Communication method and application device

Abstract

The embodiment of the application provides a communication method and an application device, wherein the method comprises the following steps: a first network element receives a first request from a second network element, the first request is used for requesting model training or model inference using a model; the first network element sends a second request to a third network element, the second request comprises first information, the first information is used for indicating information of the model, and the second request is used for requesting model training or model inference using the model. By adopting the embodiment of the application, the training and inference of the model can be realized under the condition that the network element resources for deploying the AI model are limited.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence (AI) technology, and in particular to a communication method and application device. Background Technology

[0002] The core advantages of next-generation mobile communication networks lie in their ultra-low latency, high bandwidth, and massive device connectivity, making them an ideal platform for supporting the development of AI technology. On the other hand, AI's ability to handle complex problems will greatly enrich the application scenarios of next-generation mobile communication networks.

[0003] Fine-tuning AI models online according to specific scenarios can maximize their performance in those scenarios. However, how to achieve model training and inference when network element resources for deploying AI models are limited is a problem that urgently needs to be solved. Summary of the Invention

[0004] This application discloses a communication method and application device that enables model training and inference when network element resources for deploying AI models are limited.

[0005] Firstly, this application discloses a communication method applied to a first network element. The first network element can be an access network element, a core network element, or an Operation Administration and Maintenance (OAM) network element, or it can be a component, part, or circuit within an access network element, core network element, or OAM network element; or it can be a chip or chip system applicable to an access network element, core network element, or OAM network element; or the first network element can also be a logic module or software capable of implementing all or part of the functions of an access network element, core network element, or OAM network element. The following example uses a first network element, and the method includes:

[0006] Receive a first request from a second network element, the first request being used to request model training or model inference using the model; send a second request to a third network element, the second request including first information, the first information being used to indicate information about the model, the second request being used to request model training or model inference using the model.

[0007] In this application, the second network element can be a terminal device or an access network element, and the third network element can be an internal network element of the 3GPP network or an external network element of the 3GPP network (e.g., an overthe-top (OTT) device). The terminal device or access network element can send a first request to the first network element, requesting model training or model inference. After receiving the first request, the first network element can send a second request to the third network element, requesting the third network element to train the model or use the model for model inference. It can be understood that the second network element sends a second request to the third network element through the first network element, requesting the third network element to train the model or use the model for model inference, enabling the second network element deploying the AI ​​model to achieve model training and inference under resource constraints.

[0008] In some feasible implementations, sending the second request to the third network element includes: sending the second request to the third network element when the first network element does not meet the computing power requirements for model training or model inference.

[0009] In this embodiment, the first network element can perform model training and inference using the model. When the first network element meets the computing power requirements for model training or inference, it can perform model training or inference using the model, enabling the second network element deploying the AI ​​model to perform model training and inference under resource constraints. If the first network element does not meet the computing power requirements for model training or inference, it can send a second request to the third network element, requesting the third network element to perform model training or inference using the model. The third network element performing model training or inference using the model enables the second network element deploying the AI ​​model to perform model training and inference under resource constraints.

[0010] In some feasible implementations, the method further includes: determining that the first network element does not meet the computing power requirements for model training or model inference. That is, if the first network element determines that it does not meet the computing power requirements for model training or model inference, the first network element sends a second request to the third network element.

[0011] In some feasible implementations, determining that the first network element does not meet the computational power requirements for model training or model inference includes: determining that the memory capacity of the first network element is smaller than the size of the model; and / or determining that the processor utilization rate of the first network element is higher than a first threshold; and / or determining that the processing speed of the processor of the first network element is lower than a second threshold. Thus, it can be definitively determined whether the first network element meets the computational power requirements for model training or model inference.

[0012] In some feasible implementations, the first network element is an internal network element of the 3GPP network, and the third network element is an external network element of the 3GPP network.

[0013] In this embodiment, the third network element is an external network element of the 3GPP network, and the first network element is an internal network element of the 3GPP network. In one possible scenario, the second network element cannot directly communicate with the third network element and sends a request to the third network element to train or infer the model. Therefore, the second network element sends a first request to the first network element, which then sends a second request to the third network element. This allows the second network element, which deploys the AI ​​model, to utilize the computing resources of the external network element of the 3GPP network for model training or inference, even with limited resources.

[0014] In some feasible implementations, both the first network element and the third network element are internal network elements of the 3GPP (3rd Generation Partnership Project) network.

[0015] In this embodiment, both the first network element and the third network element are internal network elements of the 3GPP (3rd Generation Partnership Project) network. In this case, the second network element can also send a first request to the first network element, which in turn sends a second request to the third network element, thereby enabling the third network element to train the model or perform model inference. This allows the second network element deploying the AI ​​model to achieve model training and inference under resource constraints.

[0016] In some feasible implementations, sending a second request to a third network element includes: sending a second request to a third network element if the first network element permits the transmission of the first information.

[0017] In this embodiment, if the first network element directly sends the second request to the third network element, there may be a problem of leaking privacy data within the 3GPP network. Therefore, the second request is only sent to the third network element if the first network element permits the transmission of the first information. If the first network element refuses to send the first information, then the first network element does not send the second request to the third network element. For example, the first network element may determine whether privacy data within the 3GPP network is involved based on the input and output data of the model. If it is involved, the first network element refuses to send the first information and does not send the second request to the third network element; if it is not involved, the first network element permits the transmission of the first information and sends the second request to the third network element.

[0018] In some feasible implementations, the method further includes: receiving a first response from the third network element, the first response indicating agreement or refusal to train the model or use the model for model inference. Thus, the first network element can know whether the third network element agrees to train the model or use the model for model inference.

[0019] In some feasible implementations, the method further includes sending a second response to the second network element, the second response indicating agreement or refusal to train the model or use the model for model inference. Thus, the second network element can determine whether the model can be trained or used for model inference by other network elements.

[0020] In some feasible implementations, the method further includes: receiving second information from the second network element, the second information being used to indicate the model. Upon receiving the second information, the model can be determined based on the second information, thereby enabling model training or model inference.

[0021] In some feasible implementations, the method further includes: receiving the training model of the model or the inference result of the model from the third network element; and sending the training model of the model or the inference result of the model to the second network element.

[0022] The training model or inference result obtained by the third network element after training the model or using the model for model inference can be sent to the second network element through the first network element, thereby realizing the training or inference of the model.

[0023] In some feasible implementations, the first request includes one or more of the following information: the identifier of the first request, the identifier of the second network element, the identifier of the agent, the identifier of the model, the number of parameters of the model, the amount of data used for model training or model inference, the performance index requirements for model training or model inference, and the input and output description of the model.

[0024] Secondly, this application provides a communication device, including units, modules, or means for performing the various steps of the first aspect or any of the implementation methods described above. The modules, units, or means can be implemented by software, hardware, or a combination of software and hardware.

[0025] Thirdly, this application provides a communication device, which can be a first network element or a device (e.g., a chip, a chip system, or a circuit) within the first network element. The communication device may include a processor, a memory, an input interface, and an output interface. The input interface is used to receive information from other communication devices besides the communication device, and the output interface is used to output information to other communication devices besides the communication device. The processor invokes a computer program stored in the memory to execute the communication method provided in the first aspect or any embodiment of the first aspect.

[0026] Fourthly, this application provides a communication system comprising at least one first network element, at least one second network element, and at least one third network element. When at least one of the aforementioned first network elements is running in the communication system, it is used to execute the communication method described in the first aspect. When at least one of the aforementioned first network elements is running in the communication system, it is used to send a first request and second information to the first network element, etc. When at least one of the aforementioned third network elements is running in the communication system, it is used to train a model or use the model for model inference, and to send the training model or inference results to the second network element, etc.

[0027] Fifthly, this application provides a computer-readable storage medium storing computer instructions that, when the computer program or computer instructions are executed, cause the methods described in the first aspect and any possible implementation thereof to be performed.

[0028] In a sixth aspect, this application provides a computer program product including executable instructions that, when the computer program product is run on a communication device, cause the methods described in the first aspect and any possible implementation thereof to be executed.

[0029] In a seventh aspect, this application provides a communication device, which includes a processor and may further include a memory, for implementing the methods described in the first aspect and any possible implementation thereof. The communication device may be a chip system, which may be composed of chips or may include chips and other discrete devices.

[0030] Eighthly, this application provides a computer program, including program code, which, when a computer runs the computer program, executes the communication method provided in the first aspect or any embodiment of the first aspect.

[0031] In a ninth aspect, this application provides a chip including a processor, the processor being configured to perform the communication method provided in the first aspect or any embodiment of the first aspect.

[0032] In a tenth aspect, this application provides a chip system including at least one processor, a memory, and an interface circuit. The memory, the interface circuit, and the at least one processor are interconnected via lines. The at least one memory stores instructions. When the instructions are executed by the processor, they implement the communication method provided in the first aspect or any embodiment of the first aspect.

[0033] It should be understood that the implementation and beneficial effects of the above-mentioned aspects can be mutually referenced.

[0034] Furthermore, in the process of performing the method described in the first aspect and any possible implementation, the processes related to sending and / or receiving information in the above methods can be understood as the process of the processor outputting information and / or the process of the processor receiving input information. When outputting information, the processor can output the information to a transceiver (or communication interface or transmitting module) so that the transceiver can transmit it. After the information is output by the processor, it may need to undergo other processing before reaching the transceiver. Similarly, when the processor receives input information, the transceiver (or communication interface or transmitting module) receives the information and inputs it into the processor. Furthermore, after the transceiver receives the information, the information may need to undergo other processing before being input into the processor.

[0035] Based on the above principles, for example, the information sent mentioned in the aforementioned method can be understood as information output by the processor. Similarly, the information received can be understood as information received by the processor from input.

[0036] Optionally, unless otherwise specified, or unless they contradict their actual function or internal logic in the relevant description, the operations of the processor, such as transmitting, sending, and receiving, can be more generally understood as processor output and receiving, input, and other operations.

[0037] Optionally, in performing the methods of the first aspect and any possible implementation described above, the processor may be a processor specifically designed to perform these methods, or it may be a processor that performs these methods by executing computer instructions stored in memory, such as a general-purpose processor. The memory may be a non-transitory memory, such as read-only memory (ROM), which may be integrated with the processor on the same chip or disposed on separate chips. This application does not limit the type of memory or the arrangement of the memory and processor. Attached Figure Description

[0038] The accompanying drawings used in the embodiments of this application are described below.

[0039] Figure 1A and Figure 1B These are schematic diagrams of a network architecture of a communication system provided in an embodiment of this application;

[0040] Figure 2 This is a schematic diagram of an AI function provided in an embodiment of this application;

[0041] Figures 3 to 7 These are interactive schematic diagrams of a communication method provided in an embodiment of this application;

[0042] Figure 8 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0043] Figure 9 This is a schematic diagram of another communication device provided in an embodiment of this application. Detailed Implementation

[0044] The technical solutions of this application embodiment can be applied to various communication systems, such as Long Term Evolution (LTE) communication systems, New Radio (NR) communication systems, LTE-Advanced (LTE-A) communication systems, Device-to-Device (D2D) communication systems, Vehicle-to-Everything (V2X) communication systems, Machine-to-Machine (M2M) communication systems, Internet of Things (IoT) communication systems, Narrow Band Internet of Things (NB-IoT) communication systems, Integrated Sensing and Communication Systems, Frequency Division Duplex (FDD) communication systems, Time Division Duplex (TDD) communication systems, Non-Terrestrial Network (NTN) communication systems, Wireless Projection Communication Systems, Integrated Access and Backhaul (IAB) communication systems, Public Land Mobile Network (PLMN) communication systems, and Non-Public Networks (NPN) communication systems. The network (NPN) communication system, as well as the communication system that evolved after 5G communication system, or non-3rd generation partnership project (3GPP) communication system, are not restricted.

[0045] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described below with reference to the accompanying drawings.

[0046] Please see Figure 1A , Figure 1A This is a schematic diagram of the network architecture of a communication system provided in an embodiment of this application. For example... Figure 1A As shown, the communication system 1000 includes a radio access network (RAN) 100 and a core network (CN) 200. RAN 100 includes at least one RAN node (e.g., Figure 1A 110a and 110b (collectively referred to as 110) and at least one terminal device (such as Figure 1A The 120A-120J models are collectively referred to as 120. Terminal devices connect wirelessly to access network devices, which in turn connect wirelessly or via wired connections to the core network. Core network devices and access network devices can be independent physical devices, or they can integrate the functions of core network devices and access network devices onto the same physical device. Alternatively, a single physical device can integrate some core network device functions and some access network device functions. Terminal devices and access network devices can be interconnected via wired or wireless connections. Figure 1A For illustrative purposes only, this communication system may also include other devices, network elements, and networks.

[0047] In this embodiment, a terminal device is an entity on the user side used to receive or transmit signals, providing voice and / or data to the user. A terminal device may be a terminal, user equipment (UE), access terminal, UE unit, UE station, mobile device, mobile station, mobile station, mobile terminal, mobile client, mobile unit, remote station, remote terminal, remote unit, wireless unit, wireless communication device, user agent, or user device, etc. The access terminal may be a cellular phone, cordless phone, session initiation protocol (SIP) phone, wireless local loop (WLL) station, personal digital assistant (PDA), handheld device with wireless communication capabilities, computing device or other processing device connected to a wireless modem, vehicle-mounted device, terminal in a future communication system, terminal in a future evolved PLMN, or terminal in a future NPN, etc.

[0048] As an example and not a limitation, in this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those that focus on a specific type of application function and require the use of other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0049] Terminal equipment can also be a communication module, satellite phone, or its components with satellite communication capabilities, or a satellite communication terminal, such as a very small aperture terminal (VSAT) (commonly referred to as a VSAT terminal), portable station, fixed station, vehicle-mounted or airborne satellite communication terminal, etc. It should be understood that a satellite communication terminal can serve as a micro base station to further provide data interfaces to accessed user equipment. Hereinafter, it will sometimes be simply referred to as a terminal.

[0050] In the embodiments of this application, the terminal device can be a terminal as a final product, such as the various terminal devices described above; it can also be a component or part with terminal functions; it can be a circuit or chip (such as a modem chip, also known as a baseband chip, or a system-on-chip (SoC) chip or system-in-package (SIP) chip containing a modem core), a chip system, or a processor) that can be applied to the terminal to perform communication functions; or it can be a logic node, logic module, or software that can implement all or part of the terminal functions. The chip system can be composed of chips, or it can include chips and other discrete devices.

[0051] Terminal devices can communicate with each other using some kind of air interface technology (such as NR or LTE). Terminal devices can also communicate with network devices using some kind of air interface technology (such as NR or LTE).

[0052] In this embodiment, the access network device may be referred to as a RAN device, an access network node, or an access network apparatus, or simply an access network. The access network device is used to connect terminal devices to the network. That is, the access network provides access services to terminal devices so that they can access (or access) the network. The access network can support both wired and wireless access.

[0053] Optionally, the access network equipment consists of multiple AN / RAN nodes. AN / RAN nodes can include, but are not limited to: access points (APs), enhanced node Bs (eNBs), home evolved node Bs (HNBs), baseband units (BBUs), next-generation node Bs (gNBs), transmission reception points (TRPs), transmission points (TPs), or other access nodes, such as wireless relay nodes or wireless backhaul nodes. AN / RAN nodes can be one or more antenna panels, or network nodes constituting gNBs or transmission points, such as BBUs or distributed units (DUs), or devices performing RAN functions in communication systems such as D2D, V2X, M2M, and U2U, such as roadside units (RSUs). AN / RAN nodes can be wireless controllers in cloud radio access network (CRAN) scenarios, open RAN (O-RAN or ORAN), or access networks in future communication systems, etc., without any limitations.

[0054] In this application embodiment, in some possible scenarios, AI applications and wireless networks can share infrastructure. The access network equipment may include Base Station Radio Access Network Function (BRF) network elements, Base Station Management Function (BMF) network elements, and Base Station Artificial Intelligence Function (BAF) network elements. Among them, the BRF network element is the real-time processing core of the radio access network, used to ensure air interface performance; the BMF network element is the global management and coordination hub of the base station, mainly used for global management of the base station; and the BAF network element is the intelligent engine of the base station, used to drive data-driven decision optimization.

[0055] In some deployments, a gNB may include a centralized unit (CU) and a dedicated unit (DU). The gNB may also include an active antenna unit (AAU). The CU implements some of the gNB's functions, and the DU implements others. For example, the CU handles non-real-time protocols and services, implementing radio resource control (RRC) and packet data convergence protocol (PDCP) layer functions. The DU handles physical layer protocols and real-time services, implementing radio link control (RLC), media access control (MAC), and physical (PHY) layer functions. The AAU implements some physical layer processing functions, radio frequency processing, and active antenna-related functions. RRC layer information is generated by the CU and is ultimately encapsulated by the DU's PHY layer to become PHY layer information, or it may be derived from PHY layer information. Therefore, in this architecture, higher-layer signaling, such as RRC layer signaling, can be considered as being sent by the DU, or by the DU+AAU. It is understood that network devices can be one or more of the following: CU nodes, DU nodes, and AAU nodes. Furthermore, a CU can be classified as a network device in the RAN or as a network device in the core network; this application does not limit this classification.

[0056] In some deployments, a gNB may include a radio unit (RU). The RU may be included in a radio frequency device or radio frequency unit, such as a remote radio unit (RRU), an AAU, or a remote radio head (RRH).

[0057] In different systems, CU (or CU-control plane (CP), CU-user plane (UP)), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. For ease of description, this application uses CU, CU-CP, CU-UP, DU, and RU as examples. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0058] In 5G communication systems, core network equipment can refer to... Figure 1B These correspond to network elements for policy control function (PCF), application function (AF), access and mobility management function (AMF), session management function (SMF), location management function (LMF), user plane function (UPF), and network data analytics function (NWDAF).

[0059] Among them, the UPF network element is responsible for managing the transmission of user plane data and quality of service (QoS) control, traffic statistics and other functions. It can perform user data packet forwarding according to the routing rules of the session management network element, such as sending uplink data to the data network or other user plane network elements, and forwarding downlink data to other user plane network elements or (R)AN network elements.

[0060] The AMF (Access Default Mode) network element is responsible for user access management, security authentication, and mobility management. The LMF (Local Mode Default Mode) network element manages and controls location service requests from target terminals and processes location-related information. The SMF (Signal Management Default Mode) network element is responsible for session management, allocating and releasing resources for terminal device sessions. The PCF (Policy and Charging Rules Function) network element is responsible for user policy management. Similar to the Policy and Charging Rules Function (PCRF) network element in LTE, it is mainly responsible for policy authorization, quality of service (QoS), and generating charging rules, and distributing these rules to the UPF (Universal Programming Default Mode) network element via the SMF network element to complete the installation of the corresponding policies and rules. The AF (Application Default Mode) network element can be a third-party application control platform or the operator's own equipment. The AF network element is responsible for application management and can provide services to multiple application servers.

[0061] The NWDAF (Network Data Analyzer) element is used for analyzing network slicing-related data and can be extended to analyze various types of network data, including network operation data collected from network functions, statistical data related to terminal devices and networks obtained from Operation Administration and Maintenance (OAM) equipment, and application data obtained from third-party applications. The analysis results generated by NWDAF can also be output to network functions, OAM equipment, or third-party applications. NWDAF is also typically responsible for training AI models. AI models trained by NWDAF can be applied to network-specific domains such as mobility management, session management, and network automation, using AI methods to replace the numerical formula-based methods in traditional network functions.

[0062] OAM network elements, also known as OAM devices, OAM entities, or OAM functions, refer to the network management work typically divided into three categories based on the actual needs of operator network operations: operation, administration, and maintenance, abbreviated as OAM. Operation mainly involves the analysis, prediction, planning, and configuration of daily network and services; maintenance mainly involves routine operational activities such as testing and fault management of the network and its services. OAM network elements can detect network operating status, optimize network connectivity and performance, improve network stability, and reduce network maintenance costs.

[0063] In this application embodiment, a network element may also be referred to as a functional network element, functional entity, node, device, etc. A network element can be a network component implemented on dedicated hardware, a software instance running on dedicated hardware, or an instance of virtualized functionality on a suitable platform; for example, the aforementioned virtualization platform can be a cloud platform. In different communication systems, the above... Figure 1BThe network elements shown, such as AMF, SMF, PCF, and NWDAF, may have other names, and this application does not impose any restrictions.

[0064] Optional, please continue to refer to Figure 1B The communication system may also include Figure 1A The data network device is not shown in the image. The data network device may be simply referred to as the data network below. The data network is used to provide business services to users.

[0065] In the embodiments of this application, the network device may be a network equipment as a final product, such as at least one of the above-mentioned access network equipment, core network equipment and data network equipment or network elements therein, or may include independent network elements, or may be a component or part with network equipment functions, or may be a communication chip (such as a processor, baseband chip or chip system, etc.) that can be applied in the network equipment.

[0066] like Figure 1A and Figure 1B The number and types of network devices and terminal devices included in the network architecture shown are merely examples, and the embodiments of this application are not limited thereto. For example, it may also include more or fewer terminal devices communicating with the network devices. Similarly, it may include more or fewer network devices communicating with the terminal devices. For the sake of brevity, they are not described one by one in the accompanying drawings.

[0067] In this embodiment, the communication system may further include over-the-top (OTT) devices, such as UE-side servers and OTT servers. OTT devices can be internet devices that provide various application services to users via the internet. OTT devices may be servers belonging to communication device manufacturers, or communication devices in the external network of the communication system, such as suppliers of AI solutions (e.g., autonomous driving solutions, XR solutions). The external network of the communication system may be an external network of the 3GPP communication network. The services provided by the OTT devices may include internet television services, app stores, etc., and are not limited thereto.

[0068] The aforementioned terminal devices and network devices, as well as OAM devices and OTT devices, can all be referred to as communication devices. They can be general-purpose devices or special-purpose devices, and the embodiments of this application do not specifically limit them.

[0069] The network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0070] To facilitate understanding of this application, the following will be combined with... Figure 2 This section introduces the functional framework for AI or machine learning (ML). This framework refers to the AI / ML functional framework for the NR air interface in versions R18 or R19 of protocol TS38.843, such as... Figure 2 As shown, the functional framework includes data collection, model training, management, inference, and model storage functions.

[0071] The collected data can include monitoring data, training data, and inference data. Training data serves as the input for training AI / ML models. Inference data serves as the input for AI / ML inference functions, and monitoring data serves as the input for managing AI / ML models or functions.

[0072] The model training function is used to train, validate, and test AI / ML models. It can also be used to generate model performance metrics, which can be used as part of the model testing process. The model training function is also used to process training data, such as data preprocessing, cleaning, formatting, and transformation. After training or updating the model, it can be transferred or delivered to a device with model storage capabilities for model storage.

[0073] The management function is responsible for managing the operations of AI / ML models or functions, such as selection, deactivation, switching, and rollback, and also for performance monitoring. The management function is also responsible for making decisions to ensure correct inference operations are performed based on data received from the data acquisition and inference functions. Devices with management functions can send management instructions to devices with inference functions, which serve as the information required for managing the inference function's input. Devices with management functions can also request model transfers or delivery requests from devices with model storage functions to store the model. Performance feedback or retraining requests serve as the input information required for the model training function.

[0074] Devices with inference capabilities use inference data as input to provide outputs, such as inference outputs, to devices that apply AI / ML models or AI / ML functions. Inference outputs can be used to monitor the performance of AI / ML models or AI / ML functions. After storing models on devices with model storage capabilities, AI / ML models can be transferred or delivered to devices with inference capabilities.

[0075] This application proposes a communication method that enables network elements deploying AI models to perform model training and inference under resource constraints.

[0076] The following descriptions will illustrate these methods through various embodiments. It should be understood that these methods can be used in combination. The technical solutions provided in this application are not limited to the processes described below. Furthermore, the scenario descriptions in this application are merely illustrative and do not limit the scope of the solutions described. The solutions in this application are applicable not only to the described scenarios but also to scenarios with similar problems.

[0077] The first network element in this application embodiment can be a network element within a 3GPP network, such as an access network element, core network element, or OAM network element; or it can be a component, part, or circuit within an access network element, core network element, or OAM network element; or it can be a chip or chip system applicable to an access network element, core network element, or OAM network element; or the first network element can also be a logic module or software capable of implementing all or part of the functions of an access network element, core network element, or OAM network element. The second network element in this application embodiment can be a terminal device or an access network element; or it can be a component, part, or circuit within a terminal device or access network element; or it can be a chip or chip system applicable to a terminal device or access network element. The third network element in this application embodiment can be a network element within a 3GPP network or a network element outside a 3GPP network (e.g., an OTT server).

[0078] Please see Figure 3 , Figure 3 This is an interactive schematic diagram of a communication method provided in an embodiment of this application. The communication device involved in this method may include a first network element, a second network element, and a third network element, etc. The method includes, but is not limited to, the following steps S301 and S302, wherein:

[0079] Step S301: The second network element sends a first request to the first network element. The first request is used to request model training or model inference.

[0080] Correspondingly, the first network element receives a first request from the second network element. That is, the second network element requests the first network element to train the model or use the model for model inference. The second network element can be a terminal device or an access network element. In one example, when the second network element is a terminal device, the first network element can be an access network element or a core network element. In another example, when the second network element is a terminal device, the first network element can be a BRF element in the access network device. In yet another example, when the second network element is an access network element, the first network element can be a core network element. In yet another example, when the second network element is a BRF element in the access network device, the first network element can be a BMF element in the access network device.

[0081] Optionally, the first request may include one or more of the following information: the identifier of the first request, the identifier of the second network element, the identifier of the agent, the identifier of the model, the number of parameters of the model, the amount of data used for model training or model inference, the performance index requirements for model training or model inference, and the input and output description of the model.

[0082] The identifier for the first request indicates the first request, and the content and / or type requested by the first request can be determined based on the identifier. The identifier for the second network element indicates the second network element; for example, if the second network element is a terminal device, the identifier for the second network element can refer to the UEID of the terminal device. The identifier for the agent indicates the agent or vendor of the second network element. The identifier for the model indicates the model, and the model and other information about the model can be determined based on the identifier. The number of parameters in the model can refer to the number of variables that need to be determined through model training. The amount of data used for model training or model inference indicates the scale of data processed by the model. The amount of data in model training can be the number of samples used to train the model and its feature dimensions. The amount of data in model inference can refer to the scale of the input data for each inference. Performance requirements for model training or model inference can include, for example, accuracy, speed, and resource consumption. The input-output description of the model can indicate the content of the model's input and output data; for example, the input data is an image, and the output data is the image category. For example, after receiving a first request, the first network element can determine the content and / or type requested by the first request based on its identifier; it can also determine the network element that sent the first request as the second network element based on its identifier, thus enabling the first network element to determine whether to interact with the second network element subsequently, such as sending a second response, inference results, or training a model; it can identify the agent or supplier of the second network element based on its agent's identifier, thus enabling the first network element to send a second request to the agent or supplier of the second network element; based on the model's identifier, the number of model parameters, the amount of data used for model training or model inference, the performance requirements for model training or model inference, and the model's input and output description, the first network element can obtain relevant information about the model, thus enabling it to determine whether it meets the computing power requirements for model training or model inference, and consequently, whether it needs to send second information to the third network element.

[0083] Optionally, the first request may further include a request type, indicating whether the request specifically requests model training or model inference. This can be indicated by a single bit. For example, a value of 1 for the request type indicates a request for model training, while a value of 0 indicates a request for model inference. Alternatively, a value of 0 indicates a request for model training, while a value of 1 indicates a request for model inference. This application does not limit the scope of this embodiment.

[0084] Step S302: The first network element sends a second request to the third network element. The second request includes first information, which is used to indicate information about the model. The second request is used to request model training or model inference.

[0085] Correspondingly, the third network element receives a second request from the first network element. That is, the first network element requests the third network element to perform model training or model inference. The third network element can be an internal network element or an external network element of the 3GPP network. In one example, the first network element is an access network element, and the third network element can be an external network element of the 3GPP network, such as a UE-side server. In another example, the first network element is a BRF network element in the access network equipment, and the third network element can be a BAF network element in the access network equipment. In yet another example, the first network element is an internal network element of the 3GPP network, such as an access network equipment or core network equipment, and the third network element can be an external network element of the 3GPP network, such as an OTT server. In yet another example, the first network element is a BMF network element in the access network equipment, and the third network element can be a BAF network element in the access network equipment.

[0086] Understandably, the first network element can determine the second request based on the first request. For example, if the first request is for requesting model training, then the second request is determined to be for requesting model training; if the first request is for requesting model inference, then the second request is determined to be for using the model for model inference.

[0087] The first piece of information may include the number of model parameters, the amount of data used for model training or model inference, and the performance requirements for model training or model inference.

[0088] In one possible implementation, the second request may include other information besides the first information, such as the identifier of the second request, the identifier of the agent, etc. In another possible implementation, when the second request includes the identifier of the second request, the identifier of the second request may be the same as or different from the identifier of the first request; this application embodiment does not limit this.

[0089] In one possible implementation, after receiving the first request, the first network element does not directly send the second request to the third network element. Instead, it first performs a judgment to determine whether to send the second request to the third network element. The following example illustrates the scenario of sending the second request to the third network element.

[0090] Example 1: If the first network element does not meet the computing power requirements for model training or model inference, the first network element sends a second request to the third network element.

[0091] In other words, if the first network element does not meet the computing power requirements for model training or inference, it cannot perform model training or inference and needs to request other communication devices to perform the training or inference. Therefore, a second request can be sent to the third network element to request it to perform model training or inference. If the first network element meets the computing power requirements for model training or inference, it can perform model training or inference without requesting other communication devices, and thus does not need to send a second request to the third network element. If the first network element meets the computing power requirements for model training or inference, after training or inferring the model, it can send the training model or inference result to the second network element. After training or inferring the model, the third network element can send the training model or inference result directly to the second network element, or send it through the first network element.

[0092] In some feasible implementations, the method may further include: a first network element determining whether it satisfies the capability for model training or model inference. If the first network element determines that it does not satisfy the capability for model training or model inference, the first network element may send a second request to a third network element.

[0093] For example, when the first network element has not yet determined whether it meets the computational requirements for model training or inference, or when the first network element has determined that it meets the requirements for model training or inference, the first network element does not send a second request to the third network element. The first network element can determine that it does not meet the computational requirements for model training or inference by: determining that the memory capacity of the first network element is less than the size of the model; and / or determining that the processor utilization rate of the first network element is higher than a first threshold; and / or determining that the processing speed of the processor of the first network element is lower than a second threshold.

[0094] The memory may include main memory and / or video memory. The model size may refer to the sum of the model and the training data (or inference data). The processor may include a CPU and / or a GPU. The first threshold may be defined in the 3GPP standard or may be predefined by the manufacturer of the first network element; the second threshold may be defined in the 3GPP standard or may be predefined by the manufacturer of the first network element, and this application embodiment does not limit this. For example, the processor's computing speed can be measured by the trillion operations per second (TOPS) available to the processor.

[0095] For example, if it is determined that the memory and / or GPU memory of the first network element can load the model and training data (or inference data), then the memory capacity of the first network element is greater than the size of the model, indicating that the storage space of the first network element is sufficient. If the memory and / or GPU memory cannot load the model and training data (or inference data), then the memory capacity of the first network element is less than the size of the model, indicating that the storage space of the first network element is insufficient. As another example, if it is determined that the CPU and / or GPU utilization rate of the first network element is lower than a first threshold, then the processor utilization rate of the first network element is lower than the first threshold, indicating that the computing power of the first network element is sufficient. If it is determined that the CPU and / or GPU utilization rate of the first network element is higher than the first threshold, then the processor utilization rate of the first network element is higher than the first threshold, indicating that the computing power of the first network element is insufficient, i.e., it does not meet the computing power requirements for model training or model inference. As yet another example, if it is determined that the available TOPS of the processor of the first network element is lower than a second threshold, then the computing speed of the first network element is lower than the second threshold. If it is determined that the available TOPS of the processor of the first network element is higher than the second threshold, then the computing speed of the first network element is higher than the second threshold. Loading a model refers to reading a model from a storage device (such as a hard drive) into memory for inference or further training. For example, in this embodiment, loading a model may mean reading a model from a storage device (e.g., a terminal device or a first network element) into the memory of the first network element so that the first network element can train the model or use the model for inference.

[0096] Example 2: If the first network element allows the transmission of the first information, the first network element sends a second request to the third network element.

[0097] Permission to send the first information means permission to send the first information to other communication devices, such as a third network element.

[0098] In other words, the first network element will only send a second request to the third network element if the first network element allows the first information to be sent; if the first network element refuses to send the first information, the first network element will not send a second request to the third network element.

[0099] For example, the first message is refused to be sent when the model's input and output involve intra-network privacy data; the first message is allowed to be sent when the model's input and output do not involve intra-network privacy data. This is because the third network element may be a network element outside the 3GPP network, while the model's input and output may involve intra-network privacy data, posing a risk of intra-network privacy data leakage. For example, the location of the second network element can be sent outside the network after authorization by the first network element, while the Reference Signal Receiving Power (RSRP) is usually reserved for use by the first network element. If the model's input and output involve RSRP, the first network element refuses to send the first message and does not send a second request to the third network element; if the model's input and output do not involve RSRP, the first network element allows the first message to be sent and sends a second request to the third network element. In this application embodiment, information belonging to privacy data and / or information not belonging to privacy data can be predefined by the 3GPP standard or pre-configured by the first network element manufacturer; this application embodiment does not limit this.

[0100] Example 3: If the first network element does not meet the computing power requirements for model training or model inference but is allowed to send the first information, the first network element sends a second request to the third network element.

[0101] As can be seen, Example 3 is a combination of Examples 1 and 2 above. When the first network element does not meet the computational power requirements for model training or inference but is permitted to send the first information, the first network element sends a second request to the third network element; when the first network element meets the computational power requirements for model training or inference or refuses to send the first information, the first network element does not send a second request to the third network element. For specific implementation details, please refer to the implementation methods of Examples 1 and 2 above, which will not be elaborated upon here.

[0102] Optionally, the method may further include: a third network element sending a first response to a first network element, the first response being used to indicate agreement or refusal to train the model or use the model for model inference.

[0103] Correspondingly, the first network element receives a first response from the third network element. That is, the first response can be a message sent by the third network element to the first network element, instructing the third network element to agree or refuse to train the model or use the model for model inference.

[0104] In one example, when the first network element is an access network element and / or a core network element or an OAM network element, the third network element can be a UE-side server. In another example, when the first network element is a BRF network element in the access network device, the third network element can be a BAF network element in the access network device. In yet another example, when the first network element is a core network element or an OAM network element, the third network element can also be an OTT server. In yet another example, when the first network element is a BMF network element in the access network device, the third network element can be a BAF network element in the access network device.

[0105] Optionally, the method may further include: the first network element sending a second response to the second network element, the second response being used to indicate agreement or refusal to train the model or use the model for model inference.

[0106] Correspondingly, the second network element receives a second response from the first network element. That is, the second response can be a message sent from the first network element to the second network element, indicating agreement or refusal to train the model or use the model for model inference. Optionally, the second response can also be specifically used to instruct the first or third network element to agree or refuse to train the model or use the model for model inference. It is understood that when the second response is used to indicate agreement to train the model or use the model for model inference, the second response can also be used to instruct the second network element to send the model.

[0107] In one example, when the second network element is a terminal device, the first network element can be an access network element, a core network element, or an OAM network element. In another example, when the second network element is a terminal device, the first network element can be a BRF network element in the access network device. In yet another example, when the second network element is an access network element, the first network element can be a core network element or an OAM network element. In yet another example, when the second network element is a BRF network element in the access network device, the first network element can be a BMF network element in the access network device.

[0108] Optionally, the second network element sends second information to the third network element, the second information being used to indicate the model.

[0109] Correspondingly, the third network element receives the second information from the second network element. The second information may indicate the model in ways including, but not limited to, indicating the model through a model identifier, indicating the model through the model's structure and parameters, or indicating the model through a model file; this application embodiment does not limit this. For example, the model file format includes, but is not limited to, .pt files, .bin files, .onnx files, and .pth files.

[0110] The second information may also include a request identifier. It is understood that the request identifier in the second information can be the identifier of the first request. The second network element sending the second information to the third network element can be sending the request identifier along with the model's structure and parameters, allowing the third network element to determine the model based on its structure and parameters, and to determine whether to train the model or use it for inference, as well as the corresponding performance requirements. Alternatively, the second network element sending the second information to the third network element can be sending the request identifier and a model identifier, allowing the third network element to determine the model based on the model identifier, and to determine whether to train the model or use it for inference, as well as the corresponding performance requirements. For example, if the second network element has previously sent the model's structure and parameters to the third network element, or if the model is deployed by the third network element, then the model can be considered to be present on the third network element. Upon receiving the model identifier, the third network element can determine the model's structure and parameters based on the model identifier. Furthermore, the second network element sending the model to the third network element can only send an indication of the adaptation layer or the number of layers that need updating, thereby reducing resource consumption. Alternatively, the second information sent to the third network element can be a request identifier and a model file, so that the third network element can obtain the model based on the model file and determine whether to train the model or use the model for model inference and the corresponding performance requirements based on the request identifier.

[0111] The method by which the second network element sends the second information to the third network element can be either that the second network element directly sends the model to the third network element, or that the second network element sends the second information to the third network element through the first network element. It can be understood that sending the second information from the second network element to the third network element through the first network element can mean that the second network element sends the second information to the first network element, and after receiving the second information from the second network element, the first network element then sends the second information to the third network element.

[0112] For example, when the second network element is a terminal device and the third network element is a UE-side server, and the second and third network elements belong to the same vendor, the second network element can directly send the second information to the third network element, or the second network element can send the second information to the third network element through the first network element. Situations where the second network element sends the second information to the third network element through the first network element also include, but are not limited to: when the second network element is a terminal device and the third network element is a BAF network element in the access network equipment; when the second network element is an access network equipment and the third network element is an OTT server; when the second network element is a BRF network element in the access network equipment and the third network element is a BAF network element in the access network equipment.

[0113] Optionally, after the third network element trains the model or uses the model for model inference, it can send third information to the second network element.

[0114] Correspondingly, the second network element receives third information from the third network element.

[0115] The third piece of information may include one or more of a request identifier, a completion status indication, and the training model or inference result. For example, the completion status indication may indicate that model inference or model training is complete, or that model inference or model training is not complete. When the completion status indicates that model inference or model training is complete, the third piece of information may also include the training model or inference result. As another example, if the third piece of information includes the training model or inference result, it may not need to include a completion status indication, because sending the training model or inference result can implicitly assume that the completion status indication indicates that model inference or model training is complete. It is understood that training a model yields a trained model, and using the model for inference yields an inference result.

[0116] The method by which a third network element sends third information to a second network element can be either that the third network element sends the third information directly to the second network element, or that the third network element sends the third information to the second network element through a first network element. It can be understood that sending third information from the third network element to the second network element through a first network element means that the third network element sends the third information to the first network element, and after receiving the third information from the third network element, the first network element then sends the third information to the second network element.

[0117] For example, when the second network element is a terminal device and the third network element is a UE-side server, and the second and third network elements belong to the same vendor, the third network element can directly send third information to the second network element, or it can send the third information to the second network element through the first network element. Situations where the third network element sends second information to the second network element through the first network element also include, but are not limited to: when the second network element is a terminal device and the third network element is a BAF network element in the access network device; when the second network element is an access network device and the third network element is an OTT server; when the second network element is a BRF network element in the access network device and the third network element is a BAF network element in the access network device.

[0118] In this application, a second request is sent from a first network element to a third network element, requesting model training or model inference. The third network element then trains or infers the model, enabling the second network element deploying the AI ​​model to perform model training and inference under resource constraints.

[0119] Optionally, if the second network element has previously sent a first request to the first network element requesting model training or model inference, when the second network element sends the first request again, the first request may further include a suggested computing power indicator to indicate the recommended network element for model training or model inference. For example, the second network element is a terminal device, the first network element is a BRF network element in the access network device, and the third network element is a BAF network element in the access network device. If the first network element (or the third network element) performed model training or model inference when the second network element previously sent the first request, then when the second network element sends the first request again, the first request may include a suggested computing power indicator to indicate the recommended first network element (or the third network element) for model training or model inference. This reduces the latency of the first network element in determining computing power (whether it meets the computing power requirements for model training or model inference) and allocating resources.

[0120] The communication methods provided in this application are described below in conjunction with different first network elements, second network elements, and third network elements. These communication methods should be included in... Figure 3 In the methods shown, please refer to the following respectively. Figures 4 to 7 The described communication method, in Figure 4 In this context, the first network element can be an access network element, a core network element, or an OAM network element; the second network element can be a terminal device; and the third network element can be a UE-side server. For example... Figure 4 As shown, the communication method may include, but is not limited to, the following steps:

[0121] Step S401: The terminal device sends a first request to the access network element.

[0122] Correspondingly, the access network element receives the first request sent by the terminal device.

[0123] The first request is used to request model training or model inference using the model. In other words, the terminal device requests the access network element to train the model or use the model for model inference.

[0124] Optionally, the first request may include one or more of the following information: the identifier of the first request, the identifier of the terminal device, the identifier of the agent, the identifier of the model, the number of parameters of the model, the amount of data used for model training or model inference, the performance index requirements for model training or model inference, and the input and output description of the model.

[0125] Optionally, the first request may further include a request type, indicating whether the request specifically requests model training or model inference. This can be indicated by a single bit. For example, a value of 1 for the request type indicates a request for model training, while a value of 0 indicates a request for model inference. Alternatively, a value of 0 indicates a request for model training, while a value of 1 indicates a request for model inference. This application does not limit the scope of this embodiment.

[0126] Optionally, in step S402, the access network element determines whether the access network element meets the computing power requirements for model training or model inference.

[0127] After receiving the first request, the access network element can determine whether it meets the computing power requirements for model training or model inference based on the model information in the first request.

[0128] Access network elements can determine whether they meet the computing power requirements for model training or inference using the following methods: determining whether the memory capacity of the access network element is less than the size of the model; and / or, determining whether the processor utilization rate of the access network element is higher than a first threshold. The memory may include system memory and / or video memory; and / or, determining whether the processor speed of the access network element is lower than a second threshold. The processor may include a CPU and / or a GPU. The first threshold can be defined in the 3GPP standard, or it may be predefined by the access network element manufacturer; the second threshold can be defined in the 3GPP standard, or it may be predefined by the access network element manufacturer, and this application embodiment does not limit this. For example, if it is determined that the memory and / or GPU memory of the access network element can load the model and training data (or inference data), then it is determined that the memory capacity of the access network element is greater than the size of the model, indicating that the storage space of the access network element is sufficient. If the memory and / or GPU memory cannot load the model and training data (or inference data), then it is determined that the memory capacity of the access network element is less than the size of the model, indicating that the storage space of the access network element is insufficient, that is, the access network element does not meet the computing power requirements for model training or model inference. As another example, if it is determined that the CPU and / or GPU utilization rate of the access network element is lower than a first threshold, then it is determined that the processor utilization rate of the access network element is lower than the first threshold, indicating that the computing power of the access network element is sufficient. If it is determined that the CPU and / or GPU utilization rate of the access network element is higher than the first threshold, then it is determined that the processor utilization rate of the access network element is higher than the first threshold, indicating that the computing power of the access network element is insufficient, that is, the access network element does not meet the computing power requirements for model training or model inference. For example, if it is determined that the TOPS available to the processor of the first network element is lower than the second threshold, it means that the computing speed of the first network element is lower than the second threshold; if it is determined that the TOPS available to the processor of the first network element is higher than the second threshold, it means that the computing speed of the first network element is higher than the second threshold.

[0129] Optionally, if the access network element determines that the access network element meets the computing power requirements for model training or model inference, then steps S403 to S406 are executed, and the access network element performs model training or uses the model for model inference, without executing other steps.

[0130] Optionally, in step S403, the access network element sends a third response to the terminal device.

[0131] Correspondingly, the terminal device receives a third response from the access network element.

[0132] The third response is used to indicate consent to model training or model inference. Further, the third response can also be used to instruct the terminal device to send the model. Optionally, the third response can specifically be used to instruct the access network element to consent to model training or model inference.

[0133] Optionally, in step S404, the terminal device sends the fourth information to the access network element.

[0134] Correspondingly, the access network element receives the fourth information from the terminal device.

[0135] The fourth piece of information is used to indicate the model.

[0136] The terminal device can send the fourth piece of information to the access network element by sending the model's structure and parameters, allowing the access network element to determine the model based on these parameters. Alternatively, the terminal device can send a model identifier to the access network element, enabling the access network element to determine the model based on the identifier. For example, if the terminal device has previously sent the model's structure and parameters to the access network element, or if the model is deployed by the access network element, then the access network element can be considered to contain the model. Upon receiving the model identifier, the access network element can then determine the model's structure and parameters based on it. Furthermore, the terminal device can send only the adaptation layer or the layer number requiring update to the access network element, thereby reducing resource consumption.

[0137] Optionally, in step S405, the access network element trains the model or uses the model for model inference.

[0138] Understandably, when the first request is used to instruct the model to be trained, the access network element will train the model based on the first request. When the first request is used to instruct the model to be used for inference, the access network element will use the model to perform inference based on the first request.

[0139] Optionally, in step S406, the access network element sends the training model or inference result to the terminal device.

[0140] Correspondingly, the terminal device receives the training model or inference results from the access network element.

[0141] It is understandable that when an access network element trains a model, it sends the training model to the terminal device; when an access network element uses the model for inference, it sends the inference result to the terminal device.

[0142] Optionally, if the access network element determines that the access network element does not meet the computing power requirements for model training or model inference, then step S407 is executed.

[0143] Optionally, in step S407, the access network element sends a third request to the core network element or the OAM network element.

[0144] Correspondingly, the core network element or OAM network element receives a third request from the access network element.

[0145] The third request is used to request model training or model inference. The third request may include some or all of the information from the first request. The third request may be the same as or different from the first request. For example, the third request may include an identifier for the third request, an identifier for the terminal device, an identifier for the agent, an identifier for the model, the number of parameters of the model, the amount of data required for model training or model inference, performance metrics required for model training or model inference, a description of the model's input and output, and the request type. In one possible implementation, when the third request includes an identifier for the third request, the identifier for the third request may be the same as or different from the identifier for the first request; this embodiment of the application does not limit this.

[0146] Optionally, in step S408, the core network element or OAM network element determines whether to allow the transmission of the first message.

[0147] The first piece of information is used to indicate the model's information. This first piece of information may include the number of model parameters, the amount of data used for model training or inference, and performance metrics required for model training or inference.

[0148] For example, a core network element or an OAM network element can determine whether to allow the transmission of the first information by determining whether the input and output of the model involve privacy data within the network. If the core network element or OAM network element determines that the input and output of the model involve privacy data within the network, then the core network element or OAM network element refuses to send the first information; if the core network element or OAM network element determines that the input and output of the model do not involve privacy data within the network, then the core network element or OAM network element allows the transmission of the first information. In this embodiment, the information belonging to privacy data and / or information not belonging to privacy data can be predefined by the 3GPP standard or pre-configured by the vendor of the core network element or OAM network element, and this embodiment does not limit this.

[0149] Optionally, if the core network element or OAM network element refuses to send the first information, steps S409 to S410 are executed, and no other steps are executed.

[0150] Optionally, in step S409, the core network element or OAM network element sends a fourth response to the access network element.

[0151] Correspondingly, the access network element receives the fourth response from the core network element or the OAM network element.

[0152] The fourth response is used to indicate a refusal to train the model or use it for model inference.

[0153] Optionally, in step S410, the access network element sends a second response to the terminal device.

[0154] Correspondingly, the terminal device receives a second response from the access network element.

[0155] The second response is used to indicate a refusal to train the model or use the model for inference.

[0156] Optionally, if the core network element or OAM network element permits the transmission of the first information, step S411 is executed.

[0157] Optionally, in step S411, the core network element or OAM network element sends a second request to the UE-side server.

[0158] Correspondingly, the UE-side server receives a second request from a core network element or an OAM network element.

[0159] The second request is used to request model training or model inference. In one possible implementation, the second request may include other information besides the first information, such as an identifier for the second request or an identifier for the agent. In one possible implementation, when the second request includes an identifier for the second request, the identifier for the second request may be the same as or different from the identifier for the first request; this embodiment does not limit this.

[0160] Optionally, in step S412, the UE-side server sends a first response to the core network element or the OAM network element.

[0161] Correspondingly, the core network element or OAM network element receives the first response from the UE-side server.

[0162] The first response is used to indicate consent to train the model or use the model for inference.

[0163] Optionally, in step S413, the core network element or OAM network element sends a fourth response to the access network element.

[0164] Correspondingly, the access network element receives the fourth response from the core network element or the OAM network element.

[0165] The fourth response is used to indicate consent to model training or model inference.

[0166] Optionally, in step S414, the access network element sends a second response to the terminal device.

[0167] Correspondingly, the terminal device receives a second response from the access network element.

[0168] The second response indicates consent to model training or model inference. Optionally, the second response may also instruct the terminal device to send the model to the UE-side server.

[0169] Optionally, in step S415, the terminal device sends second information to the UE-side server, the second information being used to indicate the model.

[0170] Correspondingly, the UE-side server receives second information from the terminal device. The second information may indicate the model in ways including, but not limited to, indicating the model through a model identifier, indicating the model through the model's structure and parameters, or indicating the model through a model file; this application embodiment does not limit this. The model file format includes, but is not limited to, .pt files, .bin files, .onnx files, and .pth files.

[0171] The second information may also include a request identifier. It is understood that the request identifier in the second information can be the identifier of the first request. The terminal device sending the second information to the UE-side server can be sending the request identifier along with the model's structure and parameters. This allows the UE-side server to determine the model based on its structure and parameters, and to determine whether to train the model or use it for model inference, as well as the corresponding performance requirements, based on the request identifier. Alternatively, the terminal device sending the second information to the UE-side server can be sending both the request identifier and a model identifier. This allows the UE-side server to determine the model based on the model identifier, and to determine whether to train the model or use it for model inference, as well as the corresponding performance requirements. For example, if the terminal device has previously sent the model's structure and parameters to the UE-side server, or if the model was deployed by the UE-side server, then the UE-side server can be considered to contain the model. Upon receiving the model identifier, the UE-side server can determine the model's structure and parameters based on the model identifier. Furthermore, the terminal device sending the model to the UE-side server can only send the adaptation layer or the number of layers that need updating, thereby reducing resource consumption. Alternatively, the second information sent to the third network element can be a request identifier and a model file, so that the third network element can obtain the model based on the model file and determine whether to train the model or use the model for model inference and the corresponding performance requirements based on the request identifier.

[0172] The terminal device can send the second information to the UE-side server either directly to the UE-side server or through access network elements, core network elements, or OAM network elements. Specifically, sending the second information through access network elements, core network elements, or OAM network elements can mean that the terminal device sends the second information to an access network element, which then forwards it to a core network element or OAM element, which in turn forwards it to the UE-side server.

[0173] Optionally, in step S416, the UE-side server performs model training or uses the model for model inference.

[0174] It is understandable that when the second request is used to instruct the model to be trained, the UE-side server will train the model based on the second request. When the second request is used to instruct the model to be used for model inference, the UE-side server will use the model for model inference based on the second request.

[0175] Optionally, in step S417, the UE-side server sends the training model or inference results to the terminal device.

[0176] Correspondingly, the terminal device receives the training model or inference results from the UE-side server.

[0177] It is understandable that when the UE-side server trains the model, it sends the training model to the terminal device; when the UE-side server uses the model for inference, it sends the inference result to the terminal device.

[0178] The training model or inference results may be included in the third information. Furthermore, the third information may also include one or more of a request identifier and a completion status indication. The completion status indication is used to indicate that model inference or model training is complete, or to indicate that model inference or model training is not complete. Optionally, the third information may include the training model or inference results, but may not need to include a completion status indication, because sending the training model or inference results can implicitly assume that the completion status indication indicates that model inference or model training is complete.

[0179] The UE-side server can send training models or inference results to the terminal device either directly or via core network elements, OAM elements, and access network elements. Specifically, sending training models or inference results via core network elements, OAM elements, and access network elements can mean the UE-side server sends the training model or inference results to the core network element or OAM element, which then sends them to the access network element, which in turn sends them to the terminal device.

[0180] exist Figure 4 In this context, the first network element may include an access network element and a core network element or an OAM network element. Furthermore, in one possible implementation, the functions of the access network element can be performed by the core network element or the OAM network element; that is, they can be executed by the core network element or the OAM network element. Figure 4 The steps performed by the access network element include, for example, steps S402 and S405. In another possible implementation, the functions of the core network element or OAM network element can be performed by the access network element; that is, the access network element can execute these functions. Figure 4 The steps performed by the core network element or OAM network element, for example, step S408. It is understood that this applies when the function of the access network element can be performed by the core network element or OAM network element, or vice versa. Figure 4 The steps for interaction between access network elements and core network elements or OAM network elements can be omitted, such as steps S407, S09, and S413.

[0181] When AI applications and wireless network infrastructure are shared, the interactive diagrams of the embodiments of this application can be as follows: Figure 5 As shown. In Figure 5 This can include terminal equipment and access network equipment. Access network equipment can include BRF network elements, BMF network elements, and BAF network elements. BMF network elements are used to manage the switching of communication and computing resources, BRF network elements are used to provide air interface services, and BAF network elements are used to provide computing power services for AI applications.

[0182] The first network element can be a BRF network element, the second network element can be a terminal device, and the third network element can be a BAF network element. For example... Figure 5 As shown, the communication method may include, but is not limited to, the following steps:

[0183] Step S501: The terminal device sends a first request to the BRF network element.

[0184] Accordingly, the BRF network element receives a first request from the terminal device. The first request is used to request model training or model inference.

[0185] The definition of the first request can be found in [reference]. Figure 4 Step S401 in the process.

[0186] Optionally, in step S502, the BRF network element determines whether it meets the computational power requirements for model training or model inference.

[0187] Upon receiving the first request, a BRF network element can determine whether it meets the computational power requirements for model training or inference based on the model information in the first request. The method for determining whether a BRF network element meets the computational power requirements for model training or inference can be referred to the foregoing and will not be repeated here.

[0188] Optionally, if the BRF network element determines that it meets the computing power requirements for model training or model inference, then steps S503 to S506 are executed, and the BRF network element performs model training or uses the model for model inference, without executing other steps.

[0189] Optionally, in step S503, the BRF network element sends a third response to the terminal device. Correspondingly, the terminal device receives the third response from the BRF network element.

[0190] The third response is used to indicate consent to model training or model inference. Further, the third response can also be used to instruct the terminal device to send the model. Optionally, the third response can specifically be used to instruct the BRF network element to consent to model training or model inference.

[0191] Optionally, in step S504, the terminal device sends the fourth information to the BRF network element.

[0192] Correspondingly, the BRF network element receives the fourth information from the terminal device.

[0193] The fourth piece of information is used to indicate the model. The way the fourth piece of information indicates the model can be referred to above, and will not be repeated here.

[0194] Optionally, in step S505, the BRF network element performs model training or uses the model for model inference.

[0195] Understandably, when the first request is used to instruct the model to be trained, the BRF network element will train the model based on the first request. When the first request is used to instruct the model to be used for inference, the BRF network element will use the model to perform inference based on the first request.

[0196] Optionally, in step S506, the BRF network element sends the training model or inference results to the terminal device.

[0197] Correspondingly, the terminal device receives the training model or inference results from the BRF network element.

[0198] It is understandable that when a BRF network element trains a model, it sends the training model to the terminal device; when a BRF network element uses the model for inference, it sends the inference result to the terminal device.

[0199] Optionally, if the BRF network element determines that it does not meet the computing power requirements for model training or model inference, then step S507 is executed, in which the BRF network element sends a second request to the BAF network element, requesting the BAF network element to perform model training or use the model for model inference.

[0200] Optionally, in step S507, the BRF network element sends a second request to the BAF network element.

[0201] Correspondingly, the BAF network element receives the second request from the BRF network element.

[0202] The second request is used to request model training or model inference. The second request may include first information, which indicates information about the model. The first information may include the number of model parameters, the amount of data required for model training or inference, and performance metrics required for model training or inference. In one possible implementation, the second request may include other information besides the first information, such as an identifier for the second request or an identifier for the agent. In one possible implementation, when the second request includes an identifier for the second request, the identifier may be the same as or different from the identifier for the first request; this embodiment does not limit this.

[0203] In one possible implementation, the BRF network element sending a second request to the BAF network element can be achieved by the BRF network element sending the second request to the BAF network element through the BMF network element. That is, the BRF network element sends a second request to the BMF network element, and after receiving the second request from the BRF network element, the BMF network element then sends the second request to the BAF network element.

[0204] Optionally, in step S508, the BAF network element determines whether it meets the computing power requirements for model training or model inference.

[0205] Upon receiving the second request, the BAF element can determine whether it meets the computational power requirements for model training or inference based on the first information in the second request. The method by which the BAF element determines whether it meets the computational power requirements for model training or inference can be referred to the foregoing and will not be repeated here.

[0206] Optionally, in step S509, the BAF network element sends a first response to the BRF network element.

[0207] Correspondingly, the BRF network element receives the first response from the BAF network element.

[0208] The first response is used to indicate consent or refusal to train the model or use the model for inference.

[0209] In one possible implementation, the method by which the BAF network element sends the training model or inference result to the BRF network element can be that the BAF network element sends the training model or inference result to the BRF network element through the BMF network element. It can be understood that sending the training model or inference result from the BAF network element to the BRF network element can mean that the BAF network element sends the training model or inference result to the BMF network element, and after receiving the training model or inference result from the BAF network element, the BMF network element then sends the training model or inference result to the BRF network element.

[0210] Optionally, in step S510, the BRF network element sends a second response to the terminal device. Correspondingly, the terminal device receives the second response from the BRF network element. The second response is used to indicate agreement or refusal to train the model or use the model for model inference. Optionally, if the first response indicates refusal to train the model or use the model for model inference, the second response can also be used to instruct the terminal device to send the model to the BRF network element.

[0211] Optionally, if the first response is used to indicate refusal to train the model or use the model for model inference, step S511 is performed.

[0212] Optionally, in step S511, the terminal device sends second information to the BAF network element. The second information is used to indicate the model.

[0213] Correspondingly, the BAF network element receives the second information from the terminal device.

[0214] The second information indication model can be referenced as described above, and will not be repeated here.

[0215] In one possible implementation, the terminal device sends the second information to the BAF network element by sending the second information to the BAF network element through the BRF network element and the BMF network element. It can be understood that sending the second information to the BAF network element through the BRF network element and the BMF network element could mean that the terminal device sends the second information to the BRF network element, the BRF network element then sends the second information to the BMF network element, and the BMF network element then sends the second information to the BAF network element.

[0216] Optionally, in step S512, the BAF network element performs model training or uses the model for model inference.

[0217] Understandably, when the second request is used to instruct the model to be trained, the BAF element will train the model based on the second request. When the second request is used to instruct the model to be used for inference, the BAF element will use the model for inference based on the second request.

[0218] Optionally, in step S513, the BAF network element sends the training model or inference results to the terminal device.

[0219] Correspondingly, the terminal device receives the training model or inference results from the BAF network element.

[0220] It is understandable that when a BAF network element trains a model, it sends the training model to the terminal device; when a BAF network element uses the model for inference, it sends the inference result to the terminal device.

[0221] The training model or inference results may be included in the third information. Furthermore, the third information may also include one or more of a request identifier and a completion status indication. The completion status indication is used to indicate that model inference or model training is complete, or to indicate that model inference or model training is not complete. Optionally, the third information may include the training model or inference results, but may not need to include a completion status indication, because sending the training model or inference results can implicitly assume that the completion status indication indicates that model inference or model training is complete.

[0222] The method by which a BAF network element sends training models or inference results to a terminal device can be that the BAF network element sends the training models or inference results to the terminal device through BMF and BRF network elements. It can be understood that sending training models or inference results to a terminal device through BMF and BRF network elements means that the BAF network element sends the training model or inference results to the BMF network element, the BMF network element sends the training model or inference results to the BRF network element, and then the BRF network element sends the training model or inference results to the terminal device.

[0223] and Figure 4 Compared to the embodiments shown, in Figure 5 In the illustrated embodiment, the BRF or BMF does not need to determine (or confirm) whether to allow the transmission of the first information. Furthermore, the computing power determination (i.e., determining whether the requirements for model training or model inference are met) can be based on the physical entity corresponding to a specific functional network element. For example, a BRF network element can determine whether it meets the requirements for model training or model inference based on its corresponding physical entity (e.g., memory, video memory, CPU, and / or GPU); a BAF network element can determine whether it meets the requirements for model training or model inference based on its corresponding physical entity (e.g., memory, video memory, CPU, and / or GPU). It is understood that the physical entity corresponding to a specific functional network element can refer to the physical entity allocated to the specific functional network element. Additionally, in one possible implementation, in this embodiment, the terminal device can include a suggested computing power indication in the first request sent when it needs to train the model or use the model for model inference the next time, based on the computing power used in the previous model training or model inference. This indication is used to recommend a BRF network element or a BAF network element to train the model or use the model for model inference. This can reduce the latency of BRF network elements in determining computing power (whether the BRF network element meets the computing power requirements for model training or model inference) and in allocating resources.

[0224] exist Figure 6 In this context, the first network element is either a core network element or an OAM network element, the second network element is an access network element, and the third network element is an OTT server. This may include, but is not limited to, the following steps:

[0225] Step S601: The access network element sends a first request to the core network element or the OAM network element.

[0226] Accordingly, the core network element or OAM network element receives the first request from the access network element. The first request is used to request model training or model inference.

[0227] Optionally, the definition of the first request can be found in [reference]. Figure 4 Step S401 in the process.

[0228] Optionally, in step S602, the core network element or OAM network element determines whether the core network element or OAM network element meets the computing power requirements for model training or model inference.

[0229] Upon receiving the first request, the core network element or OAM element can determine whether it meets the computational power requirements for model training or inference based on the model information in the first request. The method for determining whether the core network element or OAM element meets the computational power requirements for model training or inference can be found above and will not be repeated here.

[0230] Optionally, if the core network element or OAM network element determines that the core network element or OAM network element meets the computing power requirements for model training or model inference, then steps S603 to S606 are executed, and the core network element or OAM network element performs model training or uses the model for model inference, and no other steps are executed.

[0231] Optionally, in step S603, the core network element or OAM network element sends a third response to the access network element. Correspondingly, the access network element receives the third response from the core network element or OAM network element.

[0232] The third response is used to indicate consent to model training or model inference. Further, the third response can also be used to instruct access network elements to send the model. Optionally, the third response can specifically be used to instruct core network elements or OAM network elements to consent to model training or model inference.

[0233] Optionally, in step S604, the access network element sends the fourth information to the core network element or the OAM network element.

[0234] Correspondingly, the core network element or OAM network element receives the fourth information from the access network element.

[0235] The fourth piece of information is used to indicate the model. The way the fourth piece of information indicates the model can be referred to above, and will not be repeated here.

[0236] Optionally, in step S605, the core network element or OAM network element performs model training or uses the model for model inference.

[0237] Understandably, when the first request is used to instruct model training, the core network element or OAM element will train the model based on the first request. When the first request is used to instruct model inference, the core network element or OAM element will use the model for model inference based on the first request.

[0238] Optionally, in step S606, the core network element or OAM network element sends the training model or inference result to the access network element.

[0239] Correspondingly, the access network element receives the training model or inference results from the core network element or the OAM network element.

[0240] It is understandable that when a core network element or OAM element trains a model, it sends the training model to the access network element. When a core network element or OAM element uses the model for model inference, it sends the inference result to the access network element.

[0241] Optionally, in step S607, the core network element or OAM network element determines whether to allow the transmission of the first message.

[0242] The first piece of information is used to indicate the model's information. This first piece of information may include the number of model parameters, the amount of data used for model training or inference, and performance metrics required for model training or inference.

[0243] For example, a core network element or an OAM network element can determine whether to allow the first information by determining whether the model's input and output involve intra-network privacy data. If the core network element or OAM network element determines that the model's input and output involve intra-network privacy data, then the core network element or OAM network element refuses to send the first information; if the core network element or OAM network element determines that the model's input and output do not involve intra-network privacy data, then the core network element or OAM network element allows the sending of the first information. In this embodiment, the information belonging to privacy data and / or information not belonging to privacy data can be predefined by the 3GPP standard or pre-configured by the vendor of the core network element or OAM network element; this embodiment does not limit this.

[0244] Optionally, if the core network element or OAM network element refuses to send the first information, step S608 is executed, and no other steps are executed.

[0245] Optionally, in step S608, the core network element or OAM network element sends a second response to the access network element.

[0246] Correspondingly, the access network element receives a second response from the core network element or the OAM network element.

[0247] The second response is used to indicate a refusal to train the model or use the model for inference.

[0248] Optionally, if the core network element or OAM network element permits the transmission of the first information, step S609 is executed.

[0249] Optionally, in step S609, the core network element or OAM network element sends a second request to the OTT server.

[0250] Correspondingly, the OTT server receives a second request from a core network element or an OAM network element.

[0251] The second request is used to request model training or model inference. In one possible implementation, the second request may include other information besides the first information, such as an identifier for the second request or an identifier for the agent. In one possible implementation, when the second request includes an identifier for the second request, the identifier for the second request may be the same as or different from the identifier for the first request; this embodiment does not limit this.

[0252] Optionally, in step S610, the OTT server sends a first response to the core network element or the OAM network element.

[0253] Correspondingly, the core network element or OAM network element receives the first response from the OTT server.

[0254] The first response is used to indicate consent to train the model or use the model for inference.

[0255] Optionally, in step S611, the core network element or OAM network element sends a second response to the access network element.

[0256] Correspondingly, the access network element receives a second response from the core network element or the OAM network element.

[0257] The second response indicates consent to train or infer the model. Optionally, the second response may also instruct the access network element to send the model to the OTT server.

[0258] Optionally, in step S612, the access network element sends second information to the OTT server, the second information being used to indicate the model.

[0259] Correspondingly, the OTT server receives the second information from the access network element.

[0260] The second information indication model can be referenced as described above, and will not be repeated here.

[0261] The access network element can send the second information to the OTT server either directly or through a core network element or an OAM element. Specifically, sending the second information through a core network element or an OAM element means that the access network element sends the second information to the core network element or OAM element, which then forwards it to the OTT server.

[0262] Optionally, in step S613, the OTT server trains the model or uses the model for model inference.

[0263] Understandably, when the second request is to instruct the model to be trained, the OTT server will train the model based on the second request. When the second request is to instruct the model to be used for inference, the OTT server will use the model for inference based on the second request.

[0264] Optionally, in step S614, the OTT server sends the training model or inference results to the access network element.

[0265] Correspondingly, the access network element receives the training model or inference results from the OTT server.

[0266] It is understandable that when the OTT server trains the model, it sends the training model to the access network element; when the OTT server uses the model for inference, it sends the inference result to the access network element.

[0267] The training model or inference results may be included in the third information. Furthermore, the third information may also include one or more of a request identifier and a completion status indication. The completion status indication is used to indicate that model inference or model training is complete, or to indicate that model inference or model training is not complete. Optionally, the third information may include the training model or inference results, but may not need to include a completion status indication, because sending the training model or inference results can implicitly assume that the completion status indication indicates that model inference or model training is complete.

[0268] The OTT server can send training models or inference results to access network elements in two ways: either directly, or via core network elements or OAM elements. Specifically, sending training models or inference results via core network elements or OAM elements means the OTT server sends the training models or inference results to the core network elements or OAM elements, which then forward them to the access network elements.

[0269] When AI applications and wireless network infrastructure are shared, the interactive diagrams of the embodiments of this application can be as follows: Figure 7 As shown. In Figure 7 In this diagram, the first network element is a BRF network element, the second network element is a BMF network element, and the third network element is a BAF network element. For example... Figure 7 As shown, the communication method may include, but is not limited to, the following steps:

[0270] Step S701: The BRF network element sends a first request to the BMF network element.

[0271] Accordingly, the BMF network element receives the first request sent by the BRF network element. This first request is used to request model training or model inference.

[0272] The definition of the first request can be found in [reference]. Figure 4 Step S401 in the process.

[0273] Step S702: The BMF network element sends a second request to the BAF network element.

[0274] Correspondingly, the BAF network element receives the second request from the BMF network element.

[0275] The second request is used to request model training or model inference. The second request may include first information, which indicates information about the model. The first information may include the number of model parameters, the amount of data required for model training or inference, and performance metrics required for model training or inference. In one possible implementation, the second request may include other information besides the first information, such as an identifier for the second request or an identifier for the agent. In one possible implementation, when the second request includes an identifier for the second request, the identifier may be the same as or different from the identifier for the first request; this embodiment does not limit this.

[0276] Optionally, in step S703, the BAF network element determines whether it meets the computing power requirements for model training or model inference.

[0277] Upon receiving the second request, the BAF element can determine whether it meets the computational power requirements for model training or inference based on the first information in the second request. The method by which the BAF element determines whether it meets the computational power requirements for model training or inference can be referred to the foregoing and will not be repeated here.

[0278] Optionally, in step S704, the BAF network element sends a first response to the BMF network element.

[0279] Correspondingly, the BMF network element receives the first response from the BAF network element.

[0280] The first response is used to indicate agreement or refusal to train or infer the model. It is understood that if the BAF network element meets the computational power requirements for model training or inference, the first response indicates agreement to train or infer the model; if the BAF network element does not meet the computational power requirements, the first response indicates refusal to train or infer the model.

[0281] Optionally, in step S705, the BMF network element sends a second response to the BRF network element.

[0282] Accordingly, the BRF network element receives a second response from the BMF network element. It is understood that the second response is determined based on the first response. If the first response indicates agreement to train or use the model for inference, the second response indicates agreement to train or use the model for inference; if the second response indicates refusal to train or use the model for inference, the second response indicates refusal to train or use the model for inference.

[0283] If the second response indicates consent or refusal to train the model or use the model for model inference, step S706 is executed.

[0284] Optionally, in step S706, the BRF network element sends the second information to the BAF network element.

[0285] The second information indication model can be referenced as described above, and will not be repeated here.

[0286] In one possible implementation, the method by which the BRF network element sends the second information to the BAF network element can be that the BRF network element sends the second information to the BAF network element through the BMF network element. It can be understood that sending the second information from the BRF network element to the BAF network element through the BMF network element can mean that the BRF network element sends the second information to the BMF network element, and then the BMF network element sends the second information to the BAF network element.

[0287] Optionally, in step S707, the BAF network element performs model training or uses the model for model inference.

[0288] Understandably, when the second request is used to instruct the model to be trained, the BAF element will train the model based on the second request. When the second request is used to instruct the model to be used for inference, the BAF element will use the model for inference based on the second request.

[0289] Optionally, in step S708, the BAF network element sends the training model or inference result to the BRF network element.

[0290] Correspondingly, the BRF network element receives the training model or inference results from the BAF network element.

[0291] It is understandable that when a BAF network element trains a model, it sends the training model to the terminal device; when a BAF network element uses the model for inference, it sends the inference result to the terminal device.

[0292] The training model or inference results may be included in the third information. Furthermore, the third information may also include one or more of a request identifier and a completion status indication. The completion status indication is used to indicate that model inference or model training is complete, or to indicate that model inference or model training is not complete. Optionally, the third information may include the training model or inference results, but may not need to include a completion status indication, because sending the training model or inference results can implicitly assume that the completion status indication indicates that model inference or model training is complete.

[0293] The method by which a BAF network element sends a training model or inference result to a BRF network element can be that the BAF network element sends the training model or inference result to the BRF network element through a BMF network element. This can be understood as the BAF network element sending the training model or inference result to the BRF network element via a BMF network element, and then the BMF network element sending the training model or inference result to the BRF network element.

[0294] The methods of the embodiments of this application have been described in detail above, and the apparatus of the embodiments of this application is provided below.

[0295] Please see Figure 8 , Figure 8This is a schematic diagram of a communication device provided in an embodiment of this application. The communication device may include a transceiver unit 801 and a processing unit 802. The transceiver unit 801 may be a device with signal input (receiving) or output (transmitting) capabilities, used for signal transmission with other devices or other components within a device. The processing unit 802 may be a device with processing capabilities, including one or more processors, used to execute instructions (or code or programs), for example, processing communication protocols and communication data. The communication device may be a first network element. The first network element may be an access network element, a core network element, or an Operation Administration and Maintenance (OAM) network element, or it may be a component, part, or circuit within an access network element, core network element, or OAM network element, or it may be a chip or chip system applicable to an access network element, core network element, or OAM network element, or the first network element may also be a logic module or software capable of implementing all or part of the functions of an access network element, core network element, or OAM network element.

[0296] In one embodiment, when the communication device can be a first network element, wherein:

[0297] Transceiver unit 801 is used to receive a first request from the second network element, the first request being used to request model training or to use the model for model inference.

[0298] The transceiver unit 801 is also configured to send a second request to a third network element. The second request includes first information, which is used to indicate information about the model. The second request is used to request model training or model inference using the model.

[0299] It should be noted that the implementation of each unit can also be referenced accordingly. Figures 3 to 7 A corresponding description of any of the method embodiments.

[0300] Please see Figure 9 , Figure 9 This is a schematic diagram of another communication device provided in an embodiment of this application. For example... Figure 9 As shown, the communication device may include a processor 111. The processor 111, also referred to as a processing unit, can implement certain control functions. When the processor 111 runs, it causes the communication device to execute the functions described in this embodiment. Figures 3 to 7 Any method described.

[0301] like Figure 9The communication device shown may further include a storage medium 112, which may also be referred to as a storage unit or a memory. Instructions 114 are stored on the storage medium 112. These instructions 114 can be executed on the processor 111, causing the communication device to perform the functions described in this embodiment. Figures 3 to 7 Any method described.

[0302] Optionally, the processor 111 may include instructions 113, which can be executed on the processor 111 to cause the communication device to perform the actions described in this embodiment. Figures 3 to 7 Any method described.

[0303] The communication device can be a first network element, an access network element, a core network element, or an OAM network element, etc., used to implement the method described in the method embodiments. However, the scope of the device described in this application is not limited to this; the communication device can be a standalone device or part of a larger device. For example, the communication device can be:

[0304] (1) An independent integrated circuit IC, or chip, or chip system or subsystem;

[0305] (2) A collection of one or more ICs, optionally, the collection of ICs may include a storage component for storing data and / or instructions;

[0306] (3) ASIC, such as modems;

[0307] (4) Modules that can be embedded in other devices.

[0308] This application also provides a computer-readable storage medium storing instructions that, when executed by a computer or processor, can implement the relevant steps in the communication method provided in the above-described method embodiments.

[0309] This application also provides a computer program product including instructions that, when executed by a computer or processor, cause one or more steps of any of the above-described communication methods to be performed. If the constituent modules of the aforementioned devices are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0310] This application provides a chip or chip system including at least one processor for calling and executing instructions stored in a memory, causing a communication device with the chip installed to perform any of the methods described above.

[0311] This application also provides another chip, including a processor and a memory, wherein the processor is used to call and execute instructions stored in the memory, causing a communication device with the chip installed to perform any of the methods described above.

[0312] This application embodiment also provides another chip, including: an input interface, an output interface, and a processing circuit. The input interface, the output interface, and the processing circuit are connected via internal connection paths. The processing circuit is used to execute any of the methods described above. Optionally, the chip also includes a memory. The input interface, the output interface, the processor, and the memory are connected via internal connection paths. The processor is used to execute code in the memory. When the code is executed, the processor is used to execute any of the methods described above.

[0313] This application also provides another chip system, including at least one processor and a communication interface, wherein the communication interface and at least one processor are interconnected via a line, and the at least one processor is used to run computer programs or instructions to perform any of the methods described above. This chip system may be composed of chips, or may include chips and other discrete devices.

[0314] This application also provides a communication system, which includes a first communication device and an OTT device, as detailed in the following description. Figures 3 to 7 Any of the methods shown. The system may also include a second communication device, terminal equipment, or access network device, etc., without limitation.

[0315] It should be understood that the memory mentioned in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Non-volatile memory can be a hard disk drive (HDD), a solid-state drive (SSD), ROM, programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be RAM, which is used as an external cache. Memory is any other medium capable of carrying or storing desired program code having an instruction or data structure form and accessible by a computer, but is not limited thereto. The memory in the embodiments of this application can also be a circuit or any other device capable of implementing a storage function for storing program instructions and / or data.

[0316] It should also be understood that the processor mentioned in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor, or any conventional processor, etc.

[0317] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA, or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, the memory (storage module) is integrated into the processor.

[0318] It should be noted that the memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.

[0319] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments provided herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0320] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0321] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0322] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0323] The steps in the methods of this application can be adjusted, combined, or deleted according to actual needs. Each step in each embodiment can be partially performed (for example, the communication device may not perform the steps performed by the communication device in the above embodiments). The execution order of different steps can be changed. The embodiments described herein can be combined with other embodiments, different embodiments can be combined with each other, and different steps of different embodiments herein can be combined.

[0324] The modules / units in the device of this application embodiment can be merged, divided, and deleted according to actual needs.

[0325] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments.

[0326] In this application, it may refer to a communication protocol or specification, such as the 3GPP communication protocol.

[0327] In this application, unless otherwise specified, "at least one" means "one or more".

[0328] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the embodiments of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0329] In the embodiments of this application, "including" can refer to a relationship of inclusion or an equality relationship. For example, A includes B, which could mean that A includes B and may also include other content, or that A and B are the same content.

[0330] In the description of this application, unless otherwise stated, " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B can mean A or B. "And / or" in this application is merely a description of the relationship between the related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Furthermore, in the description of this application, unless otherwise stated, "multiple" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0331] In this application, the words "exemplarily" or "for example" are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design that is described as "exemplarily" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "exemplarily" or "for example" is intended to present the relevant concepts in a specific manner.

[0332] In the description of this application, "instruction" can include direct and indirect instructions, as well as explicit and implicit instructions. The information indicated by a certain piece of information is called the information to be instructed. In the specific implementation process, there are many ways to instruct the information to be instructed. For example, the information to be instructed can be directly instructed, such as by instructing the information itself or its index. Alternatively, the information to be instructed can be indirectly indicated by instructing other information, where there is a relationship between the indicated other information and the information to be instructed. Another example is that only a part of the information to be instructed can be indicated, while the other parts are known or pre-agreed upon. Furthermore, the instruction of specific information can be achieved by using a pre-agreed (such as an agreement) arrangement of various pieces of information, thereby reducing the instruction overhead to some extent.

[0333] It is understood that in the description of this application, "when," "if," and "if" all refer to the device making a corresponding action under certain objective circumstances, and are not time-limited, nor do they require the device to make a judgment action when it is implemented, nor do they mean that there are other limitations.

[0334] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

Claims

1. A communication method, characterized in that, Applied to the first network element, the method includes: Receive a first request from the second network element, the first request being used to request model training or to use the model for model inference; A second request is sent to a third network element. The second request includes first information, which is used to indicate information about the model. The second request is used to request model training or model inference using the model.

2. The method according to claim 1, characterized in that, Sending the second request to the third network element includes: If the first network element does not meet the computing power requirements for model training or model inference, a second request is sent to the third network element.

3. The method according to claim 2, characterized in that, The method further includes: It is determined that the first network element does not meet the computing power requirements for model training or model inference.

4. The method according to claim 3, characterized in that, The step of determining that the first network element does not meet the computing power requirements for model training or model inference includes: It is determined that the memory capacity of the first network element is less than the size of the model; and / or, Determine that the processor utilization rate of the first network element is higher than a first threshold; and / or, The processing speed of the processor of the first network element is lower than the second threshold.

5. The method according to any one of claims 1-4, characterized in that, The first network element is an internal network element of the 3GPP network, and the third network element is an external network element of the 3GPP network.

6. The method according to any one of claims 1-4, characterized in that, Both the first network element and the third network element are internal network elements of the 3GPP (3rd Generation Partnership Project) network.

7. The method according to any one of claims 1-6, characterized in that, Sending the second request to the third network element includes: If the first network element permits the transmission of the first information, a second request is sent to the third network element.

8. The method according to any one of claims 1-7, characterized in that, The method further includes: A first response is received from the third network element, the first response being used to indicate agreement or refusal to train the model or use the model for model inference.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: A second response is sent to the second network element, the second response being used to indicate agreement or refusal to train the model or use the model for model inference.

10. The method according to any one of claims 1-9, characterized in that, The method further includes: Receive second information from the second network element, the second information being used to indicate the model.

11. The method according to any one of claims 1-10, characterized in that, The method further includes: Receive the training model or the inference result of the model from the third network element; Send the training model of the model or the inference result of the model to the second network element.

12. The method according to any one of claims 1-11, characterized in that, The first request includes one or more of the following information: the identifier of the first request, the identifier of the second network element, the identifier of the agent, the identifier of the model, the number of parameters of the model, the amount of data used for model training or model inference, the performance index requirements for model training or model inference, and the input and output description of the model.

13. A communication system, characterized in that, The system includes a first network element and a second network element. The second network element is used to send a first request to the first network element, the first request being used to request model training or to use the model for model inference; The first network element is used to send a second request to the third network element. The second request includes first information, which is used to indicate information about the model. The second request is used to request model training or model inference using the model.

14. The system according to claim 13, characterized in that, The first network element is also configured to send a second request to the third network element if the first network element does not meet the computing power requirements for model training or model inference.

15. The system according to claim 14, characterized in that, The first network element is further configured to determine that the first network element does not meet the computing power requirements for model training or model inference.

16. The system according to claim 15, characterized in that, The first network element is further configured to determine that the capacity of its memory is less than the size of the model; and / or, It is also used to determine that the processor utilization rate of the first network element is higher than a first threshold; and / or, The processing speed of the processor used in the first network element is lower than the second threshold.

17. The system according to any one of claims 13-16, characterized in that, The first network element is an internal network element of the 3GPP network, and the third network element is an external network element of the 3GPP network.

18. The system according to any one of claims 13-16, characterized in that, Both the first network element and the third network element are internal network elements of the 3GPP (3rd Generation Partnership Project) network.

19. The system according to any one of claims 13-18, characterized in that, The first network element is further configured to send a second request to the third network element if the first network element permits the transmission of the first information.

20. The system according to any one of claims 13-19, characterized in that, The third network element is used to send a first response to the first network element, the first response being used to indicate agreement or refusal to train the model or use the model for model inference.

21. The system according to any one of claims 13-20, characterized in that, The first network element is further configured to send a second response to the second network element, the second response being used to indicate agreement or refusal to train the model or use the model for model inference.

22. The system according to any one of claims 13-21, characterized in that, The second network element is also used to send second information to the first network element, the second information being used to instruct the model.

23. The system according to any one of claims 13-22, characterized in that, The third network element is also used to send the training model of the model or the inference result of the model to the first network element; The first network element is also used to send the training model of the model or the inference result of the model to the second network element.

24. The system according to any one of claims 13-23, characterized in that, The first request includes one or more of the following information: the identifier of the first request, the identifier of the second network element, the identifier of the agent, the identifier of the model, the number of parameters of the model, the amount of data used for model training or model inference, the performance index requirements for model training or model inference, and the input and output description of the model.

25. A communication device, characterized in that, Includes a unit for performing the method according to any one of claims 1-12.

26. A communication device, characterized in that, The communication device includes a processor and a storage medium, the storage medium storing instructions that, when executed by the processor, cause the method according to any one of claims 1-12 to be implemented.

27. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes instructions that, when executed by a processor, cause the method according to any one of claims 1-12 to be implemented.

28. A computer program product, characterized in that, The computer program product includes instructions that, when executed by a processor, cause the method according to any one of claims 1-12 to be implemented.

29. A computer program, characterized in that, Includes program code that, when the computer runs the computer program, performs the method as described in any one of claims 1-12.

30. A chip, characterized in that, Includes a processor, the processor being configured to perform the method as described in any one of claims 1-12.

Images