical effects.

[0051] Figure 7 This is a flowchart illustrating a model request method provided in one embodiment. For example... Figure 7 As shown, the method provided in this embodiment is applicable to a first communication entity. The first communication entity can be a 5GC network element containing AnLF (such as NWDAF), or LMF, etc. The method includes the following steps.

[0052] S710. Based on the target model description file, send a model request message to the second communication entity. The model request information is used to request the target model.

[0053] The target model description file is the model description file corresponding to the target model, used to describe information related to the target model. The target model description file is selected by the first communication entity from at least one model description file. There can be one or more target model description files.

[0054] The model request message is a message used to request the target model, determined based on the target model description file. The second communication entity is the owner of the target model. When there are multiple target model description files, there are also multiple target models. If the owners of these multiple target models are not unique, then the second communication entity is also not unique.

[0055] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0056] Among them, the model identifier is a unique identifier, and there are no duplicate model identifiers in the network; the model description file identifier is the ID of the model description file; the model owner information includes the model owner's address information and / or the model owner's identity information, such as equipment vendor ID and network element ID; the model licensing information is used to indicate the requester information with the qualification to obtain the model (such as the qualification to operate and use / shareable qualification), such as equipment vendor ID and network element ID.

[0057] In one embodiment, the model description file further includes at least one of the following:

[0058] Version number;

[0059] Model file size;

[0060] Model storage address information;

[0061] The model stores identity information, such as ADRFID;

[0062] Environmental requirements for model operation;

[0063] Model type, such as neural network, random forest, support vector machine, etc.;

[0064] Model task type, such as classification model, regression model, generative model, segmentation model, etc.;

[0065] Model usage information, such as analysis ID;

[0066] Is it a linear model or a nonlinear model?

[0067] Can it be used for horizontal federated learning?

[0068] Can it be used for vertical federated learning?

[0069] Can it be used for multi-level federated learning?

[0070] Can it be used for federated learning?

[0071] Number of model parameters;

[0072] The data type, format, and dimensional information of the model input;

[0073] The data type, format, and dimensional information output by the model;

[0074] Data preprocessing requirements, such as whether to normalize, etc.;

[0075] The model's topology, such as the number of layers and the specific content of each layer (e.g., fully connected layers, pooling layers, etc.), and the number of parameters in each layer (e.g., the number of convolutional kernels, activation function types, etc.).

[0076] Relevant information during model training includes at least one of the following: loss function type (e.g., cross-entropy), learning rate used during training, batch size, optimizer (e.g., stochastic gradient descent (SGD), Adam optimizer), regularization parameters used during training (e.g., L2 regularization, proportion of random dropout), training duration, model training time or number of iterations, hardware configuration during training (e.g., GPU model, Tensor Processing Unit (TPU) usage, etc.), and dataset information used to train the model (e.g., network elements from which the data originates, dataset name, dataset version, etc.).

[0077] Data augmentation information, such as whether data augmentation techniques were used and their specific methods (rotation, cropping, normalization, etc.);

[0078] Does it have residual connections and attention mechanisms?

[0079] The computational power requirements for model inference and the model inference speed (e.g., the inference speed for each input sample) under different computing power conditions;

[0080] Model testing information, such as the region information of model testing, and the achieved accuracy, precision, recall, F1 score, mean squared error, mean absolute percentage error, etc.

[0081] Example code;

[0082] License information;

[0083] The model uses area information, such as tracking area information.

[0084] In one embodiment, the model request information may carry a model identifier, so that the second communication entity can determine the model requested by the first communication entity based on the model identifier.

[0085] In one embodiment, before step S710 is executed, the first communication entity may also send a file request message to the second communication entity or the third communication entity; receive a file feedback message sent by the second communication entity or the third communication entity, the file feedback message containing at least one of the following: at least one model description file, having a link to at least one model description file; selecting one or more target model description files from at least one model description file.

[0086] The file request information may include at least one of the following: requester information, model requirement information. The requester information is used by the second or third communication entity to verify whether the requester is eligible to obtain the model; the model requirement information indicates the model-related information requested by the requester.

[0087] S720: Receive the model feedback message sent by the second communication entity. The model feedback message includes information about the target model.

[0088] In one embodiment, the information of the target model includes at least one of the following: a model file of the target model, having a link to the target model.

[0089] In one embodiment, after step S720 is executed, the first communication entity can further obtain the target model based on the information of the target model, and perform inference based on the target model; determine evaluation information based on the inference result, the evaluation information being used to describe the accuracy of the target model description file; and send an evaluation message to the second communication entity, the evaluation message including the evaluation information. In this way, the accuracy of the model description file can be improved.

[0090] Figure 8 This is a flowchart illustrating a model feedback method provided in one embodiment. Figure 8 As shown, the method provided in this embodiment is applicable to a second communication entity. The second communication entity can be a 5GC network element containing an MTLF (such as an NWDAF), or an LMF, ADRF, etc. The method includes the following steps.

[0091] S810: Receive a model request message sent by the first communication entity. The model request message is a message used to request the target model, which is determined based on the target model description file.

[0092] The target model description file is the model description file corresponding to the target model, used to describe information related to the target model. The target model description file is selected by the first communication entity from at least one model description file. There can be one or more target model description files.

[0093] The model request message is a message used to request the target model, based on the target model description file. The second communication entity is the owner of the target model.

[0094] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0095] Among them, the model identifier is a unique identifier, and there are no duplicate model identifiers in the network; the model description file identifier is the ID of the model description file; the model owner information includes the model owner's address information and / or the model owner's identity information, such as equipment vendor ID and network element ID; the model licensing information is used to indicate the requester information with the qualification to obtain the model (such as the qualification to operate and use / shareable qualification), such as equipment vendor ID and network element ID.

[0096] In one embodiment, the model description file further includes at least one of the following:

[0097] Version number;

[0098] Model file size;

[0099] Model storage address information;

[0100] The model stores identity information, such as ADRFID;

[0101] Environmental requirements for model operation;

[0102] Model type, such as neural network, random forest, support vector machine, etc.;

[0103] Model task type, such as classification model, regression model, generative model, segmentation model, etc.;

[0104] Model usage information, such as analysis ID;

[0105] Is it a linear model or a nonlinear model?

[0106] Can it be used for horizontal federated learning?

[0107] Can it be used for vertical federated learning?

[0108] Can it be used for multi-level federated learning?

[0109] Can it be used for federated learning?

[0110] Number of model parameters;

[0111] The data type, format, and dimensional information of the model input;

[0112] The data type, format, and dimensional information output by the model;

[0113] Data preprocessing requirements, such as whether to normalize, etc.;

[0114] The model's topology, such as the number of layers and the specific content of each layer (e.g., fully connected layers, pooling layers, etc.), and the number of parameters in each layer (e.g., the number of convolutional kernels, activation function types, etc.).

[0115] Relevant information during model training includes at least one of the following: loss function type (e.g., cross-entropy), learning rate used during training, batch size, optimizer (e.g., stochastic gradient descent (SGD), Adam optimizer), regularization parameters used during training (e.g., L2 regularization, proportion of random dropout), training duration, model training time or number of iterations, hardware configuration during training (e.g., GPU model, Tensor Processing Unit (TPU) usage, etc.), and dataset information used to train the model (e.g., network elements from which the data originates, dataset name, dataset version, etc.).

[0116] Data augmentation information, such as whether data augmentation techniques were used and their specific methods (rotation, cropping, normalization, etc.);

[0117] Does it have residual connections and attention mechanisms?

[0118] The computational power requirements for model inference and the model inference speed (e.g., the inference speed for each input sample) under different computing power conditions;

[0119] Model testing information, such as the region information of model testing, and the achieved accuracy, precision, recall, F1 score, mean squared error, mean absolute percentage error, etc.

[0120] Example code;

[0121] License information;

[0122] The model uses area information, such as tracking area information.

[0123] In one embodiment, the model request information may carry a model identifier, so that the second communication entity can determine the model requested by the first communication entity, i.e., the target model, based on the model identifier.

[0124] In one embodiment, before step S810 is executed, the second communication entity may further receive a file request message sent by the first communication entity; based on the file request message, it sends a file feedback message to the first communication entity, the file feedback message containing at least one of the following: at least one model description file, having a link to at least one model description file. This causes the first communication entity to select one or more target model description files from at least one model description file.

[0125] The file request information may include at least one of the following: requester information, model requirement information. The requester information is used by the second communication entity to verify whether the requester is eligible to obtain the model; the model requirement information indicates the model-related information requested by the requester.

[0126] In one embodiment, at least one model description file is a file generated by the second communication entity after training the model; and / or, at least one model description file is a file obtained by the second communication entity from locally stored model description files.

[0127] S820. Determine the information of the target model based on the model request message.

[0128] In one embodiment, the information of the target model includes at least one of the following: a model file of the target model, having a link to the target model.

[0129] S830. Send a model feedback message to the first communication entity. The model feedback message includes information about the target model.

[0130] In one embodiment, after step S830 is executed, the second communication entity can also receive an evaluation message sent by the first communication entity. The evaluation message includes evaluation information used to describe the accuracy of the target model description file. Based on the evaluation information, the target model is retrained or the target model description file is regenerated. In this way, the accuracy of the model description file can be improved.

[0131] Optionally, for model description files (including model description files generated by the second communication entity after training the model, and / or model description files stored locally by the second communication entity), the second communication entity may store these model description files in the third communication entity. Specifically, the second communication entity sends a file storage request message to the third communication entity, the file storage request message including at least one of the following: the model description file generated by the second communication entity after training the model, the model description file stored locally by the second communication entity; or receives a file storage feedback message sent by the third communication entity.

[0132] Storage feedback information is used to indicate the storage status of the model description file (e.g., storage successful, storage failed).

[0133] Alternatively, after the second communication entity stores the model description files in the third communication entity, the second communication entity can also manage these model description files. Specifically, the second communication entity sends a file management request message to the third communication entity. The file management request message is used to instruct the model description files to be managed to perform at least one of the following operations: extraction operation, deletion operation; and receiving a file management feedback message sent by the third communication entity.

[0134] File management request messages include at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier. File management feedback messages are used to indicate the management status of model description files (e.g., successful deletion, failed deletion, successful extraction including the extracted model description file, and failed extraction).

[0135] Figure 9 This is a flowchart illustrating a document feedback method provided in one embodiment. For example... Figure 9 As shown, the method provided in this embodiment is applicable to a third communication entity. The third communication entity can be a 5GC network element with storage capabilities, such as an NRF or ADRF. The method includes the following steps.

[0136] S910, Receive the file request message sent by the first communication entity.

[0137] The file request information may include at least one of the following: requester information, model requirement information. The requester information is used by the third-party communication entity to verify whether the requester is eligible to obtain the model; the model requirement information indicates the model-related information requested by the requester.

[0138] S920. Based on the file request message, send a file feedback message to the first communication entity. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0139] In one embodiment, at least one model description file is a file obtained by a third communication entity from locally stored model description files.

[0140] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0141] Among them, the model identifier is a unique identifier, and there are no duplicate model identifiers in the network; the model description file identifier is the ID of the model description file; the model owner information includes the model owner's address information and / or the model owner's identity information, such as equipment vendor ID and network element ID; the model licensing information is used to indicate the requester information with the qualification to obtain the model (such as the qualification to operate and use / shareable qualification), such as equipment vendor ID and network element ID.

[0142] In one embodiment, the model description file further includes at least one of the following:

[0143] Version number;

[0144] Model file size;

[0145] Model storage address information;

[0146] The model stores identity information, such as ADRFID;

[0147] Environmental requirements for model operation;

[0148] Model type, such as neural network, random forest, support vector machine, etc.;

[0149] Model task type, such as classification model, regression model, generative model, segmentation model, etc.;

[0150] Model usage information, such as analysis ID;

[0151] Is it a linear model or a nonlinear model?

[0152] Can it be used for horizontal federated learning?

[0153] Can it be used for vertical federated learning?

[0154] Can it be used for multi-level federated learning?

[0155] Can it be used for federated learning?

[0156] Number of model parameters;

[0157] The data type, format, and dimensional information of the model input;

[0158] The data type, format, and dimensional information output by the model;

[0159] Data preprocessing requirements, such as whether to normalize, etc.;

[0160] The model's topology, such as the number of layers and the specific content of each layer (e.g., fully connected layers, pooling layers, etc.), and the number of parameters in each layer (e.g., the number of convolutional kernels, activation function types, etc.).

[0161] Relevant information during model training includes at least one of the following: loss function type (e.g., cross-entropy), learning rate used during training, batch size, optimizer (e.g., stochastic gradient descent (SGD), Adam optimizer), regularization parameters used during training (e.g., L2 regularization, proportion of random dropout), training duration, model training time or number of iterations, hardware configuration during training (e.g., GPU model, Tensor Processing Unit (TPU) usage, etc.), and dataset information used to train the model (e.g., network elements from which the data originates, dataset name, dataset version, etc.).

[0162] Data augmentation information, such as whether data augmentation techniques were used and their specific methods (rotation, cropping, normalization, etc.);

[0163] Does it have residual connections and attention mechanisms?

[0164] The computational power requirements for model inference and the model inference speed (e.g., the inference speed for each input sample) under different computing power conditions;

[0165] Model testing information, such as the region information of model testing, and the achieved accuracy, precision, recall, F1 score, mean squared error, mean absolute percentage error, etc.

[0166] Example code;

[0167] License information;

[0168] The model uses area information, such as tracking area information.

[0169] Optionally, the model description file stored locally by the third communication entity is a file sent by the second communication entity and stored in the third communication entity. Specifically, the third communication entity receives a file storage request message sent by the second communication entity, which includes at least one of the following: a model description file generated by the second communication entity after training the model; a model description file stored locally by the second communication entity; or a model description file carried in the file storage request message stored locally by the third communication entity, and sends a file storage feedback message to the second communication entity.

[0170] Storage feedback information is used to indicate the storage status of the model description file (e.g., storage successful, storage failed).

[0171] Alternatively, after the third communication entity sends a file storage feedback message to the second communication entity, the third communication entity can also manage these model description files. Specifically, the third communication entity receives a file management request message sent by the second communication entity. The file management request message instructs the model description file to be managed to perform at least one of the following operations: extraction operation, deletion operation; manage the model description file according to the file management request message, and send a file management feedback message to the second communication entity.

[0172] File management request messages include at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier. File management feedback messages are used to indicate the management status of model description files (e.g., successful deletion, failed deletion, successful extraction including the extracted model description file, and failed extraction).

[0173] Below are some examples to illustrate the methods provided in this application.

[0174] Example 1: Interaction between the first and second communication entities.

[0175] Figure 10 This is an interactive flowchart provided in Example 1. For example... Figure 10 As shown, the steps include the following.

[0176] S1001, The first communication entity sends a file request message to the second communication entity.

[0177] The file request information may include at least one of the following: requester information (such as the requester's equipment vendor ID, network element ID), and model requirement information (such as analysis ID, model topology, etc.). The requester information is used by the second communication entity to verify whether the requester is eligible to obtain the model; the model requirement information indicates the model-related information requested by the requester.

[0178] S1002, The second communication entity receives the file request message sent by the first communication entity.

[0179] Based on the document request information, the second communication entity can understand which requester needs what kind of model.

[0180] S1003. The second communication entity sends a file feedback message to the first communication entity according to the file request message. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0181] The second communication entity can obtain at least one model description file from locally stored model description files based on the file request message (such as model requirement information included in the file request message), and / or, the second communication entity can train a model and generate a corresponding model description file based on the file request message (such as model requirement information included in the file request message). Then, the second communication entity verifies that the model corresponding to the matched / generated model description file can be obtained by the requester based on the file request message (such as requester information included in the file request message).

[0182] Optionally, to improve interaction efficiency, the second communication entity can be matched locally first, and then the model can be trained if no match is found.

[0183] After the second communication entity determines at least one model description file, it sends a file feedback message to the first communication entity. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0184] S1004, The first communication entity receives the file feedback message sent by the second communication entity.

[0185] If the file feedback message includes at least one model description file, the first communication entity can directly parse the file feedback message to obtain at least one model description file; if the file feedback message includes a link with at least one model description file, the first communication entity can obtain at least one model description file based on the link with at least one model description file.

[0186] S1005. The first communication entity selects one or more target model description files from at least one model description file.

[0187] In one embodiment, the first communication entity may select one or more target model description files from at least one model description file based on internal logic and business requirements.

[0188] S1006. The first communication entity sends a model request message to the second communication entity according to the target model description file. The model request message is used to request the target model.

[0189] The target model description file is a model description file corresponding to the target model, used to describe information related to the target model. There can be one or multiple target model description files.

[0190] The model request information can carry a model identifier, so that the second communication entity can determine the model requested by the first communication entity, i.e., the target model, based on the model identifier.

[0191] In Example 1, for a second communication entity, the model description files stored / generated locally by the second communication entity are all files corresponding to its own models.

[0192] S1007. The second communication entity determines the information of the target model based on the model request message.

[0193] The target model information includes at least one of the following: the target model's model file, and a link to the target model.

[0194] S1008, the second communication entity sends a model feedback message to the first communication entity. The model feedback message includes information about the target model.

[0195] S1009, The first communication entity receives the model feedback message sent by the second communication entity.

[0196] S1010. The first communication entity obtains the target model based on the information of the target model and performs inference based on the target model.

[0197] If the target model information includes the target model file, the first communication entity can directly parse the target model from the target model information to obtain the target model; if the target model information includes a link to the target model, the first communication entity can obtain the target model based on the link to the target model.

[0198] S1011. The first communication entity determines the evaluation information based on the reasoning results. The evaluation information is used to describe the accuracy of the target model description file.

[0199] S1012, The first communication entity sends an evaluation message to the second communication entity. The evaluation message includes evaluation information.

[0200] S1013, The second communication entity receives the evaluation message sent by the first communication entity.

[0201] S1014. The second communication entity retrains the target model or regenerates the target model description file based on the evaluation information.

[0202] This can improve the accuracy of the model description file.

[0203] Example 2: Interaction between the first communication entity, the second communication entity, and the third communication entity.

[0204] Figure 11 This is an interactive flowchart provided in Example 2. For example... Figure 11 As shown, the steps include the following.

[0205] S1101, The first communication entity sends a file request message to the third communication entity.

[0206] The file request information may include at least one of the following: requester information (such as the requester's equipment vendor ID, network element ID), and model requirement information (such as analysis ID, model topology, etc.). The requester information is used by the third-party communication entity to verify whether the requester is eligible to obtain the model; the model requirement information indicates the model-related information requested by the requester.

[0207] S1102, The third communication entity receives the file request message sent by the first communication entity.

[0208] Based on the document request information, the third communication entity can understand which requester needs what kind of model.

[0209] S1103. The third communication entity sends a file feedback message to the first communication entity according to the file request message. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0210] The third communication entity can obtain at least one model description file from locally stored model description files based on the file request message (such as model requirement information included in the file request message). Then, the third communication entity verifies that the model corresponding to the above-matched / generated model description file can be obtained by the requester based on the file request message (such as requester information included in the file request message).

[0211] After the third communication entity determines at least one model description file, it sends a file feedback message to the first communication entity. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0212] S1104, The first communication entity receives the file feedback message sent by the third communication entity.

[0213] If the file feedback message includes at least one model description file, the first communication entity can directly parse the file feedback message to obtain at least one model description file; if the file feedback message includes a link with at least one model description file, the first communication entity can obtain at least one model description file based on the link with at least one model description file.

[0214] S1105. The first communication entity selects one or more target model description files from at least one model description file.

[0215] In one embodiment, the first communication entity may select one or more target model description files from at least one model description file based on internal logic and business requirements.

[0216] S1106. The first communication entity sends a model request message to the second communication entity according to the target model description file. The model request information is used to request the target model.

[0217] The target model description file is a model description file corresponding to the target model, used to describe information related to the target model. There can be one or multiple target model description files.

[0218] The model request information can carry a model identifier, so that the second communication entity can determine the model requested by the first communication entity, i.e., the target model, based on the model identifier.

[0219] In Example 2, since the third communication entity may store model description files corresponding to different models stored by the second communication entity, the target model may have multiple owners, that is, the number of second communication entities is multiple.

[0220] S1107. The second communication entity determines the information of the target model based on the model request message.

[0221] The target model information includes at least one of the following: the target model's model file, and a link to the target model.

[0222] S1108, The second communication entity sends a model feedback message to the first communication entity. The model feedback message includes information about the target model.

[0223] S1109. The first communication entity receives the model feedback message sent by the second communication entity.

[0224] S1110. The first communication entity obtains the target model based on the information of the target model and performs inference based on the target model.

[0225] If the target model information includes the target model file, the first communication entity can directly parse the target model from the target model information to obtain the target model; if the target model information includes a link to the target model, the first communication entity can obtain the target model based on the link to the target model.

[0226] S1111 The first communication entity determines the evaluation information based on the reasoning results. The evaluation information is used to describe the accuracy of the target model description file.

[0227] S1112. The first communication entity sends an evaluation message to the second communication entity. The evaluation message includes evaluation information.

[0228] S1113. The second communication entity receives the evaluation message sent by the first communication entity.

[0229] S1114. The second communication entity retrains the target model or regenerates the target model description file based on the evaluation information.

[0230] This can improve the accuracy of the model description file.

[0231] Figure 12 This is a flowchart illustrating a method for storing model description files according to one embodiment. Figure 12 As shown, the method provided in this embodiment is applicable to a second communication entity. The second communication entity can be a 5GC network element containing an MTLF (such as an NWDAF), or an LMF, ADRF, etc. The method includes the following steps.

[0232] S1210. Send a file storage request message to the third communication entity. The file storage request message includes at least one of the following: a model description file generated by the second communication entity after training the model, or a model description file stored locally by the second communication entity.

[0233] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0234] Among them, the model identifier is a unique identifier, and there are no duplicate model identifiers in the network; the model description file identifier is the ID of the model description file; the model owner information includes the model owner's address information and / or the model owner's identity information, such as equipment vendor ID and network element ID; the model licensing information is used to indicate the requester information with the qualification to obtain the model (such as the qualification to operate and use / shareable qualification), such as equipment vendor ID and network element ID.

[0235] In one embodiment, the model description file further includes at least one of the following:

[0236] Version number;

[0237] Model file size;

[0238] Model storage address information;

[0239] The model stores identity information, such as ADRFID;

[0240] Environmental requirements for model operation;

[0241] Model type, such as neural network, random forest, support vector machine, etc.;

[0242] Model task type, such as classification model, regression model, generative model, segmentation model, etc.;

[0243] Model usage information, such as analysis ID;

[0244] Is it a linear model or a nonlinear model?

[0245] Can it be used for horizontal federated learning?

[0246] Can it be used for vertical federated learning?

[0247] Can it be used for multi-level federated learning?

[0248] Can it be used for federated learning?

[0249] Number of model parameters;

[0250] The data type, format, and dimensional information of the model input;

[0251] The data type, format, and dimensional information output by the model;

[0252] Data preprocessing requirements, such as whether to normalize, etc.;

[0253] The model's topology, such as the number of layers and the specific content of each layer (e.g., fully connected layers, pooling layers, etc.), and the number of parameters in each layer (e.g., the number of convolutional kernels, activation function types, etc.).

[0254] Relevant information during model training includes at least one of the following: loss function type (e.g., cross-entropy), learning rate used during training, batch size, optimizer (e.g., stochastic gradient descent (SGD), Adam optimizer), regularization parameters used during training (e.g., L2 regularization, proportion of random dropout), training duration, model training time or number of iterations, hardware configuration during training (e.g., GPU model, Tensor Processing Unit (TPU) usage, etc.), and dataset information used to train the model (e.g., network elements from which the data originates, dataset name, dataset version, etc.).

[0255] Data augmentation information, such as whether data augmentation techniques were used and their specific methods (rotation, cropping, normalization, etc.);

[0256] Does it have residual connections and attention mechanisms?

[0257] The computational power requirements for model inference and the model inference speed (e.g., the inference speed for each input sample) under different computing power conditions;

[0258] Model testing information, such as the region information of model testing, and the achieved accuracy, precision, recall, F1 score, mean squared error, mean absolute percentage error, etc.

[0259] Example code;

[0260] License information;

[0261] The model uses area information, such as tracking area information.

[0262] S1220: Receive the file storage feedback message sent by the third communication entity.

[0263] Storage feedback information is used to indicate the storage status of the model description file (e.g., storage successful, storage failed).

[0264] Optionally, after the second communication entity stores the model description files in the third communication entity, the second communication entity can also manage these model description files. Specifically, the second communication entity sends a file management request message to the third communication entity. The file management request message is used to instruct the model description files to be managed to perform at least one of the following operations: extraction operation, deletion operation; and receiving a file management feedback message sent by the third communication entity.

[0265] File management request messages include at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier. File management feedback messages are used to indicate the management status of model description files (e.g., successful deletion, failed deletion, successful extraction including the extracted model description file, and failed extraction).

[0266] Figure 13 This is a flowchart illustrating another method for storing model description files provided in one embodiment. For example... Figure 13 As shown, the method provided in this embodiment is applicable to a third communication entity. The third communication entity can be a 5GC network element with storage capabilities, such as an NRF or ADRF. The method includes the following steps.

[0267] S1310. Receive a file storage request message sent by the second communication entity. The file storage request message includes at least one of the following: a model description file generated by the second communication entity after training the model, or a model description file stored locally by the second communication entity.

[0268] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0269] Among them, the model identifier is a unique identifier, and there are no duplicate model identifiers in the network; the model description file identifier is the ID of the model description file; the model owner information includes the model owner's address information and / or the model owner's identity information, such as equipment vendor ID and network element ID; the model licensing information is used to indicate the requester information with the qualification to obtain the model (such as the qualification to operate and use / shareable qualification), such as equipment vendor ID and network element ID.

[0270] In one embodiment, the model description file further includes at least one of the following:

[0271] Version number;

[0272] Model file size;

[0273] Model storage address information;

[0274] The model stores identity information, such as ADRFID;

[0275] Environmental requirements for model operation;

[0276] Model type, such as neural network, random forest, support vector machine, etc.;

[0277] Model task type, such as classification model, regression model, generative model, segmentation model, etc.;

[0278] Model usage information, such as analysis ID;

[0279] Is it a linear model or a nonlinear model?

[0280] Can it be used for horizontal federated learning?

[0281] Can it be used for vertical federated learning?

[0282] Can it be used for multi-level federated learning?

[0283] Can it be used for federated learning?

[0284] Number of model parameters;

[0285] The data type, format, and dimensional information of the model input;

[0286] The data type, format, and dimensional information output by the model;

[0287] Data preprocessing requirements, such as whether to normalize, etc.;

[0288] The model's topology, such as the number of layers and the specific content of each layer (e.g., fully connected layers, pooling layers, etc.), and the number of parameters in each layer (e.g., the number of convolutional kernels, activation function types, etc.).

[0289] Relevant information during model training includes at least one of the following: loss function type (e.g., cross-entropy), learning rate used during training, batch size, optimizer (e.g., stochastic gradient descent (SGD), Adam optimizer), regularization parameters used during training (e.g., L2 regularization, proportion of random dropout), training duration, model training time or number of iterations, hardware configuration during training (e.g., GPU model, Tensor Processing Unit (TPU) usage, etc.), and dataset information used to train the model (e.g., network elements from which the data originates, dataset name, dataset version, etc.).

[0290] Data augmentation information, such as whether data augmentation techniques were used and their specific methods (rotation, cropping, normalization, etc.);

[0291] Does it have residual connections and attention mechanisms?

[0292] The computational power requirements for model inference and the model inference speed (e.g., the inference speed for each input sample) under different computing power conditions;

[0293] Model testing information, such as the region information of model testing, and the achieved accuracy, precision, recall, F1 score, mean squared error, mean absolute percentage error, etc.

[0294] Example code;

[0295] License information;

[0296] The model uses area information, such as tracking area information.

[0297] S1320, Store the model description file and send a file storage feedback message to the second communication entity.

[0298] Storage feedback information is used to indicate the storage status of the model description file (e.g., storage successful, storage failed).

[0299] Optionally, after the third communication entity sends a file storage feedback message to the second communication entity, the third communication entity can also manage these model description files. Specifically, the third communication entity receives a file management request message sent by the second communication entity. The file management request message instructs the model description file to be managed to perform at least one of the following operations: extraction operation, deletion operation; manage the model description file according to the file management request message, and send a file management feedback message to the second communication entity.

[0300] File management request messages include at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier. File management feedback messages are used to indicate the management status of model description files (e.g., successful deletion, failed deletion, successful extraction including the extracted model description file, and failed extraction).

[0301] Figure 14 This is an interactive schematic diagram illustrating a method for storing model description files according to one embodiment. For example... Figure 14 As shown, the steps include the following.

[0302] S1401, the second communication entity sends a file storage request message to the third communication entity. The file storage request message includes at least one of the following: a model description file generated by the second communication entity after training the model, or a model description file stored locally by the second communication entity.

[0303] S1402, The third communication entity receives the file storage request message sent by the second communication entity.

[0304] S1403, Third Communication Entity Storage Model Description File.

[0305] S1404, The third communication entity sends a file storage feedback message to the second communication entity.

[0306] S1405, The second communication entity receives the file storage feedback message sent by the second communication entity.

[0307] S1406, The second communication entity sends a first file management request message to the third communication entity.

[0308] The first file management request message is used to instruct the extraction operation of the model description file to be managed. The first file management request message includes at least one of the following: model identifier, model description file identifier, and model description file extraction identifier.

[0309] S1407, The third communication entity sends the first file management feedback message to the second communication entity.

[0310] The first file management feedback message includes at least one of the following: the extracted model description file, with a link to the extracted model description file.

[0311] S1408, The second communication entity sends a second file management request message to the third communication entity.

[0312] The second file management request message is used to instruct the deletion of a model description file to be managed. The second file management request message includes at least one of the following: model identifier, model description file identifier, or model description file deletion identifier.

[0313] S1409, The third communication entity sends a second file management feedback message to the second communication entity.

[0314] The steps S1406-S1407 and S1408-S1409 above are optional steps.

[0315] Figure 15 This is a schematic diagram of a model request device provided in one embodiment. This device can be configured in a first communication entity, such as... Figure 15 As shown, the device includes a transmitting module 1501 and a receiving module 1502.

[0316] The sending module 1501 is configured to send a model request message to the second communication entity according to the target model description file. The model request message is used to request the target model.

[0317] The receiving module 1502 is configured to receive model feedback messages sent by the second communication entity. The model feedback messages include information about the target model.

[0318] The model request device provided in this embodiment is for implementing... Figure 7 The model request method of the embodiment shown is implemented in a similar principle and with similar technical effects to the above embodiments, and will not be repeated here.

[0319] In one embodiment, the information of the target model includes at least one of the following: a model file of the target model, having a link to the target model.

[0320] In one embodiment, combined with Figure 15 , Figure 16 This is a schematic diagram of another model request device provided in one embodiment, such as... Figure 16 As shown, it also includes: processing module 1503.

[0321] The sending module 1501 is also configured to send a file request message to a second communication entity or a third communication entity;

[0322] The receiving module 1502 is further configured to receive a file feedback message sent by a second communication entity or a third communication entity, wherein the file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file;

[0323] Processing module 1503 is configured to select one or more target model description files from at least one model description file.

[0324] In one embodiment, the file request information includes at least one of the following: requester information, model requirement information.

[0325] In one embodiment, model requirement information is used to indicate the model-related information requested by the requester.

[0326] In one embodiment, the processing module 1503 is further configured to obtain the target model based on the information of the target model, and perform inference based on the target model; determine evaluation information based on the inference result, the evaluation information being used to describe the accuracy of the target model description file;

[0327] The sending module 1501 is also configured to send an evaluation message to the second communication entity, the evaluation message including evaluation information.

[0328] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0329] In one embodiment, model licensing information is used to indicate information about the requester who is qualified to acquire the model.

[0330] In one embodiment, the model description file further includes at least one of the following information: version number; model file size; model storage address information; model storage identity information; environmental requirements for model operation; model type; model task type; model purpose information; whether it is a linear or nonlinear model; whether it can be used for horizontal federated learning; whether it can be used for vertical federated learning; whether it can be used for multi-layer federated learning; whether it can be used for federated learning; number of model parameters; data type, format, and dimensional information of model input; data type, format, and dimensional information of model output; data preprocessing requirements; model topology; relevant information during model training; data augmentation information; whether there are residual connections and attention mechanisms; computational power requirements and inference speed of model inference under different computing power; model testing information; sample code; license information; and model usage area information.

[0331] Figure 17 This is a schematic diagram of a model feedback device provided in one embodiment. This device can be configured in a second communication entity, such as... Figure 17 As shown, the device includes: a receiving module 1701, a processing module 1702, and a transmitting module 1703.

[0332] The receiving module 1701 is configured to receive a model request message sent by the first communication entity. The model request message is a message used to request the target model, which is determined based on the target model description file.

[0333] Processing module 1702 is configured to determine the information of the target model based on the model request message;

[0334] The sending module 1703 is configured to send a model feedback message to the first communication entity. The model feedback message includes information about the target model.

[0335] The model feedback device provided in this embodiment is for implementing... Figure 8 The model feedback method of the embodiment shown is similar in principle and technical effect to the model feedback device provided in this embodiment, and will not be repeated here.

[0336] In one embodiment, the information of the target model includes at least one of the following: a model file of the target model, having a link to the target model.

[0337] In one embodiment, the receiving module 1701 is further configured to receive a file request message sent by the first communication entity;

[0338] The sending module 1703 is further configured to send a file feedback message to the first communication entity based on the file request message. The file feedback message contains at least one of the following: at least one model description file, and has a link to at least one model description file.

[0339] In one embodiment, the file request information includes at least one of the following: requester information, model requirement information.

[0340] In one embodiment, model requirement information is used to indicate the model-related information requested by the requester.

[0341] In one embodiment, at least one model description file is a file generated by the second communication entity after training the model; and / or, at least one model description file is a file obtained by the second communication entity from locally stored model description files.

[0342] In one embodiment, the receiving module 1701 is further configured to receive an evaluation message sent by the first communication entity, the evaluation message including evaluation information used to describe the accuracy of the target model description file;

[0343] The processing module 1702 is also configured to retrain the target model or regenerate the target model description file based on the evaluation information.

[0344] In one embodiment, the model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

[0345] In one embodiment, model licensing information is used to indicate information about the requester who is qualified to acquire the model.

[0346] In one embodiment, the model description file further includes at least one of the following information: version number; model file size; model storage address information; model storage identity information; environmental requirements for model operation; model type; model task type; model purpose information; whether it is a linear or nonlinear model; whether it can be used for horizontal federated learning; whether it can be used for vertical federated learning; whether it can be used for multi-layer federated learning; whether it can be used for federated learning; number of model parameters; data type, format, and dimensional information of model input; data type, format, and dimensional information of model output; data preprocessing requirements; model topology; relevant information during model training; data augmentation information; whether there are residual connections and attention mechanisms; computational power requirements and inference speed of model inference under different computing power; model testing information; sample code; license information; and model usage area information.

[0347] In one embodiment, the sending module 1703 is further configured to send a file storage request message to a third communication entity. The file storage request message includes at least one of the following: a model description file generated by the second communication entity after training the model, and a model description file stored locally by the second communication entity.

[0348] The receiving module 1701 is also configured to receive file storage feedback messages sent by a third communication entity.

[0349] In one embodiment, the sending module 1703 is further configured to send a file management request message to a third communication entity. The file management request message is used to instruct the model description file to be managed to perform at least one of the following operations: extraction operation, deletion operation;

[0350] The receiving module 1701 is also configured to receive file management feedback messages sent by a third communication entity.

[0351] In one embodiment, the file management request message includes at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier.

[0352] This application also provides a communication entity, including a processor, which is configured to implement the method provided in any embodiment of this application when executing a computer program. The communication entity in this embodiment can be a first communication entity, a second communication entity, or a third communication entity.

[0353] Figure 18 This is a schematic diagram of the structure of a communication entity provided in one embodiment. For example... Figure 18 As shown, the communication entity includes a processor 60, a memory 61, and a communication interface 62; the number of processors 60 in the communication entity can be one or more. Figure 18 Taking a processor 60 as an example; the processor 60, memory 61, and communication interface 62 in the communication entity can be connected via a bus or other means. Figure 18 Taking the bus connection as an example, a bus can refer to one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus that uses any of the various bus architectures.

[0354] The memory 61, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the embodiments of this application. The processor 60 executes at least one functional application and data processing of the communication entity by running the software programs, instructions, and modules stored in the memory 61, thereby implementing the methods described above.

[0355] Memory 61 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on terminal usage. Furthermore, memory 61 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, memory 61 may include memory remotely located relative to processor 60, which can be connected to a communication entity via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, networks, mobile communication networks, and combinations thereof.

[0356] Communication interface 62 can be configured to receive and send data.

[0357] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods provided in any embodiment of this application.

[0358] The computer storage medium in this application embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. Computer-readable storage media include (a non-exhaustive list): electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), electrically erasable, programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0359] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, the data signals carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0360] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, radio frequency (RF), or any suitable combination thereof.

[0361] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination of programming languages, including object-oriented programming languages ​​(such as Java, Smalltalk, C++, Ruby, and Go) and conventional procedural programming languages ​​(such as the "C" language or similar programming languages). The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network (including a Local Area Network (LAN) or a Wide Area Network (WAN)), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0362] Those skilled in the art will understand that the term user terminal encompasses any suitable type of wireless user equipment, such as mobile phones, portable data processing devices, portable web browsers, or vehicle-mounted mobile stations.

[0363] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although this application is not limited thereto.

[0364] Embodiments of this application can be implemented by executing computer program instructions through the data processor of a mobile device, for example, in a processor entity, or through hardware, or through a combination of software and hardware. The computer program instructions can be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0365] Any block diagram of logical flow in the accompanying drawings of this application may represent program steps, or may represent interconnected logic circuits, modules, and functions, or may represent a combination of program steps and logic circuits, modules, and functions. The computer program may be stored in memory. The memory may be of any type suitable to the local technical environment and may be implemented using any suitable data storage technology, such as, but not limited to, read-only memory (ROM), random access memory (RAM), optical storage devices and systems (Digital Multifunction Discs, DVDs, or CDs), etc. Computer-readable media may include non-transitory storage media. The data processor may be of any type suitable to the local technical environment, such as, but not limited to, general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and processors based on multi-core processor architectures.

Claims

1. A model request method, characterized in that, Applied to a first communication entity, the method includes: According to the target model description file, a model request message is sent to the second communication entity, wherein the model request information is used to request the target model; Receive a model feedback message sent by the second communication entity, the model feedback message including information about the target model.

2. The method according to claim 1, characterized in that, The information of the target model includes at least one of the following: a model file of the target model, and a link to the target model.

3. The method according to claim 1, characterized in that, Before sending a model request message to the second communication entity according to the target model description file, the method further includes: Send a file request message to the second or third communication entity; Receive a file feedback message sent by the second communication entity or the third communication entity, the file feedback message containing at least one of the following: at least one model description file, having a link to at least one model description file; Select one or more of the target model description files from the at least one model description file.

4. The method according to claim 3, characterized in that, The file request information includes at least one of the following: requester information, model requirement information.

5. The method according to claim 4, characterized in that, The model requirement information is used to indicate the model-related information requested by the requester.

6. The method according to claim 1, characterized in that, After receiving the model feedback message sent by the second communication entity, the method further includes: Based on the information of the target model, obtain the target model and perform inference based on the target model; Evaluation information is determined based on the reasoning results, and the evaluation information is used to describe the accuracy of the target model description file; An evaluation message is sent to the second communication entity, the evaluation message including the evaluation information.

7. The method according to claim 1, characterized in that, The model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

8. The method according to claim 7, characterized in that, The model licensing information is used to indicate the information of the requester who is qualified to obtain the model.

9. The method according to claim 7, characterized in that, The model description file also includes at least one of the following: version number; model file size; model storage address information; model storage identity information; environment requirements for model operation; model type; model task type; model purpose information; Is it a linear or nonlinear model? Is it suitable for horizontal federated learning? Is it suitable for vertical federated learning? Is it suitable for multi-layer federated learning? Is it suitable for federated learning? Number of model parameters; Data type, format, and dimensionality information of model inputs; Data type, format, and dimensionality information of model outputs; Data preprocessing requirements; Model topology; Relevant information during model training; Data augmentation information; Whether there are residual connections and attention mechanisms; Computational power requirements and inference speed of the model under different computing power levels; Model testing information; Sample code; License information; Information on the region where the model is used.

10. A model feedback method, characterized in that, Applied to a second communication entity, the method includes: Receive a model request message sent by a first communication entity, wherein the model request message is a message used to request a target model based on a target model description file; Based on the model request message, determine the information of the target model; A model feedback message is sent to the first communication entity, the model feedback message including information about the target model.

11. The method according to claim 10, characterized in that, The information of the target model includes at least one of the following: a model file of the target model, and a link to the target model.

12. The method according to claim 10, characterized in that, Before receiving the model request message sent by the first communication entity, the method further includes: Receive the file request message sent by the first communication entity; Based on the file request message, a file feedback message is sent to the first communication entity, the file feedback message containing at least one of the following: at least one model description file, and having a link to at least one model description file.

13. The method according to claim 12, characterized in that, The file request information includes at least one of the following: requester information, model requirement information.

14. The method according to claim 13, characterized in that, The model requirement information is used to indicate the model-related information requested by the requester.

15. The method according to claim 12, characterized in that, The at least one model description file is a file generated by the second communication entity after training the model; and / or, the at least one model description file is a file obtained by the second communication entity from a locally stored model description file.

16. The method according to claim 10, characterized in that, After sending the model feedback message to the first communication entity, the method further includes: Receive an evaluation message sent by the first communication entity, the evaluation message including evaluation information, the evaluation information being used to describe the accuracy of the target model description file; Based on the evaluation information, the target model is retrained or the target model description file is regenerated.

17. The method according to claim 10, characterized in that, The model description file includes at least one of the following: model identifier; model description file identifier; model owner information; model license information.

18. The method according to claim 17, characterized in that, The model licensing information is used to indicate the information of the requester who is qualified to obtain the model.

19. The method according to claim 17, characterized in that, The model description file also includes at least one of the following: version number; model file size; model storage address information; model storage identity information; environment requirements for model operation; model type; model task type; model purpose information; Is it a linear or nonlinear model? Is it suitable for horizontal federated learning? Is it suitable for vertical federated learning? Is it suitable for multi-layer federated learning? Is it suitable for federated learning? Number of model parameters; Data type, format, and dimensionality information of model inputs; Data type, format, and dimensionality information of model outputs; Data preprocessing requirements; Model topology; Relevant information during model training; Data augmentation information; Whether there are residual connections and attention mechanisms; Computational power requirements and inference speed of the model under different computing power levels; Model testing information; Sample code; License information; Information on the region where the model is used.

20. The method according to claim 15, characterized in that, The method further includes: Send a file storage request message to a third communication entity, the file storage request message including at least one of the following: a model description file generated by the second communication entity after training the model, and a model description file stored locally by the second communication entity; Receive the file storage feedback message sent by the third communication entity.

21. The method according to claim 20, characterized in that, After receiving the file storage feedback message sent by the third communication entity, the method further includes: Send a file management request message to the third communication entity, the file management request message being used to instruct the model description file to be managed to perform at least one of the following operations: extraction operation, deletion operation; Receive the file management feedback message sent by the third communication entity.

22. The method according to claim 21, characterized in that, The file management request message includes at least one of the following: model identifier, model description file identifier, model description file extraction identifier, and model description file deletion identifier.

23. A communication entity, characterized in that, include: processor; The processor is configured to implement the method as described in any one of claims 1-22 when executing a computer program.

24. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-22.

Images