Model operation result acquisition method and apparatus

Abstract

This application discloses a method and apparatus for obtaining model operation results. The method includes: firstly, a network device receives a first request instruction from an OTT device, the first request instruction being used to request a target operation result, the first request instruction including descriptive information of the target operation result, the descriptive information of the target operation result being used to determine a first model and a first operation type; subsequently, the target operation result is obtained by performing an operation corresponding to the first operation type on the first model; finally, the target operation result is sent to the OTT device. This avoids the difficulty in achieving the required AI model training and inference caused by bandwidth limitations of the OTT device itself.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method and apparatus for obtaining model operation results. Background Technology

[0002] Artificial Intelligence (AI) is a new feature in next-generation mobile communication network technology. On one hand, 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 AI technology development. On the other hand, AI's ability to handle complex problems will greatly enrich the application scenarios of next-generation mobile communication networks. Currently, over-the-top (OTT) devices, due to their bandwidth limitations and latency issues, struggle to achieve the necessary AI model training and inference. Summary of the Invention

[0003] This application provides a method and apparatus for obtaining model operation results. In this embodiment, the OTT device sends a request instruction to the network device for the model training or inference results, requesting to obtain the corresponding model training or inference results from the network device. This can avoid the difficulty in achieving the required AI model training and inference caused by its own bandwidth limitations and other issues.

[0004] In a first aspect, this application provides a method for obtaining model operation results, applied to a first node. The method includes: receiving a first request instruction from a second node, the first request instruction being used to request a target operation result, the first request instruction including descriptive information of the target operation result, the descriptive information of the target operation result being used to determine a first model and a first operation type, the first node being a network-side device, and the second node being a cross-platform OTT device; obtaining the target operation result, the target operation result being obtained by performing an operation corresponding to the first operation type on the first model; and sending the target operation result to the second node.

[0005] Based on the above embodiments, the OTT device sends a request command to the network device regarding the model training or inference results, requesting to obtain the corresponding target operation results from the network device. This avoids the difficulty in achieving the required AI model training and inference due to limitations in bandwidth. Furthermore, the interaction information between the OTT device and the network device only includes request commands and response information, without involving other privacy information. This allows for the provision of operation results of the intranet AI model based on requests from external OTT devices while improving data privacy and security.

[0006] The steps to obtain the result of the target operation can be performed by the first node or by the first node through interaction with other nodes.

[0007] The steps for the first node to obtain the result of the target operation may include the following:

[0008] In one feasible implementation, the first node stores model registration information and obtains the target operation result. Specifically, this can be done through the following steps: determining the first model and the first operation type based on the description information of the target operation result and the model registration information; determining the first model operation data based on the first model; and performing the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0009] Based on the above implementation method, the first node only needs to select and activate the corresponding model, determine the model operation type and model operation data through the description information of the target operation result and the model registration information, and determine the corresponding target operation result accordingly. This can avoid other interaction delays when interacting with other devices.

[0010] The description information of the target operation result can include two types of description information: parameter description information and model description information.

[0011] In one feasible implementation, if the description information of the target operation result includes parameters and parameter quality requirements, and the first model operation data is the data to be inferred, then the first model and the first operation type are determined based on the description information of the target operation result and the model registration information. Specifically, this can be done through the following steps: determining the first model and the first operation type based on the parameters, parameter quality requirements, and model registration information, wherein the first operation type is model inference; based on this, for performing the operation corresponding to the first operation type on the first model based on the first model operation data, this can be done through the following steps: inputting the data to be inferred into the first model for processing to obtain the inference result, wherein the inference result is the target operation result.

[0012] In one feasible implementation, if the description information of the target operation result includes model indication information and model quality requirements, and the first model operation data is training sample data, then the first model and the first operation type are determined based on the description information of the target operation result and the model registration information. Specifically, this can be done through the following steps: determining the first model and the first operation type based on the model indication information, model quality requirements, and model registration information, wherein the first operation type is model inference; based on this, for performing the operation corresponding to the first operation type on the first model based on the first model operation data, this can be done through the following steps: training the first model based on the training sample data to obtain the trained first model, wherein the trained first model is the target operation result.

[0013] Based on the aforementioned embodiments, since model operations mainly include model inference and model training operations, clarifying the corresponding operation type based on the description information of the target operation result facilitates the subsequent acquisition of the first model operation data, activation of the first model, and execution of the operation corresponding to the first operation type.

[0014] The steps by which the first node obtains the result of the target operation through interaction with other nodes may include the following:

[0015] In one feasible implementation, a first request instruction is sent to a third node; a target operation result is received from the third node, wherein the target operation result is determined by performing an operation corresponding to a first operation type on a first model based on first model operation data, the first model operation data is determined based on the first model, and the first model and the first operation type are determined based on the description information and model registration information of the target operation result.

[0016] Based on the above implementation method, the first request instruction is sent to other nodes to obtain the target operation result, so that the first node only acts as a node to interact with the OTT device, which reduces the computing overhead of the first node. At the same time, it avoids data leakage within the network when the third node interacts directly with the OTT device, thus ensuring the privacy and security of data within the network.

[0017] In one feasible implementation, the first node stores model registration information, which may be the registration information of models from other nodes in the communication system. Obtaining the target operation result includes: determining the identifier and first operation type of the first model based on the description information of the target operation result and the model registration information; determining the description information of the first model operation data based on the first model; sending a model operation instruction to the third node, the model operation instruction instructing the first model to perform an operation corresponding to the first operation type based on the first model operation data, the model operation instruction carrying the identifier of the first model; sending a first data transmission instruction to the fourth node, the first data transmission instruction instructing the first model operation data to be sent to the third node, the first data transmission instruction carrying the description information of the first model operation data; and receiving the target operation result from the third node.

[0018] Based on the above implementation method, after the first node selects the model, it obtains the target operation results through other nodes. This can reduce the computational pressure on each node and at the same time ensure the privacy and security of data within the network.

[0019] In one feasible embodiment, the first node stores model registration information, which may be the registration information of models from other nodes in the communication system. Obtaining the target operation result includes: determining the identifier and first operation type of the first model based on the description information of the target operation result and the model registration information; determining the description information of the first model operation data based on the first model; sending a first model operation transmission instruction, a first model transmission instruction, and a second data transmission instruction to the third node. The first model operation transmission instruction instructs the model operation instruction to be sent to the fourth node, and the model operation instruction instructs the first model to perform an operation corresponding to the first operation type based on the first model operation data. The first model transmission instruction instructs the first model to be sent to the fourth node, and the second data transmission instruction instructs the first model operation data to be sent to the fourth node. The first model transmission instruction carries the identifier of the first model, and the second data transmission instruction carries the description information of the first model operation data. Finally, the first node receives the target operation result from the fourth node.

[0020] Based on the above implementation method, after the first node selects the model, it obtains the target operation results through other nodes. This can reduce the computational pressure on each node and at the same time ensure the privacy and security of data within the network.

[0021] Furthermore, in one feasible embodiment, before obtaining the target operation result, a permission determination for obtaining the target operation result can be made based on the description information of the target operation result; and when it is determined that the permission determination for obtaining the target operation result is passed, the operation of obtaining the target operation result is executed. At the same time, the description information of the target operation result may also include usage information, and the permission determination for obtaining the target operation result is assisted based on the usage information.

[0022] Based on the above implementation method, before performing the operation of obtaining the target operation result, a permission judgment is made based on the description information of the target operation result, and the operation of obtaining the target operation result is only performed after the permission judgment is passed. This can further protect the privacy and security of network data.

[0023] In one feasible embodiment, the target operation result can be de-identified before being sent to the second node.

[0024] Based on the above implementation methods, desensitizing the target operation results can prevent the leakage of privacy data and further protect the privacy and security of network data.

[0025] In one feasible embodiment, the description information of the target operation result also includes the quality requirements of the target operation result. After sending the target operation result to the second node, feedback information is also received from the second node. If the feedback information includes first information and second request instruction, a new target operation result is obtained. The new target operation result is obtained by performing an operation corresponding to the first operation type on the second model. The second model has the same function as the first model. The first information is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result. Then, the new target operation result is sent to the second node.

[0026] Based on the above implementation method, when the target operation result does not meet the request quality requirements, the first node can also re-acquire a new target operation result, which can improve the flexibility and reliability of the first node in processing the request instruction.

[0027] Based on the foregoing, in one feasible embodiment, if the second model is a model obtained by adjusting the parameters of the first model, then the first node or other nodes that store model registration information also need to update the registration information of the first model according to the registration information of the second model.

[0028] Based on the above implementation, if the new target operation result is obtained by adjusting the parameters of the first model to obtain the second model, then the registration information of the original first model is updated according to the registration information of the second model at this time. In this way, when processing the same request in the future, the model operation can be directly performed based on the second model, which improves the efficiency of processing the same request in the future.

[0029] Secondly, this application provides a method for obtaining model operation results. The method is applied to a second node and includes: sending a first request instruction to a first node, the first request instruction being used to request a target operation result, the first request instruction including descriptive information of the target operation result, the descriptive information of the target operation result being used to determine a first model and a first operation type, the first node being a network-side device and the second node being an OTT device; receiving the target operation result from the first node, the target operation result being obtained by performing an operation corresponding to the first operation type on the first model.

[0030] For the same reasons as in the aforementioned embodiments, the OTT device sends a request instruction to the network device for the model training or inference results, requesting to obtain the corresponding target operation results from the network device. This can avoid the difficulty in achieving the required AI model training and inference due to its own bandwidth limitations and other issues.

[0031] In one feasible implementation, if it is determined that the target operation result does not meet the quality requirements of the target operation result, then a first message and a second request instruction are sent to the second node. The first message is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result. The new target operation result is received from the first node. The new target operation result is obtained by performing an operation corresponding to the first operation type on the second model. The second model has the same function as the first model. If it is determined that the target operation result meets the quality requirements of the target operation result, then a second message is sent to the second node. The second message is used to characterize that the target operation result meets the quality requirements of the target operation result.

[0032] For the same reasons as in the aforementioned embodiments, when the second node determines that the target operation result does not meet the requested quality requirements, it can enable the first node to re-obtain a new target operation result, thereby improving the accuracy of the second node obtaining the target operation result through the first node.

[0033] Thirdly, a communication device is provided, which includes units or modules for performing the methods that may be implemented in either the first or second aspect described above.

[0034] Fourthly, embodiments of this application provide a communication device, which includes at least one processor and a memory; wherein the memory is used to store computer programs or instructions; and at least one processor is used to execute the computer programs or instructions in the memory, such that the methods that may be implemented in any of the first and second aspects described above are executed.

[0035] Fifthly, embodiments of this application provide a communication system, which includes a transmitting end and a receiving end, wherein the transmitting end is used to perform the method of any one of the first aspects described above, and the receiving end is used to perform the method of any one of the second aspects described above.

[0036] Sixthly, embodiments of this application provide a computer-readable storage medium storing computer instructions that, when executed, cause the computer to perform any of the methods described above.

[0037] In a seventh aspect, embodiments of this application provide a computer program product, the computer program product including: computer program code, which, when executed by a computer, causes the computer to perform a method as described above.

[0038] Eighthly, embodiments of this application provide a chip coupled to a memory for reading and executing program instructions in the memory, so that the device in which the chip is located implements any of the methods described above.

[0039] Understandably, the index generation apparatus of the third or fourth aspect provided above is used to execute the methods provided in either the first or second aspect, and the communication system of the fifth aspect, the computer storage medium of the sixth aspect, the computer program product of the seventh aspect, and the chip of the eighth aspect are all used to implement the methods provided in either the first or second aspect. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding methods, and will not be repeated here. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of a model operation result acquisition system provided in an embodiment of this application;

[0041] Figure 2 This application provides a schematic diagram of the structure of a communication system according to an embodiment of the present application.

[0042] Figure 3 A functional architecture diagram of AI air interface technology provided in this application embodiment;

[0043] Figure 4 A flowchart illustrating a method for obtaining model operation results provided in an embodiment of this application;

[0044] Figure 5 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0045] Figure 6 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0046] Figure 7 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0047] Figure 8 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0048] Figure 9 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0049] Figure 10 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0050] Figure 11 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment;

[0051] Figure 12 This is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0052] Figure 13 This is a schematic diagram of another communication device provided in an embodiment of this application;

[0053] Figure 14 This is a schematic diagram of the processing logic of a processing circuit provided in an embodiment of this application. Detailed Implementation

[0054] The technical solutions in the embodiments of this application will be described below with reference to the accompanying drawings. The terms "system" and "network" in the embodiments of this application can be used interchangeably. Unless otherwise stated, " / " indicates that the objects before and after are in an "or" relationship; for example, A / B can represent A or B. "And / or" in this application is merely a description of the relationship between 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" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single 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 one or multiple. Furthermore, to facilitate a clear description of the technical solutions in the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish between network elements and similar items with essentially the same function. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0055] References to "one embodiment" or "some embodiments" in the embodiments described in this application mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0056] The following detailed embodiments further illustrate the objectives, technical solutions, and beneficial effects of this application. It should be understood that the following are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made based on the technical solutions of this application should be included within the scope of protection of this application.

[0057] In the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0058] The following describes the scenarios involved in the embodiments of this application.

[0059] Please see Figure 1 , Figure 1 This application provides a schematic diagram of a model operation result acquisition system, including an OTT device outside the communication system and network-side devices and terminal devices inside the wireless communication system. It is understood that the OTT device primarily interacts with the network-side devices.

[0060] OTT devices refer to devices that are not directly affiliated with a specific communication system but provide services or content via the Internet. These devices and the services they provide do not rely on the infrastructure of a specific telecommunications operator but utilize existing Internet connections (such as mobile data, Wi-Fi, wired broadband, etc.) to transmit data. For example, OTT devices and services are primarily developed and operated by third-party companies, and are not limited to a specific telecommunications network, allowing them to operate across multiple different network platforms.

[0061] The communication system can be a 5th generation (5G) mobile communication system, a future evolution system, or a converged system of multiple communication technologies, or it can be applied to existing communication systems. The application scenarios of the technical solutions provided in this application can include various scenarios, such as machine-to-machine (M2M), macro-micro communication, enhanced mobile broadband (eMBB), ultra-reliable and low-latency communication (uRLLC), and massive machine-type communication (mMTC). These scenarios may include, but are not limited to, communication scenarios between terminal devices, communication scenarios between network-side devices, and communication scenarios between network-side devices and terminal devices. Among these, network-side devices include access network devices and core network devices.

[0062] For example, please refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of a communication system provided in an embodiment of this application. Figure 1 As shown, the communication system includes core network (CN) equipment, access network (AN) equipment, and terminal devices. Core network equipment and access network equipment can be referred to as network-side equipment.

[0063] Access network equipment refers to a device deployed in a wireless access network to provide wireless communication functions for terminal devices. For example, the access network equipment in the embodiments of this application, as a node in the wireless access network, can also be referred to as an access network element, base station (BS), radio access network (RAN) node (or device, or network element), access point (AP), small tower, etc. Meanwhile, network-side equipment includes, but is not limited to: next-generation base stations (g nodeB, gNB), evolved node B (eNB), radio network controller (RNC), node B (NB), base station controller (BSC), base transceiver station (BTS), home base station (e.g., home evolved nodeB, or home node B, HNB), baseband unit (BBU), wireless fidelity (WiFi) access point, world interoperability for microwave access (WiMAX) base station, transmitting and receiving point (TRP), transmitting point (TP), or mobile switching center, etc. In systems employing different wireless access technologies, the names of devices with base station functions may vary. For example, in 5G communication systems, they are called RAN or gNB (5G NodeB); in LTE systems, they are called evolved NodeB (eNB or eNodeB); and in third-generation (3G) communication systems, they are called Node B, etc. For ease of description, in all embodiments of this application, the aforementioned devices providing wireless communication functions for terminal devices are collectively referred to as access network devices or base stations.

[0064] In some deployments of access network equipment, the access network equipment may include centralized units (CUs) and distributed units (DUs), etc. In other deployments, the CU can be further divided into CU-control plane (CP) and CU-user plane (UP), etc. In still other deployments, the access network equipment can also be a radio unit (RO). In yet another deployment, the access network equipment can be an open radio access network (ORAN) architecture, etc. For example, when the access network equipment is an ORAN architecture, the access network equipment in this application embodiment can be an access network element in ORAN, or a module of an access network element, etc. In an ORAN system, CU can also be called open (O)-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.

[0065] Core network equipment performs core network functions, which can be either 5G core network or evolved 5G core network. Taking the 5G core network as an example, CN 200 includes network elements for access and mobility management functions (AMF) responsible for mobility management and access management services; network elements for session management functions (SMF) responsible for session management; network elements for user plane packet routing and forwarding and quality of service (QoS) control; network elements for policy control functions (PCF); network elements for network exposure functions (NEF) responsible for providing network-related status information to application services; and network elements for service communication proxy (SCP) responsible for managing and coordinating communication between network functions (NFs). These core network elements can work independently or be combined to implement certain control functions. For example, AMF, SMF, PCF, NEF, and SCP can be combined into a single core network device.

[0066] Terminal equipment, also known as user equipment (UE) or terminal, is a device with wireless transceiver capabilities. It can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on water (such as on ships); and it can also be deployed in the air (such as on airplanes, balloons, and satellites). UE can include, but is not limited to: user equipment, subscriber unit, user station, mobile station, mobile station, remote station, remote terminal equipment, mobile terminal equipment, user terminal equipment, wireless communication equipment, user agent, user device, cellular phone, smartphone, cordless phone, session initiation protocol (SIP) phone, wireless data card, wireless local loop (WLL) station, personal digital assistant (PDA) computer, tablet computer, handset with wireless communication capabilities, computing device, processing device connected to a wireless modem, laptop computer, in-vehicle equipment, wearable device, machine type communication (MTC) terminal, terminal equipment in the Internet of Things (IoT), home appliances, virtual reality devices, terminal equipment in future 5G networks, or terminal equipment in future evolved PLMNs, etc. Furthermore, terminal equipment typically contains communication modules, circuits, or chips that perform corresponding communication functions. Terminal equipment can also be configured with program instructions for performing corresponding communication functions.

[0067] The prior art involved in the embodiments of this application is described below.

[0068] Please see Figure 3 , Figure 3 This application provides a functional architecture diagram of an AI air interface technology. During the 3GPPRel18-19 phase, the following were proposed for AI air interface technology: Figure 3The NR air interface AI / ML functional framework is shown. The framework aims to encompass a general functional architecture, addressing both model identifier (ID)-based lifecycle management (LCM) and function-based LCM. It primarily includes functions such as management, data collection, model training, inference, and model storage, along with corresponding data / information / instruction flows (i.e., arrows). For example, the data collection node sends training data to the model training node, monitoring data to the management node, and inference data to the inference node. The management node sends performance feedback / retraining request instructions to the model training node. After training / updating, the model training node sends the trained / updated model to the model storage node for storage. The management node sends model transfer / delivery requests to the model storage node, which then transfers / delivers the model to the inference node. The management node and inference node exchange management instructions and inference output.

[0069] For data privacy and security reasons, the aforementioned LCM (Local Management Communication) for AI models is performed within the communication network and does not consider how to provide the operation results of the in-network AI model based on requests from external OTT (Over-The-Top) devices while ensuring data privacy and security. Therefore, this embodiment obtains the corresponding target operation result by sending a request command for the target operation result from an external OTT device, and only feeds back the target operation result to the external OTT device. This allows for the provision of the operation results of the in-network AI model based on requests from external OTT devices while ensuring data privacy and security.

[0070] The embodiments of this application are described in detail below:

[0071] Example 1: The following is a detailed description of the model operation result acquisition method based on all the functional nodes included in the method. These functional nodes include the model operation result request node, the model operation result exposure node, the model operation management node, the data acquisition node, the model storage node, and the model operation node.

[0072] The main functions of the model operation result request node are to initiate a first request instruction and provide feedback on the target operation result based on the response. The model operation result exposure node includes: receiving the first request instruction; determining whether the requested target operation result is permissible to be provided externally and deciding whether to forward the first request instruction; anonymizing the target operation result; initiating a response to the target operation result; receiving and analyzing the feedback on the target operation result and deciding whether to forward the feedback. The functions of the model operation management node include: receiving and analyzing the first request instruction, selecting a model, and determining the first operation type; initiating data transmission, model transmission, model activation, model operation, model update, and model switching instructions; and forwarding the target operation result. The data acquisition node receives data transmission instructions, acquires data, and transmits it to the designated model operation node. The model storage node registers model information with the model operation management node and transmits the corresponding model to the designated model operation node according to the model transmission instructions. The model operation node receives the model and data, completes the corresponding model operation according to the model activation and operation instructions, and provides feedback on the target operation result to the model management node.

[0073] Please see Figure 4 , Figure 4 This is a flowchart illustrating a method for obtaining model operation results provided in an embodiment of this application, as shown below. Figure 4 As shown, the method includes the following steps:

[0074] S101, The model operation result request node sends a first request instruction. Correspondingly, the model operation result exposure node receives the first request instruction, which is used to request the target operation result and includes descriptive information about the target operation result.

[0075] The descriptive information of the target operation result is used to determine the first model and the first operation type. The first request instruction (descriptive information of the target operation result) may specifically include a request identifier, a description of the target operation result (such as parameters or model), quality requirements (parameter accuracy, model format), and purpose information.

[0076] S102. The model operation result exposure node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0077] S103. When the model operation result exposure node determines that the permission judgment for obtaining the target operation result has failed, it sends a first rejection instruction. Correspondingly, the model operation result request node receives the first rejection instruction, which indicates that it refuses to obtain the target operation result.

[0078] The first rejection instruction may include a request identifier and a rejection identifier.

[0079] S104. When the model operation result exposure node determines that the permission judgment for obtaining the target operation result has passed, it sends a first request instruction. Correspondingly, the model operation management node receives the first request instruction.

[0080] S105. The model operation management node determines the identifier of the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0081] Understandably, the model storage node first registers the model with the model operation management node. Model registration information can include: model identifier, model input, model output, model performance (such as accuracy), functional description, format, update time, and source. After selecting the corresponding first model, the model operation management node can directly determine the model storage node where the first model resides.

[0082] S106. The model operation management node determines the description information of the first model operation data based on the first model.

[0083] S107. The model operation management node sends the seventh data transmission command. Correspondingly, the data acquisition node receives the seventh data transmission command.

[0084] The seventh data transmission instruction is used to instruct the first model operation data to be sent to the model operation node. The data transmission instruction may include a request identifier, input description, output description, data volume, format, precision, preprocessing method, and the device identifier corresponding to the model operation node.

[0085] S108. The model operation management node sends the fifth model transmission command. Correspondingly, the model storage node receives the fifth model transmission command.

[0086] The fifth model transmission instruction is used to instruct the first model to be sent to the model operation node. The model transmission instruction may include a request identifier, the identifier of the first model, its format, precision, and the device identifier corresponding to the model operation node.

[0087] S109. The model operation management node sends model operation instructions. Correspondingly, the model operation node receives the model operation instructions.

[0088] The model operation instructions may include a request identifier, an identifier for the first model, a binary activation instruction (0 / 1), and a binary inference / training operation instruction (0 / 1), etc.

[0089] S110, the data acquisition node sends the first model operation data. Correspondingly, the model operation node receives the first model operation data.

[0090] When the data acquisition node sends the first model operation data, it can also carry a request identifier and a data identifier.

[0091] S111, the model storage node sends the first model. Correspondingly, the model operation node receives the first model.

[0092] When the model storage node sends the first model, it can also carry a request identifier and a model identifier.

[0093] S112. The model operation node performs the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0094] S113. The model operation node sends the target operation result. Correspondingly, the model operation management node receives the target operation result.

[0095] S114. The model operation management node sends the target operation result. Correspondingly, the model operation result exposure node receives the target operation result.

[0096] S115, The node exposing the model operation results performs a desensitization operation on the target operation results.

[0097] S116. The model operation result exposure node sends the target operation result. Correspondingly, the model operation result request node receives the target operation result.

[0098] When the model operation result exposure node sends the target operation result, it can also carry a request identifier.

[0099] S117. The model operation result request node determines the feedback information based on the target operation result.

[0100] S118. The model operation result request node sends feedback information. Correspondingly, the model operation result exposure node receives feedback information.

[0101] The feedback information may include a request identifier, second information, or first information and second request instructions.

[0102] S119. The model operation result exposure node analyzes the feedback information.

[0103] S120, The model operation result exposure node sends feedback information. Correspondingly, the model operation management node receives the feedback information.

[0104] Understandably, the feedback information sent by the node exposing the model operation results is either when the feedback information includes the first information or when the feedback information includes both the first information and the second request instruction.

[0105] S121. The model operation management node obtains the new target operation result after updating or switching the model through interaction with the model operation node, data acquisition node and model storage node.

[0106] The current step is performed when the feedback information includes both the first information and the second request instruction. The model update instruction corresponding to the model update may include a request identifier, a model identifier, and an update target (e.g., accuracy reaching a certain threshold). The model switching instruction corresponding to the model switching may include a request identifier, an identifier of the second model, a binary activation instruction (0 / 1), and a binary inference / training operation (0 / 1) instruction.

[0107] S122, The model operation management node sends the new target operation result. Correspondingly, the model operation result exposure node receives the new target operation result.

[0108] S123, The model operation result exposure node performs a desensitization operation on the new target operation result.

[0109] S124. The model operation result exposure node sends the new target operation result. Correspondingly, the model operation result request node receives the new target operation result.

[0110] The embodiments of this application are described below based on the combination of physical nodes and functional nodes.

[0111] Example 2: First, please refer to Figure 5 , Figure 5 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 5 As shown, the method includes the following steps:

[0112] S201, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation.

[0113] The first request instruction includes a description of the target operation result, which is used to determine the first model and the first operation type. The first node is a network-side device, which may include the aforementioned access network device (such as a base station) and core network device. The second node is an OTT device.

[0114] S202, The first node obtains the result of the target operation.

[0115] The target operation result is obtained by performing the operation corresponding to the first operation type on the first model. The first node can obtain the target operation result either by itself or through other nodes. The first operation type can be model inference or model training.

[0116] S203. The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0117] In this system, the first node only needs to obtain the target operation result according to the instruction sent by the second node to request the target operation result, and finally send the target operation result to the second node. At this time, the interaction between the first node and the second node only includes the instruction to request the target operation result and the target operation result. There is no risk of exposing other data, thus ensuring the privacy and security of the data.

[0118] Example 3: The above method mainly outlines the steps of the model operation result acquisition method. The following detailed description focuses on an example where the first node acquires the target operation result based on itself. In this case, the first node can be a core network device or an access network device (such as a base station), and can also serve as a model operation result exposure node, model operation result management node, data acquisition node, model storage node, and model operation node among functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node among functional nodes.

[0119] Please see Figure 6 , Figure 6 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 6 As shown, the method includes the following steps:

[0120] S301, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0121] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0122] S302, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0123] The description of the target operation result can include the parameters or model indication information that need to be obtained (target operation result description). The first node can then make a permission judgment based on the required parameters or model indication information. Furthermore, the description of the target operation result can also include the purpose information of the target operation result. In this case, the permission judgment can be aided by the purpose information. For example, if the second node needs to obtain the location information of a UE, and it is determined that the exposure of this location information will not pose a risk to the UE, then this information can usually be allowed to be fed back to the second node after being approved by the first node. If the second node needs the Reference Signal Received Power (RSRP) of a UE, and it is determined that the exposure of RSRP will pose a risk to the UE, this information is usually not allowed to be fed back to the second node. The purpose-based judgment could be that if the first node obtains RSRP only for participating in certain model training, then it can also be allowed to be fed back to the second node; if the first node obtains RSRP to infer other information about the UE, then it is usually not allowed to be fed back to the second node. It is understandable that the aforementioned permission judgment based on the target operation result description and purpose information can follow certain industry standards and laws and regulations.

[0124] S303. When the first node determines that the permission judgment for obtaining the target operation result has not passed, it sends a first rejection instruction to the second node. The first rejection instruction is used to indicate that the acquisition of the target operation result is rejected.

[0125] If the permission to obtain the target operation result is not passed, the target operation result cannot be obtained for the second node. In this case, an instruction indicating that the acquisition of the target operation result is refused needs to be sent to the second node.

[0126] S304. When the first node determines that the permission judgment for obtaining the target operation result has passed, it determines the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0127] The process of obtaining the target operation result will only proceed after the permission judgment for obtaining the target operation result has been passed. The description information of the target operation result may include a request identifier, a description of the target operation result, quality requirements for the target operation result, and usage information. The description of the target operation result may include parameter or model indication information. Parameters can refer to the inference result of a certain model to be obtained, such as obtaining location information that meets a certain condition. Model indication information can refer to a model description that implements a certain function, such as a model that can obtain corresponding location information based on a certain condition. The corresponding quality requirements for the target operation result may include parameter quality requirements and model quality requirements. Parameter quality requirements may include parameter accuracy, parameter format, etc., while model quality requirements may include model accuracy, model format, etc. Parameter accuracy can refer to the accuracy of the parameters, i.e., the degree of agreement between the model's output value or category label and the actual value; parameter format can refer to the data structure and encoding method of the parameters. Model accuracy can refer to the prediction accuracy of the model on a given dataset; model format can refer to the storage method and structure of the model file, which determines how the model is loaded, parsed, and executed.

[0128] Optionally, if the description information of the target operation result includes parameters and parameter quality requirements, and the first model operation data is the data to be inferred, then the first model and the first operation type are determined based on the parameters and parameter quality requirements. In this case, the first operation type is model inference, and the first model is the model used for inference.

[0129] Specifically, the first node stores the model and its registration information. This registration information can be metadata related to the model saved after training. Based on this registration information, the descriptive information of the target operation result for each model can be determined, thereby finding a model that supports inference to obtain the aforementioned parameters and determining whether the model's output accuracy and format meet the parameter quality requirements. Optionally, a corresponding rule mapping can be preset, using a pre-defined rule base to directly map request parameters to the corresponding inference model. For example, if the first request instruction is to predict a score, a score prediction model is selected; if the first request instruction is to request geographic location information, a geographic-aware score prediction model is preferred. Subsequently, corresponding conditional statements can be set according to the parameter quality requirements to select the most suitable model. Optionally, machine learning algorithms can be used to automatically select the best-matching inference model based on historical requests and model performance records, or an internal model recommendation system can be built to recommend the most suitable inference model based on request parameters and context information. After determining the corresponding model, its applicability can be further confirmed, such as verifying whether the model's functionality fully meets the request requirements, evaluating the model's historical performance, and ensuring that its accuracy, confidence, format, and other indicators meet the request's quality requirements.

[0130] Optionally, if the description information of the target operation result includes model indication information and model quality requirements, and the first model operation data is training sample data, then the first model and the first operation type are determined based on the model indication information and model quality requirements. In this case, the first operation type is model training, and the first model is the initial model that needs to be trained.

[0131] For the same reasons mentioned above, the first node stores model registration information. At this point, the description information of each model can be determined based on the model registration information, thereby finding models that support the functions included in the model's instruction information and determining whether the model's input and output formats meet the model quality requirements. Optionally, a corresponding rule mapping can be preset, that is, using a preset rule base to directly map the requested model to the corresponding model to be trained. For example, if the first request instruction requests a model for predicting scores, then a regression model is selected. Optionally, machine learning algorithms can also be used to automatically select the best-matching model to be trained based on historical requests and model performance records, or an internal model recommendation system can be built to recommend the most suitable model to be trained based on the request and context information. After determining the corresponding model, the model's applicability can be further confirmed, such as verifying whether the model's functions fully meet the request requirements, evaluating the model's historical performance, and ensuring that its accuracy, format, and other indicators meet the quality requirements of the request.

[0132] Optionally, after selecting the first model, model activation is required before performing the operation corresponding to the first operation type on the first model. Therefore, in this embodiment, the first node needs to activate the model after completing the selection of the first model.

[0133] S305, the first node determines the first model operation data based on the first model.

[0134] Once the first model to be operated on is determined, the corresponding first model operation data can be determined based on the type of the model.

[0135] Optionally, if the first model is a model that needs to be used for inference, the operation data of the first model can be determined based on the input description, data format, data precision, data preprocessing method, etc. If the first model is a model that needs to be used for training, the operation data can also be determined based on the input description, output description, data volume, data format, data precision, data preprocessing method, etc.

[0136] For example, if the first request instruction is to request the user's interest rating for a specific movie, then the corresponding first model can be a rating prediction model. If the input of the rating prediction model includes the user's viewing history, click behavior, and movie attributes, then the first model operation data that needs to be determined can include the user's viewing history and click behavior extracted from the user behavior log; movie attributes extracted from the movie database, and the first model operation data needs to meet the corresponding data format and precision. At the same time, it can also be preprocessed based on the corresponding data preprocessing method.

[0137] For another example, if the first request instruction is to request a model that can achieve personalized movie recommendations, then the corresponding first model can be a recommendation system model (deep learning recommendation model). In this case, the input of the recommendation system model can be user behavior data and movie attribute data, and the output can be the interest rating for each movie. The first model operation data that needs to be determined can include user viewing history, click behavior, playback duration and other behavioral data extracted from user behavior logs, movie attribute data such as movie category and actors, and user rating tags for each movie.

[0138] It is understandable that the data obtained above can also be obtained based on context information, which refers to additional information related to the first request instruction.

[0139] S306, the first node performs the operation corresponding to the first operation type on the first model based on the first model operation data, and obtains the target operation result.

[0140] Optionally, when the first operation type is model reasoning, the first node inputs the data to be reasoned into the first model for processing to obtain the reasoning result, which is the target operation result.

[0141] Optionally, when the first operation type is model training, the first model is trained based on the training sample data to obtain the trained first model, and the trained first model is the result of the target operation.

[0142] S307, The first node performs a desensitization operation on the target operation result.

[0143] Anonymization refers to the transformation or concealment of critical information to protect personal privacy and ensure data security. Critical information can refer to datasets containing personal information, such as names, ID numbers, phone numbers, and addresses, or it can be critical business information. Anonymization can include operations such as replacing, randomizing, and encrypting critical information. After anonymization, the privacy and security of data within the network can be further guaranteed.

[0144] S308. The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0145] S309, The second node determines the feedback information based on the target operation result.

[0146] Optionally, if it is determined that the target operation result does not meet the quality requirements of the target operation result, the feedback information is determined to include a first information and a second request instruction. The first information is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result.

[0147] If it is determined that the target operation result meets the quality requirements of the target operation result, then the feedback information is determined to include second information, which is used to characterize that the target operation result meets the quality requirements of the target operation result.

[0148] Furthermore, when it is determined that the result of the target operation does not meet the quality requirements of the target operation result, the determined feedback information may only include the first information, in which case the first node does not need to reacquire the result of the target operation.

[0149] S310, the second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0150] S311, The first node analyzes the feedback information.

[0151] The first node analyzes the feedback information primarily to determine its specific content. There are three possibilities: the feedback information includes the first piece of information and the second request instruction; the feedback information includes the first piece of information; or the feedback information includes the second piece of information.

[0152] S312, The first node determines the second model.

[0153] Optionally, if the feedback information includes the first information and the second request instruction, the first node needs to obtain the new target operation result. Therefore, it needs to redetermine a second model with the same function as the first model based on the first request instruction. The second model might be a new model different from the first model, or it might be a model obtained by adjusting and updating the parameters of the first model. If the second model is a model obtained by adjusting and updating the parameters of the first model, then the first node needs to update the registration information about the first model based on the information of the second model.

[0154] If the feedback information includes only the first information or includes the second information, then the first node does not need to perform the operation of obtaining the new target operation result.

[0155] S313, The first node determines the second model operation data based on the second model.

[0156] The steps for the first node to determine the second model operation data based on the second model are the same as the aforementioned method for obtaining the first model operation data based on the first model, and will not be repeated here. It is understood that the second model operation data may be the same as or different from the first model operation data.

[0157] S314. The first node performs the operation corresponding to the first operation type on the second model based on the second model operation data to obtain a new target operation result.

[0158] Understandably, after determining the second model, the subsequent steps for obtaining the target operation result based on the second model are the same as those for obtaining the target operation result based on the first model. They will not be repeated here.

[0159] S315. The first node performs a desensitization operation on the new target operation result.

[0160] S316. The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0161] Understandably, after receiving a new target operation result, the second node can also determine new feedback information based on that result and send it back to the first node. If the new feedback information also includes the first information and the second request instruction, the first node can again execute the aforementioned steps to obtain a new target operation result. This will not be elaborated further here.

[0162] Example 4: The above method mainly outlines the steps of the model operation result acquisition method. The following detailed description focuses on a first embodiment where the first node sends a first request instruction to other nodes to acquire the target operation result. In this case, the first node can be a core network device, and also serves as the model operation result exposure node among the functional nodes; the second node can be an OTT device, and also serves as the model operation result request node among the functional nodes; the third node can be an access network device, and also serves as the model operation management node, model storage node, and model operation node among the functional nodes; the fourth node can be a terminal device, and also serves as the data acquisition node among the functional nodes.

[0163] Please see Figure 7 , Figure 7 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 7 As shown, the method includes the following steps:

[0164] S401, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0165] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0166] S402, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0167] S403. When the first node determines that the permission judgment for obtaining the result of the target operation has failed, it sends a first rejection instruction. Correspondingly, the second node receives the first rejection instruction, which indicates that the acquisition of the result of the target operation is refused.

[0168] S404. When the first node determines that the permission judgment for obtaining the result of the target operation has passed, it sends a first request instruction. Correspondingly, the third node receives the first request instruction.

[0169] S405. The third node determines the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0170] The third node stores model registration information. The method by which the third node determines the first model and the first operation type based on the description information of the target operation result and the model registration information is the same as the method by which the first node determines the first model and the first operation type based on the description information of the target operation result in the aforementioned embodiment, and will not be repeated here.

[0171] S406, The third node determines the description information of the first model operation data based on the first model.

[0172] The descriptive information of the first model operation data may include input description, output description, data volume, format, precision, and preprocessing method.

[0173] S407, the third node sends a first data transmission instruction, which instructs the first model operation data to be sent to the third node. Correspondingly, the fourth node receives the first data transmission instruction.

[0174] The first data transmission instruction carries descriptive information about the first model operation data.

[0175] S408, the fourth node sends the first model operation data. Correspondingly, the third node receives the first model operation data.

[0176] After receiving the first data transmission instruction, the fourth node needs to obtain the first model operation data based on the description information of the first model operation data carried in the first data transmission instruction, and then send the first model operation data to the third node.

[0177] S409. The third node performs the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0178] Similarly, after completing the model selection for the first model, the third node also needs to activate the model.

[0179] S410, the third node sends the target operation result. Correspondingly, the first node receives the target operation result.

[0180] S411, The first node performs a desensitization operation on the target operation result.

[0181] S412, The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0182] S413, The second node determines the feedback information based on the target operation result.

[0183] S414, The second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0184] S415, The first node analyzes the feedback information.

[0185] S416, The first node sends feedback information. Correspondingly, the third node receives the feedback information.

[0186] Understandably, since there are three possibilities for the feedback information: the feedback information includes the first information and the second request instruction; the feedback information includes the first information; and the feedback information includes the second information. Therefore, if the feedback information includes both the first and second request instructions, or only the first information, the first node will send the feedback information to the third node.

[0187] S417, The third node determines the second model.

[0188] Similarly, if the feedback information includes the first information and the second request instruction, the third node will re-acquire the new target operation result.

[0189] S418, The third node determines the description information of the second model operation data based on the second model.

[0190] S419, The third node sends a third data transmission command. Correspondingly, the fourth node receives the third data transmission command.

[0191] The third data transmission instruction is used to instruct the second model operation data to be sent to the third node, and the third data transmission instruction carries descriptive information about the second model operation data.

[0192] S420, the fourth node sends the second model operation data. Correspondingly, the third node receives the second model operation data.

[0193] S421. The third node performs the operation corresponding to the first operation type on the second model based on the second model operation data to obtain a new target operation result.

[0194] S422, The third node sends the new target operation result. Correspondingly, the first node receives the new target operation result.

[0195] S423, The first node performs a desensitization operation on the new target operation result.

[0196] S424. The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0197] Example 5: The above method mainly outlines the steps of obtaining model operation results. The following detailed description of a second embodiment where the first node sends a first request instruction to other nodes to obtain the target operation result. In this case, the first node can be a core network device, and can also serve as a model operation result exposure node in the functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node in the functional nodes; the third node can be an access network device, and can also serve as a model operation management node in the functional nodes; the fourth node can be a terminal device, and can also serve as a data acquisition node and model storage node in the functional nodes; the fifth node can be a terminal device-side server, and can also serve as a model operation node in the functional nodes.

[0198] Please see Figure 8 , Figure 8 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 8 As shown, the method includes the following steps:

[0199] S501, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0200] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0201] S502, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0202] S503. When the first node determines that the permission judgment for obtaining the result of the target operation has failed, it sends a first rejection instruction. Correspondingly, the second node receives the first rejection instruction, which indicates that the acquisition of the result of the target operation is refused.

[0203] S504. When the first node determines that the permission judgment for obtaining the result of the target operation has passed, it sends a first request instruction. Correspondingly, the third node receives the first request instruction.

[0204] S505, the third node determines the identifier of the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0205] Before the third node determines the identifier of the first model and the first operation type based on the description information of the target operation result, the fourth node should register the model with the third node, that is, manage and record the model. This allows the third node to store the model's registration information, facilitating the determination of the first model's identifier and the fourth node through the model's registration information.

[0206] S506, The third node determines the description information of the operation data of the first model based on the first model.

[0207] The descriptive information of the first model's operational data may include input description, output description, data volume, format, precision, and preprocessing method. It is understandable that the third node primarily determines the descriptive information of the first model's operational data based on the registration information of the first model.

[0208] S507, the third node sends a fourth data transmission command, a second model operation transmission command, and a second model transmission command. Correspondingly, the fourth node receives the fourth data transmission command, the second model operation transmission command, and the second model transmission command.

[0209] The fourth data transmission instruction is used to instruct the first model operation data to be sent to the fifth node, the second model operation transmission instruction is used to instruct the model operation instruction to be sent to the fifth node, the model operation instruction is used to instruct the first model to perform the operation corresponding to the first operation type based on the first model operation data, the second model transmission instruction is used to instruct the first model to be sent to the fifth node, the fourth data transmission instruction carries the description information of the first model operation data, and the second model transmission instruction carries the identifier of the first model.

[0210] S508, the fourth node sends the first model operation data, the first model, and the model operation instructions. Correspondingly, the fifth node receives the first model operation data, the first model, and the model operation instructions.

[0211] Upon receiving the fourth data transmission instruction, the fourth node needs to retrieve the first model operation data based on the description information of the first model operation data carried in the fourth data transmission instruction. Upon receiving the second model transmission instruction, it needs to retrieve the first model from its local storage based on the identifier of the first model carried in the second model transmission instruction. Optionally, the model operation instruction may also carry a corresponding model activation instruction.

[0212] S509. The fifth node performs the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0213] S510, the fifth node sends the target operation result. Correspondingly, the fourth node receives the target operation result.

[0214] S511, the fourth node sends the target operation result. Correspondingly, the third node receives the target operation result.

[0215] S512, the third node sends the target operation result. Correspondingly, the first node receives the target operation result.

[0216] S513, The first node performs a desensitization operation on the target operation result.

[0217] S514. The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0218] S515, The second node determines the feedback information based on the target operation result.

[0219] S516, The second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0220] S517. The first node analyzes the feedback information.

[0221] S518, The first node sends feedback information. Correspondingly, the third node receives the feedback information.

[0222] Understandably, since there are three possibilities for the feedback information: the feedback information includes the first information and the second request instruction; the feedback information includes the first information; and the feedback information includes the second information. Therefore, if the feedback information includes both the first and second request instructions, or only the first information, the first node will send the feedback information to the third node.

[0223] S519, The third node analyzes the feedback information.

[0224] S520: The third node sends a model update command, or the third node sends a model switching command, a fifth data transmission command, and a third model transmission command. Correspondingly, the fourth node receives the model update command, or the fourth node receives the model switching command, the fifth data transmission command, and the third model transmission command.

[0225] The fifth data transmission instruction instructs the sending of the second model operation data to the fifth node, while the third model transmission instruction instructs the sending of the second model to the fifth node. The fifth data transmission instruction carries a description of the second model operation data, and the third model transmission instruction carries an identifier for the second model. Understandably, when the third node sends a model update instruction, the fifth node will subsequently adjust and update the parameters of the first model to determine the second model. Then, based on the second model operation data, it will perform the operation corresponding to the first operation type on the second model. In this case, the second model operation data can be the same as the first model operation data. When the third node sends a model switching instruction, the second model is a new model redefined by the third node, and the description of the second model operation data is redefined based on the information of the second model.

[0226] S521, the fourth node sends a model update command, or the fourth node sends a model switching command, second operation data, and a second model. Correspondingly, the fifth node receives the model update command, or the fifth node receives the model switching command, second operation data, and a second model.

[0227] S522, the fifth node performs the operation corresponding to the first operation type on the second model based on the second model operation data, and obtains a new target operation result.

[0228] Understandably, if the fifth node receives a model update instruction, it needs to adjust and update the parameters of the first model to obtain the second model. At this time, the operation data of the second model can be the operation data of the first model.

[0229] S523, the fifth node sends the new target operation result. Correspondingly, the fourth node receives the new target operation result.

[0230] S524, the fourth node sends the new target operation result. Correspondingly, the third node receives the new target operation result.

[0231] Understandably, if the third node sends the model update command first, then after the fifth node completes the model update, the fourth node will also update the registration information of the first model to the third node based on the information of the second model.

[0232] S525, the third node sends the new target operation result. Correspondingly, the first node receives the new target operation result.

[0233] S526. The first node performs a desensitization operation on the new target operation result.

[0234] S527. The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0235] Example 6: The above method mainly outlines the steps of the model operation result acquisition method. The following detailed description focuses on a third embodiment where the first node sends a first request instruction to other nodes to acquire the target operation result. In one feasible embodiment, the first node can be a core network device, and can also serve as a model operation result exposure node and a model operation result management node among functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node among functional nodes; the third node can be an access network device, and can also serve as a model storage node and a model operation node among functional nodes; the fourth node can be a terminal device, and can also serve as a data acquisition node among functional nodes.

[0236] In another feasible embodiment, the first node can also be an access network device, and can also serve as a model operation result exposure node and a model operation result management node in the functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node in the functional nodes; the third node can also be a terminal device, and can also serve as a model storage node and a model operation node in the functional nodes; the fourth node can also be a sensor, and can also serve as a data acquisition node in the functional nodes.

[0237] Please see Figure 9 , Figure 9 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 9 As shown, the method includes the following steps:

[0238] S601, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0239] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0240] S602, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0241] S603. When the first node determines that the permission judgment for obtaining the result of the target operation has failed, it sends a first rejection instruction. Correspondingly, the second node receives the first rejection instruction, which indicates that the acquisition of the result of the target operation is refused.

[0242] S604. When the first node determines that the permission judgment for obtaining the target operation result has passed, it determines the identifier of the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0243] The model registration information stored in the first node is obtained after the third node registers its model with the first node. Based on this model registration information, the corresponding first model and third node can be selected.

[0244] S605, The first node determines the description information of the first model operation data based on the first model.

[0245] S606, The first node sends the model operation command. Correspondingly, the third node receives the model operation command.

[0246] S607, The first node sends the first data transmission command. Correspondingly, the fourth node receives the first data transmission command.

[0247] The first data transmission instruction is used to instruct the first model operation data to be sent to the third node, and the first data transmission instruction carries descriptive information of the first model operation data.

[0248] S608, the fourth node sends the first model operation data. Correspondingly, the third node receives the first model operation data.

[0249] S609. The third node performs the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0250] S610, the third node sends the target operation result. Correspondingly, the first node receives the target operation result.

[0251] S611, The first node performs a desensitization operation on the target operation result.

[0252] S612, The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0253] S613, The second node determines the feedback information based on the target operation result.

[0254] S614, The second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0255] S615, The first node analyzes the feedback information.

[0256] S616. The first node sends a model update command, or the third node sends a model switching command. Correspondingly, the fourth node receives the model update command, or the fourth node receives the model switching command.

[0257] S617, The first node sends the third data transmission command. Correspondingly, the fourth node receives the third data transmission command.

[0258] The third data transmission instruction is used to instruct the second model operation data to be sent to the third node. It can be understood that when the first node sends a model switching instruction to the third node, it will also send the third data transmission instruction to the fourth node.

[0259] S618, the fourth node sends the second model operation data. Correspondingly, the third node receives the second model operation data.

[0260] S619. The third node performs the operation corresponding to the first operation type on the second model based on the second model operation data to obtain a new target operation result.

[0261] Understandably, if the third node receives a model update instruction, it needs to adjust and update the parameters of the first model to obtain the second model. At this time, the operation data of the second model can be the operation data of the first model.

[0262] S620: The third node sends the new target operation result. Correspondingly, the first node receives the new target operation result.

[0263] Understandably, if the first node sends a model update command first, then after the third node completes the model update, the third node will also update the registration information of the first model to the first node based on the information of the second model.

[0264] S621, The first node performs a desensitization operation on the new target operation result.

[0265] S622, The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0266] Example 7: The above method mainly outlines the steps of the model operation result acquisition method. The following detailed description focuses on a fourth embodiment where the first node sends a first request instruction to other nodes to acquire the target operation result. In this case, the first node can be a core network device, and can also serve as a model operation result exposure node and a model operation result management node among the functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node among the functional nodes; the third node can be a terminal device, and can also serve as a data acquisition node and a model storage node among the functional nodes; the fourth node can be a server on the terminal device side, and can also serve as a model operation node among the functional nodes.

[0267] Please see Figure 10 , Figure 10 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 10 As shown, the method includes the following steps:

[0268] S701, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0269] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0270] S702, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0271] S703. When the first node determines that the permission judgment for obtaining the result of the target operation has failed, it sends a first rejection instruction. Correspondingly, the second node receives the first rejection instruction, which indicates that the acquisition of the result of the target operation is refused.

[0272] S704. When the first node determines that the permission judgment for obtaining the target operation result has passed, it determines the identifier of the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0273] The model registration information is determined after the third node registers the model with the first node. Based on this model registration information, the identifier of the first model and the third node can be determined.

[0274] S705, The first node determines the description information of the first model operation data based on the first model.

[0275] S706, the first node sends a first model operation transmission command, a first model transmission command, and a second data transmission command. Correspondingly, the third node receives the first model operation transmission command, the first model transmission command, and the second data transmission command.

[0276] The first model operation transmission instruction is used to instruct the model operation instruction to be sent to the fourth node. The model operation instruction is used to instruct the first model to perform an operation corresponding to the first operation type based on the first model operation data. The first model transmission instruction is used to instruct the first model to be sent to the fourth node. The second data transmission instruction is used to instruct the first model operation data to be sent to the fourth node. The first model transmission instruction carries the identifier of the first model. The second data transmission instruction carries the description information of the first model operation data.

[0277] S707, the third node sends the first model operation data, the first model, and the model operation instructions. Correspondingly, the fourth node receives the first model operation data, the first model, and the model operation instructions.

[0278] S708, the fourth node performs the operation corresponding to the first operation type on the first model based on the first model operation data, and obtains the target operation result.

[0279] S709, the fourth node sends the target operation result. Correspondingly, the third node receives the target operation result.

[0280] S710, the third node sends the target operation result. Correspondingly, the first node receives the target operation result.

[0281] S711, The first node performs a desensitization operation on the target operation result.

[0282] S712, The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0283] S713, The second node determines the feedback information based on the target operation result.

[0284] S714, the second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0285] S715, The first node analyzes the feedback information.

[0286] S716, the first node sends a model update command, or the first node sends a model switching command, a sixth data transmission command, and a fourth model transmission command. Correspondingly, the fourth node receives the model update command, or the fourth node receives the model switching command, the sixth data transmission command, and the fourth model transmission command.

[0287] The sixth data transmission instruction instructs the sending of the second model operation data to the fourth node, while the fourth model transmission instruction instructs the sending of the second model to the fourth node. The sixth data transmission instruction carries a description of the second model operation data, and the fourth model transmission instruction carries an identifier for the second model. Current and subsequent steps are executed only if the feedback information includes the first information and the second request instruction.

[0288] S717, the third node sends a model update command, or the third node sends a model switching command, second operation data, and a second model. Correspondingly, the fourth node receives the model update command, or the fourth node receives the model switching command, second operation data, and a second model.

[0289] S718, the fourth node performs the operation corresponding to the first operation type on the second model based on the second model operation data, and obtains a new target operation result.

[0290] Understandably, if the fourth node receives a model update instruction, it needs to adjust and update the parameters of the first model to obtain the second model. At this time, the operation data of the second model can be the operation data of the first model.

[0291] S719, the fourth node sends the new target operation result. Correspondingly, the third node receives the new target operation result.

[0292] Understandably, if the third node sends the model update command first, then after the fifth node completes the model update, the fourth node will also update the registration information of the first model to the third node based on the information of the second model.

[0293] S720: The third node sends the new target operation result. Correspondingly, the first node receives the new target operation result.

[0294] S721, The first node performs a desensitization operation on the new target operation result.

[0295] S722, The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0296] Example 8: The above method mainly outlines the steps of the model operation result acquisition method. The following detailed description focuses on the fifth example where the first node sends a first request instruction to other nodes to acquire the target operation result. In this case, the first node can be a core network device, and can also serve as a model operation result exposure node and a model operation result management node among the functional nodes; the second node can be an OTT device, and can also serve as a model operation result request node among the functional nodes; the third node can be a terminal device, and can also serve as a data acquisition node, a model storage node, and a model operation node among the functional nodes.

[0297] Please see Figure 11 , Figure 11 A flowchart illustrating another method for obtaining model operation results provided in this application embodiment is shown below. Figure 11 As shown, the method includes the following steps:

[0298] S801, the second node sends a first request instruction. Correspondingly, the first node receives the first request instruction, which is used to request the result of the target operation and includes descriptive information about the result of the target operation.

[0299] The descriptive information of the target operation result is used to determine the first model and the first operation type.

[0300] S802, The first node makes a judgment on whether to obtain the target operation result based on the description information of the target operation result.

[0301] S803. When the first node determines that the permission judgment for obtaining the result of the target operation has failed, it sends a first rejection instruction. Correspondingly, the second node receives the first rejection instruction, which indicates that the acquisition of the result of the target operation is refused.

[0302] S804. When the first node determines that the permission judgment for obtaining the target operation result has passed, it determines the identifier of the first model and the first operation type based on the description information of the target operation result and the model registration information.

[0303] The model registration information is determined after the third node registers the model with the first node. Based on this model registration information, the identifier of the first model and the third node can be determined.

[0304] S805, The first node determines the description information of the first model operation data based on the first model.

[0305] S806, The first node sends the model operation command. Correspondingly, the third node receives the model operation command.

[0306] The model operation instructions may include the identifier of the first model and a description of the first model operation data.

[0307] S807. The third node performs the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0308] S808: The third node sends the target operation result. Correspondingly, the first node receives the target operation result.

[0309] S809, The first node performs a desensitization operation on the target operation result.

[0310] S810: The first node sends the target operation result. Correspondingly, the second node receives the target operation result.

[0311] S811, The second node determines the feedback information based on the target operation result.

[0312] S812, the second node sends feedback information. Correspondingly, the first node receives the feedback information.

[0313] S813, The first node analyzes the feedback information.

[0314] S814. The first node sends a model update command or a model switching command. Correspondingly, the third node receives the model update command or the third node receives the model switching command.

[0315] The model switching command can carry the identifier of the second model and the description of the operation data of the second model.

[0316] S815, the third node performs the operation corresponding to the first operation type on the second model based on the second model operation data, and obtains a new target operation result.

[0317] Understandably, if the third node receives a model update command, it needs to adjust and update the parameters of the first model to obtain the second model. In this case, the operation data of the second model can be the operation data of the first model. Understandably, if the first node sends a model update command first, then after the third node completes the model update, the third node will also update the registration information of the first model based on the information of the second model.

[0318] S816, the third node sends the new target operation result. Correspondingly, the first node receives the new target operation result.

[0319] S817, The first node performs a desensitization operation on the new target operation result.

[0320] S818, The first node sends the new target operation result. Correspondingly, the second node receives the new target operation result.

[0321] Specifically, in this application, the OTT device sends a request instruction to the network device regarding the model training or inference results, requesting to obtain the corresponding target operation results from the network device. This avoids the difficulty in achieving the required AI model training and inference due to limitations in bandwidth. Simultaneously, after receiving the request instruction from the OTT device, the network device first performs an approval judgment and then performs an anonymization operation before responding to the request instruction. This avoids the request instruction including any private data, thus improving data privacy and security while providing the operation results of the intranet AI model based on the OTT device's request.

[0322] Please see Figure 12 , Figure 12 This is a schematic diagram of a communication device provided in an embodiment of this application. This communication device can be used to execute any of the methods described in the foregoing embodiments.

[0323] like Figure 12 As shown, the communication device includes a processing module 1201 and a transceiver module 1202. The processing module 1201 may be one or more processors, and the transceiver module 1202 may be a transceiver or a communication interface. This communication device can be used to implement the functions of the first node, second node, third node, fourth node, and fifth node involved in any of the above method embodiments. Optionally, the communication device may further include a storage module 1203 for storing the program code and data of the communication device. The processing module 1201 can read instructions and / or data from the storage unit to enable the device to perform the actions of different devices in the aforementioned method embodiments.

[0324] In a first example, the communication device can function as a first device or a chip within a first device, and execute the steps performed by the transmitting end in embodiments one through four of the above method. The transceiver module 1202 supports communication between the first node and the second node, etc. The processing module 1201 can be used to support the execution of actions performed by the first node in the above method embodiments, excluding sending and receiving.

[0325] Specifically, the transceiver module 1202 is used to: receive a first request instruction from the second node, the first request instruction being used to request a target operation result, the first request instruction including descriptive information of the target operation result, the descriptive information of the target operation result being used to determine the first model and the first operation type, the first node being a network-side device, and the second node being a cross-platform OTT device; the processing module 1201 and / or the transceiver module 1202 are used to: obtain the target operation result, the target operation result being obtained by performing an operation corresponding to the first operation type on the first model; the transceiver module 1202 is also used to: send the target operation result to the second node.

[0326] In one feasible implementation, the first node stores model registration information. In terms of obtaining the target operation result, the processing module 1201 is specifically used to: determine the first model and the first operation type based on the description information of the target operation result and the model registration information; determine the first model operation data based on the first model; and perform the operation corresponding to the first operation type on the first model based on the first model operation data to obtain the target operation result.

[0327] In one feasible implementation, the description information of the target operation result includes parameters and parameter quality requirements. The first model operation data is the data to be inferred. The processing module 1201 is specifically used to: determine the first model and the first operation type based on the parameters, parameter quality requirements and model registration information. The first operation type is model inference. The data to be inferred is input into the first model for processing to obtain the inference result, which is the target operation result.

[0328] In one feasible implementation, the description information of the target operation result includes model indication information and model quality requirements. The first model operation data is training sample data. The processing module 1201 is specifically used to: determine the first model and the first operation type based on the model indication information, model quality requirements and model registration information, wherein the first operation type is model inference; train the first model based on the training sample data to obtain the trained first model, wherein the trained first model is the target operation result.

[0329] In one feasible implementation, the transceiver module 1202 is specifically used for: sending a first request instruction to a third node; receiving a target operation result from the third node, wherein the target operation result is determined by performing an operation corresponding to a first operation type on a first model based on first model operation data, the first model operation data is determined based on the first model, and the first model and the first operation type are determined based on the description information and model registration information of the target operation result.

[0330] In one feasible implementation, the first node stores model registration information. The processing module 1201 is specifically used to: determine the identifier and first operation type of the first model based on the description information of the target operation result and the model registration information; determine the description information of the first model operation data based on the first model; the transceiver module 1202 is specifically used to: send a model operation instruction to the third node, the model operation instruction being used to instruct the first model to perform an operation corresponding to the first operation type based on the first model operation data, the model operation instruction carrying the identifier of the first model; send a first data transmission instruction to the fourth node, the first data transmission instruction being used to instruct the first model operation data to be sent to the third node, the first data transmission instruction carrying the description information of the first model operation data; and receive the target operation result from the third node.

[0331] In one feasible implementation, the first node stores model registration information. The processing module 1201 is specifically used to: determine the identifier and first operation type of the first model based on the description information of the target operation result and the model registration information; determine the description information of the first model operation data based on the first model; the transceiver module 1202 is specifically used to: send a first model operation transmission instruction, a first model transmission instruction, and a second data transmission instruction to the third node. The first model operation transmission instruction is used to instruct the model operation instruction to be sent to the fourth node. The model operation instruction is used to instruct the first model to perform an operation corresponding to the first operation type based on the first model operation data. The first model transmission instruction is used to instruct the first model to be sent to the fourth node. The second data transmission instruction is used to instruct the first model operation data to be sent to the fourth node. The first model transmission instruction carries the identifier of the first model, and the second data transmission instruction carries the description information of the first model operation data; and receive the target operation result from the fourth node.

[0332] In one feasible implementation, the processing module 1201 is further configured to: make a permission judgment on obtaining the target operation result based on the description information of the target operation result; and when it is determined that the permission judgment for obtaining the target operation result is passed, perform the operation of obtaining the target operation result.

[0333] In one feasible implementation, the processing module 1201 is further configured to: perform a desensitization operation on the target operation result.

[0334] In one feasible implementation, the transceiver module 1202 is further configured to: receive feedback information from the second node; the processing model and / or the transceiver module 1202 is further configured to: if the feedback information includes first information and a second request instruction, obtain a new target operation result, wherein the new target operation result is obtained by performing an operation corresponding to the first operation type on the second model, the second model has the same function as the first model, the first information is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result; the transceiver module 1202 is further configured to: send the new target operation result to the second node.

[0335] In one feasible implementation, the processing module 1201 is further configured to: update the registration information of the first model based on the registration information of the second model.

[0336] In a second example, the communication device can function as a second device or a chip within a second device, and execute the steps performed by the receiving end in embodiments one through six of the above method. The transceiver module 1202 supports communication between the first node and the second node, etc. The processing module 1201 can be used to support actions performed by the second node in the above method embodiments, other than sending and receiving.

[0337] Specifically, the transceiver module 1202 is used to: send a first request instruction to the first node, the first request instruction being used to request a target operation result, the first request instruction including description information of the target operation result, the description information of the target operation result being used to determine the first model and the first operation type, the first node being a network-side device, and the second node being an OTT device; and receive the target operation result from the first node, the target operation result being obtained by performing an operation corresponding to the first operation type on the first model.

[0338] In one feasible implementation, the processing module 1201 is used to: determine whether the target operation result meets the quality requirements of the target operation result.

[0339] If the target operation result does not meet the quality requirements of the target operation result, the transceiver module 1202 is further configured to: send a first message and a second request instruction to the second node, wherein the first message is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result; receive a new target operation result from the first node, wherein the new target operation result is obtained by performing an operation corresponding to the first operation type on the second model, and the second model has the same function as the first model;

[0340] If the target operation result meets the quality requirements of the target operation result, the transceiver module 1202 is further configured to: send second information to the second node, the second information being used to characterize that the target operation result meets the quality requirements of the target operation result.

[0341] Furthermore, a processor may include a controller, an arithmetic logic unit (ALU), and registers. For example, the controller is primarily responsible for instruction decoding and issuing control signals for the operations corresponding to the instructions. The ALU is primarily responsible for performing fixed-point or floating-point arithmetic operations, shift operations, and logical operations, and can also perform address operations and translations. Registers are primarily responsible for storing register operands and intermediate operation results temporarily stored during instruction execution. In specific implementations, the processor's hardware architecture can be an ASIC architecture, a microprocessor without interlocked piped stages architecture (MIPS), an advanced reduced instruction set machine (RISC) machine (ARM) architecture, or a network processor (NP) architecture, etc. The processor can be single-core or multi-core.

[0342] The storage module can be an internal storage module of the chip, such as a register or cache. Alternatively, the storage module can be an external storage module, such as ROM or other types of static storage devices that can store static information and instructions, such as RAM.

[0343] It should be noted that the functions of the processor and interface can be implemented through hardware design, software design, or a combination of both; no restrictions are imposed here.

[0344] Furthermore, it should be noted that the aforementioned transceiver unit and / or processing unit can be implemented through virtual modules. For example, the processing unit can be implemented through software functional units or virtual devices, and the transceiver unit can be implemented through software functions or virtual devices. Alternatively, the processing unit or transceiver unit can also be implemented through physical devices. For example, if the device is implemented using a chip / chip circuit, the transceiver unit can be an input / output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operation) and output operations (corresponding to the aforementioned sending operation); the processing unit is an integrated processor, microprocessor, or integrated circuit.

[0345] It is understandable that the aforementioned communication device can also serve as a chip in the third, fourth, and fifth devices corresponding to the third, fourth, and fifth nodes, respectively, which will not be elaborated here.

[0346] Figure 13 This is a schematic diagram of another communication device provided in an embodiment of this application. Figure 13As shown, the communication device may include one or more components: a processor 1301, a memory 1302, a communication interface 1303, and a computer-readable medium 1304. The processor 1301, memory 1302, communication interface 1303, and computer-readable medium 1304 are interconnected via a bus and perform communication with each other. The memory 1302 may store one or more computer programs, which may be configured to implement the methods described in the above embodiments when executed by one or more processors 1301. The communication interface 1303 is used for communication via an internal bus or an external transmission medium.

[0347] Processor 1301 may include one or more processing cores. Processor 1301 connects to various parts within the communication device using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 1302, and by calling data stored in memory 1302. Optionally, processor 1301 may be implemented using at least one hardware form selected from microprocessors, microcontrollers, digital signal processing (DSP), field-programmable gate arrays (FPGAs), and programmable logic arrays (PLAs). Processor 1301 may integrate one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. It is understood that the aforementioned modem may also not be integrated into processor 1301, but may be implemented separately through a communication chip.

[0348] The memory 1302 may include random access memory (RAM) or read-only memory (ROM). The memory 1302 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 1302 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), and instructions for implementing the various method embodiments described above. The data storage area may also store data created during the use of the communication device.

[0349] Optionally, the processor 1301, memory 1302, and computer-readable medium 1304 may perform functions such as: encoding, decoding, rate matching, rate dematching, scrambling, descrambling, modulation, demodulation, layer mapping, fast Fourier transform, inverse fast Fourier transform, inverse discrete Fourier transform, precoding, resource element mapping, channel equalization, deresource element mapping, digital beamforming, cyclic prefix addition, cyclic prefix removal, AI inference, etc.

[0350] It is understood that the communication device may include more or fewer structural elements than those shown in the block diagram above.

[0351] This application provides a communication system, which includes a first device corresponding to a first node and a second device corresponding to a second node.

[0352] This application provides a computer-readable storage medium storing computer instructions that, when executed, cause the computer to perform any of the methods described above.

[0353] This application provides a computer program product, which includes computer program code. When the computer program code is run, it causes the computer to perform any of the methods described above.

[0354] This application provides a chip coupled to a memory for reading and executing program instructions in the memory, causing the device in which the chip resides to implement any of the methods described above. The chip may include a communication and processing circuit, which may include one or more hardware components providing a physical structure that performs various processes related to wireless communication (e.g., signal reception and / or signal transmission). The communication and processing circuit may include two or more transmit / receive chains. The functions implemented by the communication and processing circuit can also be processed on a computer-readable medium.

[0355] Please see Figure 14 , Figure 14 This is a schematic diagram of the processing logic of a processing circuit provided in an embodiment of this application. The processing circuit can be used to process information on a target operation result request (first request instruction) to obtain a judgment on whether to approve or reject the target operation result request. If the model operation result request is approved, the processing circuit can further output model information, data information, and operation type. The processing circuit can further process the above information to obtain the target operation result, and finally process the target operation result to give a target operation result response.

[0356] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not described in detail in a particular embodiment can be found in the relevant descriptions of other embodiments. It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0357] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above 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 devices or units may be electrical or other forms.

[0358] The units described above 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 according to actual needs.

[0359] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for obtaining model operation results, characterized in that, The method is applied to the first node, and the method includes: The system receives a first request instruction from a second node. The first request instruction is used to request a target operation result. The first request instruction includes a description of the target operation result. The description of the target operation result is used to determine a first model and a first operation type. The first node is a network-side device, and the second node is a cross-platform OTT device. The target operation result is obtained by performing the operation corresponding to the first operation type on the first model; The target operation result is sent to the second node.

2. The method according to claim 1, characterized in that, The first node stores model registration information, and obtaining the target operation result includes: The first model and the first operation type are determined based on the description information of the target operation result and the model registration information; Determine the operation data of the first model based on the first model; Based on the operation data of the first model, perform the operation corresponding to the first operation type on the first model to obtain the target operation result.

3. The method according to claim 2, characterized in that, The description information of the target operation result includes parameters and parameter quality requirements. The first model operation data is data to be inferred. Determining the first model and the first operation type based on the description information of the target operation result and the model registration information includes: Based on the parameters, parameter quality requirements, and model registration information, the first model and the first operation type are determined, wherein the first operation type is model inference. The step of performing the operation corresponding to the first operation type on the first model based on the first model operation data includes: The data to be inferred is input into the first model for processing to obtain the inference result, which is the target operation result.

4. The method according to claim 2, characterized in that, The descriptive information of the target operation result includes model indication information and model quality requirements. The first model operation data is training sample data. Determining the first model and the first operation type based on the descriptive information of the target operation result and the model registration information includes: The first model and the first operation type are determined based on the model indication information, model quality requirements, and model registration information, wherein the first operation type is model inference. The step of performing the operation corresponding to the first operation type on the first model based on the first model operation data includes: The first model is trained based on the training sample data to obtain the trained first model, which is the result of the target operation.

5. The method according to claim 1, characterized in that, Obtaining the target operation result includes: Send the first request instruction to the third node; The system receives the target operation result from the third node. The target operation result is determined by performing an operation corresponding to the first operation type on the first model based on the first model operation data. The first model operation data is determined based on the first model. The first model and the first operation type are determined based on the description information and model registration information of the target operation result.

6. The method according to claim 1, characterized in that, The first node stores model registration information, and obtaining the target operation result includes: The identifier of the first model and the type of the first operation are determined based on the description information of the target operation result and the model registration information; Determine the descriptive information of the first model operation data based on the first model; Send a model operation instruction to the third node. The model operation instruction is used to instruct the first model to perform an operation corresponding to the first operation type based on the first model operation data. The model operation instruction carries the identifier of the first model. Send a first data transmission instruction to the fourth node. The first data transmission instruction is used to instruct the first model operation data to be sent to the third node. The first data transmission instruction carries description information of the first model operation data. Receive the target operation result from the third node.

7. The method according to claim 1, characterized in that, The first node stores model registration information, and obtaining the target operation result includes: The identifier of the first model and the type of the first operation are determined based on the description information of the target operation result and the model registration information; Determine the descriptive information of the first model operation data based on the first model; Send a first model operation transmission instruction, a first model transmission instruction, and a second data transmission instruction to the third node. The first model operation transmission instruction is used to instruct the model operation instruction to be sent to the fourth node. The model operation instruction is used to instruct the first model to perform the operation corresponding to the first operation type based on the first model operation data. The first model transmission instruction is used to instruct the first model to be sent to the fourth node. The second data transmission instruction is used to instruct the first model operation data to be sent to the fourth node. The first model transmission instruction carries the identifier of the first model. The second data transmission instruction carries the description information of the first model operation data. Receive the target operation result from the fourth node.

8. The method according to any one of claims 1-7, characterized in that, Before obtaining the target operation result, the method further includes: The permission to obtain the target operation result is determined based on the description information of the target operation result; When it is determined that the permission judgment for obtaining the target operation result has passed, the operation of obtaining the target operation result is performed.

9. The method according to any one of claims 1-8, characterized in that, Before sending the target operation result to the second node, the method further includes: The target operation result is desensitized.

10. The method according to any one of claims 1-9, characterized in that, The description information of the target operation result includes the quality requirements of the target operation result. After sending the target operation result to the second node, the method further includes: Receive feedback information from the second node; If the feedback information includes the first information and the second request instruction, then a new target operation result is obtained. The new target operation result is obtained by performing the operation corresponding to the first operation type on the second model. The second model has the same function as the first model. The first information is used to characterize that the target operation result does not meet the quality requirements of the target operation result. The second request instruction is used to request a new target operation result. The new target operation result is sent to the second node.

11. The method according to claim 7, characterized in that, If the second model is a model obtained by adjusting the parameters of the first model, the method further includes: The registration information of the first model is updated based on the registration information of the second model.

12. A method for obtaining model operation results, characterized in that, The method is applied to the second node, and the method includes: A first request instruction is sent to the first node. The first request instruction is used to request the target operation result. The first request instruction includes description information of the target operation result. The description information of the target operation result is used to determine the first model and the first operation type. The first node is a network-side device, and the second node is an OTT device. The target operation result is received from the first node, wherein the target operation result is obtained by performing an operation corresponding to the first operation type on the first model.

13. The method according to claim 9, characterized in that, The description information of the target operation result includes the quality requirements of the target operation result. After receiving the target operation result from the first node, the method further includes: If it is determined that the target operation result does not meet the quality requirements of the target operation result, then a first message and a second request instruction are sent to the second node. The first message is used to characterize that the target operation result does not meet the quality requirements of the target operation result, and the second request instruction is used to request a new target operation result. Receive a new target operation result from the first node. The new target operation result is obtained by performing the operation corresponding to the first operation type on the second model. The second model has the same function as the first model. If it is determined that the target operation result meets the quality requirements of the target operation result, then a second message is sent to the second node, the second message being used to characterize that the target operation result meets the quality requirements of the target operation result.

14. A communication device, characterized in that, Includes units or modules for implementing the method as described in any one of claims 1 to 13.

15. A communication device, the device comprising a processor, a memory, and a computer program stored in the memory, characterized in that, The processor is configured to retrieve the executable program code computer program stored in the memory to perform the method as described in any one of claims 1-13.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-13.

17. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-13.

18. A chip, characterized in that, The chip is coupled to a memory for reading and executing program instructions stored in the memory to implement the method as described in any one of claims 1-11, or to implement the method as described in any one of claims 12-13.

Images