Communication method and related apparatus

Abstract

A communication method and a related apparatus, relating to the technical field of communications. The method comprises: sending a first message to a network device, wherein the first message is used for requesting the network device to deploy a first model, and the first message comprises identification information of the first model and resource information of a terminal device; and receiving a second message from the network device, wherein the second message comprises first instruction information or second instruction information, the first instruction information instructs the network device to independently execute a reasoning task of the first model, and the second instruction information instructs the terminal device and the network device to jointly execute the reasoning task, and instructs the terminal device and the network device to respectively deploy a first sub-model and a second sub-model in the first model. The method can reduce a transmission delay in the process of obtaining a model reasoning result by a terminal.

Description

Communication methods and related devices

[0001] This application claims priority to Chinese Patent Application No. 202410704253.0, filed with the China National Intellectual Property Administration on May 31, 2024, entitled "Communication Method and Related Apparatus", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of communication technology, and in particular to a communication method and related apparatus. Background Technology

[0003] With the development of artificial intelligence (AI) technology, AI models are gradually evolving from traditional small-scale neural network models to large-scale neural network models based on Transformers. At the same time, terminal applications based on large models are constantly emerging, such as question answering systems, text-to-image processing, text-to-video processing, and various natural language processing and multimodal tasks.

[0004] Large models, with their massive number of parameters and complex computational structures, place higher demands on the computing and storage capabilities of devices compared to traditional models. Large models are typically stored and deployed on application servers. When users need to use terminal applications based on these models, the user terminal uploads data to the application server via the network, and the application server performs the computation and returns the results to the user. During this process, the user uses the terminal application, the model inference task is executed, and data and results are transmitted between the user terminal, the network, and the application server, resulting in high transmission latency. Summary of the Invention

[0005] This application provides a communication method and related apparatus that can reduce the transmission latency of the terminal in obtaining model inference results.

[0006] Firstly, embodiments of this application provide a communication method applied to a terminal device. It is understood that this method can be executed by a communication device, which can be the terminal device itself, or a chip (system) or circuit used in the terminal device; this application does not limit this. The method includes:

[0007] Send a first message to the network device, the first message being used to request the network device to deploy a first model, the first message including the identification information of the first model and the resource information of the terminal device;

[0008] The system receives a second message from a network device, the second message including a first instruction or a second instruction, the first instruction indicating that the network device independently performs the inference task of the first model, the second instruction indicating that the terminal device and the network device jointly perform the inference task, and instructing the terminal device and the network device to respectively deploy the first sub-model and the second sub-model in the first model.

[0009] In this application, the first model may include an artificial intelligence model or a machine learning model. The identification information of the first model can be understood as information indicating and identifying the first model, specifically including at least one of the following: identification information of the first model, information indicating the accessibility of the first model, etc. The resource information of the terminal device may include storage resource information and computing resource information of the terminal device, specifically including the available memory of the terminal device chip, the computing power of the terminal device chip, etc.

[0010] Based on information such as the parameter scale and model structure of the first model, as well as the storage and computing resources of the terminal device, the network device can determine whether the first model needs to be split. If splitting is required, the first sub-model and the second sub-model obtained from the split are determined, and the first instruction information or the second instruction information is generated.

[0011] In this application, the terminal device receives first or second instruction information from the network device. Since this instruction information is obtained by the network device based on the first model and the terminal device's resource information, it can instruct the network device to independently execute the inference task or the terminal device and the network device to jointly execute the inference task. Therefore, the inference result obtained by the terminal device based on this instruction information does not originate from the service device where the first model resides. The data required for inference also does not need to be transmitted to the service device, thus reducing transmission latency during the process of the terminal device obtaining the inference result using the first model. When performing multiple inferences, transmission latency can be significantly reduced. This communication method can support the distributed deployment of large models, enabling the transfer and deployment of large models to network devices or between network devices and terminal devices. It supports various inference methods, such as independent inference by network devices and independent inference by network devices and terminal devices, expanding the model scale that the terminal can support.

[0012] In one possible implementation, before sending the first message to the network device, the method further includes:

[0013] Based on the resource information of the first model and the terminal device, determine whether the terminal device can independently execute the inference task;

[0014] Sending the first message to the network device includes:

[0015] If the terminal device cannot perform the inference task independently, the first message is sent to the network device.

[0016] In this embodiment, the terminal device determines whether it can independently execute the inference task of the first model based on information such as the parameter scale of the first model, as well as the storage and computing resource information of the terminal device. If it is determined that the inference task cannot be executed independently, the terminal device can request the network device to deploy the first model and determine whether to split a part of the model for deployment in the terminal device. This allows the network side to execute the inference task independently or the network side and the terminal side to execute the inference task collaboratively, without having to obtain the inference result from the service device, thus reducing transmission latency.

[0017] In one possible implementation, before determining whether the terminal device can independently execute the inference task based on the first model and the resource information of the terminal device, the method further includes:

[0018] A third message is received, the third message including the first model and the identification information of the first model.

[0019] In scenarios where the transfer and deployment of the first model is triggered by a terminal device, the third message can be sent by the service device where the first model resides. After receiving the third message, the terminal device can determine whether it can independently complete the inference task based on the first model and its own resource information. This allows the first model to be partially or completely deployed to the network device even when the terminal device's own resources are insufficient, reducing the transmission latency for obtaining subsequent inference results. When the first model is deployed in both the network device and the terminal device, the terminal device's resources can also be utilized, improving the terminal device's resource utilization and further reducing transmission latency.

[0020] In one possible implementation, before sending the first message to the network device, the method further includes:

[0021] Receive a fourth message, the fourth message including the identification information of the first network element performing the inference task in the network device and the identification information of the first model;

[0022] Sending the first message to the network device includes:

[0023] In response to the fourth message, the first message is sent to the network device, where the network device is the first network element.

[0024] The first network element can be a MIF in the RAN domain, an NWDAF in the core network, or an EMS or NMS. The identification information of the first network element can include at least one of the following: the IP address of the first network element; the subnet identifier where the first network element is located and the unique identifier within the subnet corresponding to the first network element; and the ID information that can uniquely indicate the first network element in the entire network.

[0025] In this embodiment, after receiving the fourth message, the terminal device can send its own resource information to the network device, so that the network device can generate a first instruction information or a second instruction information based on the terminal device's resource information and the first model, and clarify whether the network device or the network device and the terminal device will perform the inference task of the first model. This enables the first model to be deployed in the network device in whole or in part, so that the terminal device no longer needs to obtain the inference result from the service device, thereby reducing transmission latency and improving service efficiency.

[0026] In one possible implementation, the method further includes:

[0027] A fifth message is sent to the service device where the first model is located, the fifth message including the first indication information or the second indication information.

[0028] In this embodiment of the application, after receiving the second information from the network device, the terminal device sends a fifth message to the service device. The message can provide feedback to the service device that the first model has been transferred and deployed in the network device, and provide feedback on the subject that will subsequently perform the inference task and the first network element that will actually perform the inference task. This allows the service device to obtain relevant information and avoids having to perform the inference task of the first model again. As a result, it does not need to obtain the inference result from the service device, thus reducing transmission latency.

[0029] In one possible implementation, the second message may also include identification information of the first network element in the network device that performs the inference task.

[0030] In one possible implementation, the method further includes:

[0031] If the second message includes the first indication information, a sixth message is sent to the network device, the sixth message being used to request the network device to independently execute the inference task, the sixth message including first data required to execute the inference task;

[0032] Receive the first inference result, which is obtained based on the first data.

[0033] In this embodiment, the terminal device sends first data to the network device, enabling the network device to perform inference based on the first data and the first model, obtain a first inference result, and feed it back to the terminal device. This process is completed independently by the network device, without needing to send data to the service device or obtain the inference result from the service device, thus reducing the transmission latency of the entire inference process.

[0034] In one possible implementation, the method further includes:

[0035] If the second message includes the second indication information, a seventh message is sent to the network device. The seventh message is used to request the terminal device to jointly execute the inference task with the network device. The seventh message includes second data. The second data is obtained by the terminal device inferring the first data based on the first sub-model, or the second data is obtained by preprocessing the first data. The first data is the data required to execute the inference task.

[0036] In this embodiment of the application, the terminal device sends the second data to the network device, and the terminal device and the network device jointly perform the inference task to obtain the inference result. There is no need to send data to the service device and obtain the inference result from the service device, which reduces the transmission latency of the entire inference process.

[0037] In one possible implementation, the method further includes:

[0038] When the second indication information indicates that the network device obtains an intermediate layer feature based on the second sub-model, the third data is obtained by reasoning the second data based on the first sub-model.

[0039] Receive fourth data from the network device, the fourth data being obtained by the network device from reasoning about the second data based on the second sub-model;

[0040] The second reasoning result is obtained based on the third and fourth data.

[0041] In this embodiment, the terminal device and the network device determine the specific process for jointly executing the inference task based on the second instruction information. The terminal device and the network device respectively perform partial inference using the first sub-model and the second sub-model to obtain the final inference result. This eliminates the need to send data to the service device and obtain inference results from the service device, reducing the transmission latency of the entire inference process. Since the terminal device participates in inference, its resources can be utilized, improving resource utilization. The independent inference performed by the terminal device and the network device reduces the inference time required and improves inference efficiency.

[0042] In one possible implementation, the first model includes an artificial intelligence model or a machine learning model.

[0043] In the embodiments of this application, distributed deployment of large models can be realized, supporting multiple inference methods such as network-side inference and network-side and terminal-side collaborative inference, which can reduce transmission latency.

[0044] Secondly, embodiments of this application provide a communication method applied to a network device. It is understood that this method can be executed by a communication device, which can be a network device, or a chip (system) or circuit used in a network device; this application does not limit this. The method includes:

[0045] Receive a first message from the terminal device. The first message is used to request the network device to deploy a first model. The first message includes the identification information of the first model and the resource information of the terminal device.

[0046] Based on the resource information of the first model and the terminal device, a second message is generated. The second message includes a first instruction message or a second instruction message. The first instruction message indicates that the network device independently executes the inference task of the first model. The second instruction message indicates that the terminal device and the network device jointly execute the inference task. It also indicates that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model.

[0047] Send a second message to the terminal device.

[0048] In this application, the network device generates first or second instruction information based on the resource information of the first model and the terminal device. This allows the first model to be deployed on the network device or in both the network device and the terminal device. Consequently, the inference task can be executed independently by the network device or jointly by the terminal device and the network device. The data required for inference and the obtained inference results no longer need to be transmitted to the service device where the first model resides, thereby reducing transmission latency during the process of the terminal device using the first model to obtain the inference results. When multiple inference operations are performed, the transmission latency can be significantly reduced.

[0049] In one possible implementation, before receiving the first message from the terminal device, the method further includes:

[0050] A fourth message is sent to the terminal device. The fourth message is used to obtain the resource information of the terminal device. The fourth message includes the identification information of the first network element in the network device that performs the inference task and the identification information of the first model.

[0051] In this embodiment, the network device sends a fourth message to the terminal device, which enables the terminal device to send its own resource information to the network device. Based on the terminal device's resource information and the first model, the network device generates a first instruction message or a second instruction message to clarify whether the network device or the network device and the terminal device will subsequently perform the inference task of the first model. This allows the first model to be deployed entirely or partially in the network device, so that the terminal device no longer needs to obtain the inference result from the service device, thereby reducing transmission latency and improving service efficiency.

[0052] In one possible implementation, before receiving the first message from the terminal device, the method further includes: sending an eighth message to the service device where the first model is located, the eighth message including the identification information of the first network element in the network device that performs the inference task.

[0053] In this embodiment of the application, the first network element is determined in advance by the network device. The network device feeds back the identification information of the first network element to the service device, so that the service device can know the network element in the network device that handles the deployment and inference of the first model, so that the subsequent service device does not have to perform the inference task of the first model again, thereby reducing the inference transmission latency.

[0054] In one possible implementation, the method further includes: receiving a third message, the third message including the first model and the identification information of the first model.

[0055] In this embodiment, after receiving the third message, the network device can store or call the first model, obtain resource information of the terminal device, and generate first or second indication information. This enables the first model to be partially or completely deployed in the network device, reducing the transmission latency for obtaining subsequent inference results. When the first model is deployed in both the network device and the terminal device, the resources of the terminal device can also be utilized, improving the resource utilization rate of the terminal device and further reducing transmission latency.

[0056] In one possible implementation, the first message also includes the first model.

[0057] In one possible implementation, the method further includes:

[0058] Based on the identification information of the first model, determine whether the first model has been stored in the network device;

[0059] If the first model is not stored in the network device, the first model is stored.

[0060] If the first model has been stored in the network device, the first model is invoked.

[0061] Identify the first network element in the network device that performs the inference task of the first model.

[0062] In one possible implementation, the method further includes:

[0063] If the second message includes the first indication information, a sixth message is received from the terminal device, the sixth message being used to request the network device to independently execute the inference task, the sixth message including first data required to execute the inference task;

[0064] Based on the first data, the reasoning task is executed independently to obtain the first reasoning result;

[0065] The first inference result is sent to the terminal device.

[0066] In one possible implementation, the method further includes:

[0067] If the second message includes the second indication information, a seventh message is received from the terminal device. The seventh message is used to request the terminal device and the network device to jointly execute the inference task. The seventh message includes second data. The second data is obtained by the terminal device from the first data based on the first sub-model, or the second data is obtained by the terminal device from the first data after preprocessing. The first data is the data required to execute the inference task.

[0068] In one possible implementation, the method further includes:

[0069] The fourth data is obtained by reasoning from the second data based on the second sub-model;

[0070] The fourth data is sent to the terminal device.

[0071] In this embodiment, the network device determines the specific process for jointly executing the inference task based on the second instruction information, and the terminal device performs partial inference using the second sub-model and the first sub-model respectively to obtain the final inference result. This eliminates the need to send data to the service device and obtain inference results from the service device, reducing the transmission latency of the entire inference process. Since the terminal device participates in inference, its resources can be utilized, improving resource utilization. The independent inference performed by the terminal device and the network device reduces the inference time required and improves inference efficiency.

[0072] Thirdly, embodiments of this application provide a communication device that includes a unit for performing the method as described in any of the first aspects.

[0073] In one possible design, the device includes:

[0074] A communication unit is configured to send a first message to a network device, the first message being a request for the network device to deploy a first model, the first message including identification information of the first model and resource information of the terminal device;

[0075] The communication unit is further configured to receive a second message from the network device, the second message including a first instruction message or a second instruction message, the first instruction message indicating that the network device independently executes the inference task of the first model, the second instruction message indicating that the terminal device and the network device jointly execute the inference task, and instructing the terminal device and the network device to respectively deploy the first sub-model and the second sub-model in the first model.

[0076] In one possible implementation, the device further includes:

[0077] A processing unit is used to generate the first message.

[0078] Regarding the processing unit and communication unit described in the third aspect and any possible implementation, the steps performed thereon can be referred to the corresponding implementations in the first aspect.

[0079] For the technical effects of the third aspect and any possible implementation, please refer to the description of the technical effects corresponding to the first aspect and the corresponding implementation.

[0080] Fourthly, embodiments of this application provide a communication device that includes a unit for performing the method as described in any of the second aspects.

[0081] In one possible design, the device includes:

[0082] A communication unit is configured to receive a first message from a terminal device, the first message being a request for the network device to deploy a first model, the first message including identification information of the first model and resource information of the terminal device;

[0083] The communication unit is also used to send the second message to the terminal device;

[0084] The processing unit is configured to generate a second message based on the resource information of the first model and the terminal device. The second message includes either the first instruction information or the second instruction information. The first instruction information indicates that the network device independently executes the inference task of the first model, and the second instruction information indicates that the terminal device and the network device jointly execute the inference task. It also indicates that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model.

[0085] Regarding the processing unit and communication unit described in the fourth aspect and any possible implementation, the steps performed thereon can be referred to the corresponding implementation in the second aspect.

[0086] Regarding the technical effects of the fourth aspect and any possible implementation, refer to the description of the technical effects corresponding to the second aspect and the corresponding implementation.

[0087] Optionally, in the communication apparatus described in any of the third to fourth aspects and any of the possible embodiments:

[0088] In one implementation, the communication device is a communication equipment. When the communication device is a communication equipment, the communication unit can be a transceiver or an input / output interface; the processing unit can be at least one processor. Optionally, the transceiver can be a transceiver circuit. Optionally, the input / output interface can be an input / output circuit.

[0089] In another implementation, the communication device is a chip (system) or circuit used in a communication device. When the communication device is a chip (system) or circuit used in a communication device, the communication unit can be a communication interface (input / output interface), interface circuit, output circuit, input circuit, pin, or related circuit on the chip (system) or circuit; the processing unit can be at least one processor, processing circuit, or logic circuit.

[0090] Fifthly, embodiments of this application provide a communication device including a processor. The processor is coupled to a memory and can be used to execute instructions in the memory to implement the methods of any one of the first to second aspects and any possible implementations described above. Optionally, the communication device further includes a memory. Optionally, the communication device further includes a communication interface, and the processor is coupled to the communication interface.

[0091] Sixthly, embodiments of this application provide a communication device, including: a logic circuit and a communication interface. The communication interface is used to receive or send information; the logic circuit is used to receive or send information through the communication interface, causing the communication device to perform the method of any one of the first to second aspects and any possible implementation thereof.

[0092] In a seventh aspect, embodiments of this application provide a computer-readable storage medium for storing a computer program (also referred to as code or instructions); when the computer program is run on a computer, the methods described in any of the first to second aspects and any possible implementations are implemented.

[0093] Eighthly, embodiments of this application provide a computer program product, the computer program product comprising: a computer program (also referred to as code or instructions); and, when the computer program is run, causing a computer to perform the method of any one of the first to second aspects and any possible implementation thereof.

[0094] Ninthly, embodiments of this application provide a chip including a processor configured to execute instructions, which, when executed, cause the chip to perform the methods described in any one of the first to second aspects and any possible implementations thereof. Optionally, the chip further includes a communication interface configured to receive or transmit signals.

[0095] In a tenth aspect, embodiments of this application provide a communication system, the communication system including at least one communication device as described in the third aspect, or the fourth aspect, or the fifth aspect, or the sixth aspect, or the ninth aspect.

[0096] Eleventhly, embodiments of this application provide a communication system, the communication system including a terminal device and a network device, the terminal device being used to perform the methods of the first aspect and any possible implementation described above, and the network device being used to perform the methods of the second aspect and any possible implementation described above.

[0097] Furthermore, in the process of performing the methods described in any of the first to second aspects and any possible embodiments described above, the processes related to sending and / or receiving information in the above methods can be understood as the process of the processor outputting information, and / or the process of the processor receiving input information. When outputting information, the processor can output the information to a transceiver (or communication interface, or transmitting module) so that the transceiver can transmit it. After the information is output by the processor, it may need to undergo other processing before reaching the transceiver. Similarly, when the processor receives input information, the transceiver (or communication interface, or transmitting module) receives the information and inputs it to the processor. Furthermore, after the transceiver receives the information, the information may need to undergo other processing before being input to the processor.

[0098] Based on the above principles, for example, the information sent mentioned in the aforementioned method can be understood as information output by the processor. Similarly, the information received can be understood as information received by the processor from input.

[0099] Optionally, unless otherwise specified, or unless they contradict their actual function or internal logic in the relevant description, the operations of the processor, such as transmitting, sending, and receiving, can be more generally understood as processor output and receiving, input, and other operations.

[0100] Optionally, in the process of performing the methods described in any of the first to second aspects and any possible embodiments above, the processor may be a processor specifically designed to perform these methods, or it may be a processor that performs these methods by executing computer instructions stored in memory, such as a general-purpose processor. The memory may be a non-transitory memory, such as read-only memory (ROM), which may be integrated with the processor on the same chip or disposed on separate chips. This application does not limit the type of memory or the arrangement of the memory and processor.

[0101] In one possible implementation, at least one of the aforementioned memories is located outside the device.

[0102] In yet another possible implementation, at least one of the aforementioned memories is located within the device.

[0103] In another possible implementation, a portion of the memory of the at least one memory is located inside the device, while another portion is located outside the device.

[0104] In this application, the processor and memory may also be integrated into a single device, that is, the processor and memory can be integrated together.

[0105] In this embodiment, the network device can generate first instruction information or second instruction information based on the resource information of the first model and the terminal device, so that the first model is deployed on the network device or deployed in the network device and the terminal device. Accordingly, the network device can independently execute the inference task or the terminal device and the network device can jointly execute the inference task. The data required for inference and the obtained inference results do not need to be transmitted to the service device where the first model is located, thereby reducing the transmission latency during the process of the terminal device using the first model to obtain the inference results. Attached Figure Description

[0106] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0107] Figure 1 is a schematic diagram of a communication system provided in an embodiment of this application;

[0108] Figure 2 is a flowchart illustrating a communication method provided in an embodiment of this application;

[0109] Figure 3 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0110] Figure 4 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0111] Figure 5 is a flowchart illustrating another communication method provided in an embodiment of this application;

[0112] Figure 6a is a schematic diagram of the process by which a terminal device obtains inference results after the first model is transferred and deployed according to an embodiment of this application.

[0113] Figure 6b is another schematic diagram of the process by which the terminal device obtains the inference result after the first model is transferred and deployed according to the embodiment of this application;

[0114] Figure 6c is another schematic diagram of the process by which the terminal device obtains the inference result after the first model is transferred and deployed according to the embodiment of this application;

[0115] Figure 7 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0116] Figure 8 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0117] Figure 9 is a schematic diagram of the structure of a chip provided in an embodiment of this application. Detailed Implementation

[0118] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described below with reference to the accompanying drawings.

[0119] The terms "first" and "second," etc., used in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0120] The term "embodiment" as used herein means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that, unless otherwise specified or logically conflicting, the terminology and / or descriptions between the various embodiments of this application are consistent and can be mutually referenced, and technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0121] It should be understood that in this application, "at least one (item)" means one or more, "more than one" means two or more, "at least two (items)" means two or three or more, and "and / or" is used to describe the relationship between related objects, indicating that there can be three relationships. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the related objects before and after are in an "or" relationship. "At least one (item) 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 (item) of a, b, or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0122] It should be noted that, in this application, "instruction" can include direct instruction, indirect instruction, explicit instruction, and implicit instruction. When describing a certain instruction information for the purpose of instructing A, it can be understood that the instruction information carries A, directly instructs A, or indirectly instructs A.

[0123] In this application, the information indicated by the instruction information is called the information to be instructed. In specific implementations, there are many ways to indicate the information to be instructed, such as, but not limited to, directly indicating the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly indicate the information to be instructed by indicating other information, where there is a correlation between the other information and the information to be instructed. It can also indicate only a part of the information to be instructed, while the other parts are known or pre-agreed upon. For example, the instruction of specific information can be achieved by using a pre-agreed (e.g., protocol-defined) arrangement of various information, thereby reducing instruction overhead to some extent. The information to be instructed can be sent as a whole or divided into multiple sub-information units, and the sending period and / or timing of these sub-information units can be the same or different. This application does not limit the specific sending method. The sending period and / or timing of these sub-information units can be predefined, for example, according to a protocol, or configured by the transmitting device by sending configuration information to the receiving device.

[0124] It should be noted that in this application, "send" can be understood as "output" and "receive" can be understood as "input". "Send information to A", where "to A" simply indicates the direction of information transmission, and A is the destination, does not limit "send information to A" to a direct transmission over the air interface. "Send information to A" includes sending information directly to A, as well as sending information indirectly to A through a transmitter. Therefore, "send information to A" can also be understood as "outputting information destined for A". Similarly, "receive information from A" indicates that the source of the information is A, including receiving information directly from A, as well as receiving information indirectly from A through a receiver. Therefore, "receive information from A" can also be understood as "inputting information from A".

[0125] This application provides a communication method and related apparatus that can reduce the transmission latency of a terminal device in obtaining model inference results.

[0126] The following describes a communication system provided in an embodiment of this application. As shown in Figure 1, the communication system may include a terminal device, a network device, and a service device. The network device can connect the terminal device and the service device.

[0127] Terminal equipment, also known as user equipment (UE) or a terminal, is a device with wireless transceiver capabilities. It can be deployed on land (indoors or outdoors, handheld, wearable, or vehicle-mounted), on water (e.g., on ships), or in the air (e.g., on airplanes, balloons, or satellites). Terminal equipment can be mobile phones, tablets, computers with wireless transceiver capabilities, virtual reality (VR) terminals, augmented reality (AR) terminals, wireless terminals in industrial control, self-driving, remote medical care, smart grids, transportation safety, smart cities, smart homes, and so on. It can also be understood that this terminal equipment could be a terminal device in future 6G networks or a terminal device in future evolved PLMNs.

[0128] Network devices can collect, analyze, and predict data from terminal and service devices. These devices may include management network elements, such as the network data analytics function (NWDAF), element management system (EMS), network management system (NMS) in the core network, mobile intelligent function (MIF) in the access network, or other units with big data analytics or artificial intelligence processing capabilities. NWDAF, for example, possesses various intelligent computing functions such as AI training and inference. EMS manages one or more network elements of a specific category and can also be called a domain management system or single-domain management system. NMS is responsible for network operation, management, and maintenance and can also be called a cross-domain management system. MIF generally refers to the network function responsible for intelligent computing in a wireless network. Furthermore, these network devices may also include access network devices.

[0129] Access network equipment can be a next-generation node B (gNB), a next-generation evolved node B (ng-eNB), or access network equipment in future 6G communications. Access network equipment can be any device with wireless transceiver capabilities, including but not limited to the base station (BS) mentioned above. The base station can also be a base station in future communication systems such as sixth-generation communication systems. Optionally, the access network equipment can be an access node, wireless relay node, or wireless backhaul node in a wireless local area network (WiFi) system. Optionally, the access network equipment can be a wireless controller in a cloud radio access network (CRAN) scenario. Optionally, the access network equipment can be a wearable device or vehicle-mounted device. Optionally, the access network equipment can also be a small cell, a transmission reception point (TRP) (or transmission point), etc. It is understood that the access network equipment can also be a base station in a future evolved public land mobile network (PLMN), etc.

[0130] In some deployments, base stations (such as gNBs) can consist of centralized units (CUs) and distributed units (DUs). This means the functions of the base station in the access network are split, with some functions deployed in a CU and the remaining functions in a DU. Multiple DUs sharing a single CU can save costs and facilitate network expansion. In other base station deployments, the CU can be divided into a CU-control plane (CP) and a CU-user plane (UP). In still other deployments, the base station can be a radio unit (RU). In yet another deployment, the base station can be an open radio access network (ORAN) architecture, etc. This application does not limit the specific type of base station. For example, when the base station is an ORAN architecture, the base station shown in the embodiments of this application can be an access network device in the ORAN, or a module within the access network device, etc. In the ORAN system, CU can also be called open (O)-CU, DU can also be called O-DU, CU-DU can also be called O-CU-DU, CU-UP can also be called O-CU-UP, and RU can also be called O-RU.

[0131] Service equipment refers to third-party network application servers other than network equipment and terminal equipment.

[0132] It should be understood that Figure 1 exemplarily illustrates a core network device representing a management network element, a base station, a terminal device, and a service device, as well as the communication links between the various communication devices. Optionally, the communication system may include multiple base stations, and the coverage area of ​​each base station may include other numbers of terminal devices, such as more or fewer terminal devices, etc., which is not limited in this application.

[0133] The aforementioned communication devices, such as the network device, terminal device, and service device in Figure 1, can be configured with multiple antennas. These multiple antennas may include at least one transmitting antenna for transmitting signals and at least one receiving antenna for receiving signals, etc. This application embodiment does not limit the specific structure of each communication device. Optionally, the communication system may also include other network entities such as a network controller and a mobility management entity; this application embodiment is not limited to these.

[0134] It is understood that the communication system diagram shown in Figure 1 is only an example. For other forms of communication system diagrams, please refer to the relevant standards or protocols, etc., which will not be described in detail here.

[0135] The various embodiments shown below can be applied to the communication system shown in Figure 1, or to other forms of communication systems, which will not be described further below.

[0136] This application provides a communication method applicable to the field of communication technology. To more clearly describe the solution of this application, some knowledge related to machine learning models (or artificial intelligence models) will be introduced below.

[0137] The model mentioned in this application, namely the machine learning model or artificial intelligence model, can be considered as an algorithm that enables computers to "learn" automatically. The model includes two core stages: training and inference.

[0138] During the training phase, the model learns to recognize and generate patterns through a large amount of data and algorithms. Large models, through deep learning techniques and multi-layered neural networks, learn and optimize from the massive amounts of input data, and adjust the model's parameters to enable it to make accurate predictions on the input data.

[0139] The inference phase builds upon the completed training, where the trained model is used to predict or classify new, unseen data. Large models can handle various types of input and output corresponding predictions during the inference phase. Inference can be performed in production environments, such as classifying images, speech, or text in real-world applications, or for other tasks like language generation and translation.

[0140] Using models to solve real-world problems involves two main parts: model deployment and model inference. Model deployment refers to running the trained model in a specific environment. After deployment, the input data is used to perform inference using the model, yielding inference results that can then be applied to the real-world scenario.

[0141] Please refer to Figure 2, which is a flowchart illustrating a communication method provided in an embodiment of this application. This communication method is applied in the field of communication technology. It is understood that this communication method can be executed by a communication device, which can be a network device, terminal device, or service device, or a chip (system) or circuit used in these devices; this application does not limit its scope. The communication method includes, but is not limited to, the following steps:

[0142] S201: The terminal device sends a first message to the network device, and the network device receives the first message accordingly.

[0143] S202: The network device generates a second message based on the first model and the resource information of the terminal device.

[0144] S203: The network device sends a second message to the terminal device, and the terminal device receives the second message accordingly.

[0145] In this embodiment, the first message is used to request the network device to deploy a first model. The first message includes the identification information of the first model and the resource information of the terminal device. The second message includes either a first instruction or a second instruction. The first instruction indicates that the network device independently performs the inference task of the first model, and the second instruction indicates that the terminal device and the network device jointly perform the inference task of the first model, and instructs the terminal device and the network device to deploy the first sub-model and the second sub-model of the first model, respectively.

[0146] The first model may include an artificial intelligence model or a machine learning model.

[0147] The identification information of the first model can be understood as information indicating and identifying the first model, specifically including at least one of the following: the identification information of the first model, information indicating the accessibility of the first model, etc. The accessibility of the first model may include the address information of the first model, etc. For example, when the identification information of the first model includes the identification information of the first model, if the first model corresponds to a service in an APP application, then the identification information of the first model may include the application identifier of the APP and the service type of the service. This application embodiment does not limit this.

[0148] The resource information of a terminal device may include its storage and computing resources, specifically including the available memory and computing power of its chip. For example, the resource information may include the available memory, power consumption, and computing frequency of the terminal device's chip.

[0149] The first sub-model in the first model may include one or more models, and the second sub-model in the first model may also include one or more models. The first sub-model and the second sub-model are different. Specifically, the first sub-model and the second sub-model can be respectively a shallow model and a deep model in the first module. The shallow model in the first model is closer to the input layer of the first model than the deep model. Alternatively, the first sub-model and the second sub-model can be two independent sub-models that can be deployed in different locations within the first model.

[0150] Understandably, network devices can determine whether the first model needs to be split based on information such as the parameter scale and model structure of the first model, as well as resource information such as the storage and computing resources of the terminal devices. If splitting is necessary, the network devices determine the parts of the model to be deployed on the terminal devices and network devices respectively, based on the difficulty of splitting, and generate first or second instruction information. This first and second instruction information can be referred to as collaborative reasoning instructions.

[0151] When the second instruction information instructs the terminal device to deploy the first sub-model, the terminal device's storage and computing resources are sufficient to support inference of the first sub-model within the terminal device.

[0152] Optionally, the second instruction information may further instruct the specific rules for the joint execution of the inference task by the terminal device and the network device, such as the input of data, the processing method of the features obtained based on the first sub-model, the processing method of the features obtained based on the second sub-model, and the method of obtaining the final inference result. This application embodiment does not limit this.

[0153] Optionally, after receiving the second message, if the second message includes second instruction information, the terminal device can deploy the first sub-model in the first model. Subsequently, when it is necessary to execute the inference task of the first model, it can perform partial inference based on the first sub-model and the second instruction information, and jointly execute the inference task of the first model with the network device to obtain the inference result.

[0154] Understandably, when the first model is deployed in the service device, before the terminal device obtains the inference result of the first model, it needs to send the data to the service device via the network device. The service device performs the inference task, obtains the inference result, and transmits the inference result to the terminal device via the network device. This process requires information to be forwarded through multiple nodes, resulting in high transmission latency.

[0155] In this embodiment, the network device can generate first or second instruction information based on the resource information of the first model and the terminal device. This allows the first model to be deployed on the network device or on both the network device and the terminal device. Consequently, the inference task can be executed independently by the network device or jointly by the terminal device and the network device. The data required for inference and the obtained inference results no longer need to be transmitted to the service device where the first model resides, thereby reducing transmission latency during the process of the terminal device using the first model to obtain the inference results. When multiple inference operations are performed, the transmission latency can be significantly reduced.

[0156] In one possible embodiment, before performing step S201 described above, the following steps may also be performed:

[0157] S204: The terminal device establishes a connection with the service device where the first model is located through the network device.

[0158] S205: The terminal device registers, and the service device authenticates the terminal device based on the application corresponding to the first model, and the terminal device authorizes user data.

[0159] The first model is stored and deployed in the aforementioned service device, which can be an application server. The service device can be used to implement business logic generation and management functions for specific applications or services. The trained model, integrated into an application developed to solve business problems, can be deployed in the service device. The service device can provide application services to terminals by executing the inference tasks of the first model. Optionally, the first model can correspond to an application within the service device, or it can correspond to a specific service within an application within the service device.

[0160] Optionally, when the application corresponding to the first model in the service device can provide personalized services, the first model is a model fine-tuned from the basic model using user data authorized by the terminal device. Otherwise, the first model is the basic model provided by the service device.

[0161] Understandably, after the terminal device registers, the service device can send the prepared first model to the network device or the terminal device, which facilitates the subsequent deployment of the first model in the network device or in both the network device and the terminal device, thereby reducing the transmission latency of the terminal device in obtaining the inference results of the first model.

[0162] In this embodiment, the service device distributes the first model, and the terminal device sends a first message to the network device. This allows the network device to generate either a first instruction or a second instruction based on the first message and send the instruction to the terminal device. When the instruction indicates that the network device should independently execute the inference task, the first model can be deployed within the network device, and the inference result can be obtained without going through the service device, reducing transmission latency. When the instruction indicates that the terminal device and the network device should jointly execute the inference task, the terminal device can deploy a first sub-model from the first model and perform partial inference based on the instruction and the first sub-model when the inference task needs to be executed, jointly obtaining the inference result with the network device, thus deploying the first model in both the terminal device and the network device. This process reduces transmission latency and utilizes the resources of the terminal device, improving resource utilization.

[0163] Please refer to Figure 3, which is a flowchart illustrating another communication method provided in an embodiment of this application. It is understood that the steps in this embodiment can be considered reasonable variations or supplements to the embodiment in Figure 2 above; or, it is understood that the communication method in this embodiment can also be considered an embodiment that can be executed independently, and this application does not impose any limitations on it. In this communication method, the transfer deployment of the first model is triggered by the terminal device.

[0164] This communication method includes, but is not limited to, the following steps:

[0165] S301: The terminal device sends a first message to the network device, and the network device receives the first message accordingly.

[0166] S302: The network device generates a second message based on the first model and the resource information of the terminal device.

[0167] S303: The network device sends a second message to the terminal device, and the terminal device receives the second message accordingly.

[0168] The steps S301 to S303 described above are the same as steps S201 to S203 in the embodiment shown in Figure 2 above, and will not be repeated here.

[0169] In one possible embodiment, after performing step S301 as described above, the present application embodiment may further perform the following steps:

[0170] S304: The network device determines whether the first model has been stored in the network device based on the identification information of the first model; if the first model has not been stored in the network device, the first model is stored; if the first model has been stored in the network device, the first model is invoked.

[0171] The network device determines the first network element in the network device that performs the inference task of the first model.

[0172] It is understandable that the aforementioned first network element can be a MIF in the RAN domain, an NWDAF in the core network, or an EMS or NMS, etc., and the resources of the first network element are sufficient to perform the inference task of the first network element.

[0173] It is understood that the second message in step S303 above may also include the identification information of the first network element.

[0174] The network device can determine whether the first model has been stored in its storage unit, store or call the first model, and allocate the first network element as a service network element so that it can subsequently use the first network element to receive inference requests from terminal devices and execute the inference task of the first model.

[0175] In one possible embodiment, before performing step S301 described above, the present application embodiment may further perform the following steps:

[0176] S305: The terminal device determines whether it can independently perform the inference task based on the first model and the terminal device's resource information.

[0177] It is understandable that step S301 above can specifically involve the terminal device sending a first message to the network device when the terminal device cannot independently execute the inference task of the first model. Correspondingly, the network device receives the first message.

[0178] In this process, the terminal device determines whether it can independently execute the inference task of the first model based on information such as the parameter scale of the first model, as well as the storage and computing resource information of the terminal device. For example, based on the parameter scale of the first model, the terminal device determines whether the video memory requirement for executing the inference task is greater than the available memory of the terminal device. If it is greater, the terminal device determines that it cannot independently execute the inference task. As another example, the terminal device can estimate whether the energy consumption required to execute the inference task and obtain the inference result exceeds the maximum load of the terminal device based on the chip's computing frequency and power. If it does, the terminal device determines that it cannot independently execute the inference task. As yet another example, the terminal device estimates whether the chip's computing power can meet the service latency requirements when executing the inference task. If it cannot, the terminal device determines that it cannot independently execute the inference task. It is understood that the specific judgment process for determining whether the terminal device can independently execute the inference task can be determined according to actual needs. The above judgment process is only a partial example and does not constitute a limitation on the embodiments of this application.

[0179] In one possible embodiment, before performing step S305 as described above, the following steps may also be performed:

[0180] S306: The service device sends a third message, and the terminal device receives the third message accordingly.

[0181] The third message includes the first model and its identification information.

[0182] The relevant description of the recognition information of the first model can be found in the above description, and will not be repeated here.

[0183] It is understandable that the first message in step S301 above may also include the first model.

[0184] It is understood that in this embodiment, the transfer and deployment of the first model is triggered by the terminal device. Therefore, after the terminal registers, the service device can send the first model and its identification information to the terminal device by sending a third message. After receiving the third message, the terminal device can execute the above step S305 to determine whether the terminal device can independently execute the inference task of the first model based on the resource information of the first model and the terminal device, so as to determine how to obtain the inference result of the first model.

[0185] In this embodiment, the service device proactively sends the first model and identification information to the terminal device. The terminal device can determine whether it can independently complete the inference task based on the first model and its own resource information. This allows the first model to be partially or fully deployed in the network device even when the terminal device's own resources are insufficient, reducing the transmission latency for obtaining the subsequent inference results. When the first model is deployed in both the network device and the terminal device, the terminal device's resources can be utilized, improving resource utilization and further reducing transmission latency.

[0186] In one possible embodiment, after performing step S303 as described above, the present application embodiment may further perform the following steps:

[0187] S307: The terminal device sends a fifth message to the service device where the first model is located, and the service device receives the fifth message accordingly.

[0188] The fifth message includes either the first instruction information or the second instruction information mentioned above.

[0189] It is understood that, in the embodiments of this application, when the second information includes the first indication information, the fifth information includes the first indication information. When the second information includes the second indication information, the fifth information includes the second indication information.

[0190] The fifth piece of information may further include the identification information of the first network element. This identification information may include at least one of the following: the IP address of the first network element; the subnet identifier of the first network element and a unique identifier within the subnet corresponding to the first network element; and ID information that uniquely identifies the first network element across the entire network. This application does not impose any limitations on this.

[0191] Through the embodiments of this application, after receiving the second information from the network device, the terminal device sends a fifth message to the service device. This message can provide feedback to the service device that the first model has been transferred and deployed in the network device, and also provide feedback on the subject that will subsequently perform the inference task and the first network element that will actually perform the inference task. This allows the service device to obtain relevant information and avoids having to perform the inference task of the first model again. As a result, it does not need to obtain the inference result from the service device, thus reducing transmission latency.

[0192] Please refer to Figure 4, which is a flowchart illustrating another communication method provided in an embodiment of this application. It is understood that the steps in this embodiment can be considered reasonable variations or supplements to the embodiment in Figure 2 above; or, it is understood that the communication method in this embodiment can also be considered an embodiment that can be executed independently, and this application does not limit this. In this communication method, the transfer deployment of the first model is triggered by a network device.

[0193] This communication method includes, but is not limited to, the following steps:

[0194] S401: The terminal device sends a first message to the network device, and the network device receives the first message accordingly.

[0195] S402: The network device generates a second message based on the first model and the resource information of the terminal device.

[0196] S403: The network device sends a second message to the terminal device, and the terminal device receives the second message accordingly.

[0197] The steps S401 to S403 described above are the same as steps S201 to S203 in the embodiment shown in Figure 2 above, and will not be repeated here.

[0198] In one possible embodiment, before performing step S401 described above, the present application embodiment may further perform the following steps:

[0199] S404: The network device sends a fourth message to the terminal device, and the terminal device receives the fourth message accordingly.

[0200] It is understandable that step S401 above can specifically be, in response to the fourth message, the terminal device sends the first message to the network device.

[0201] The fourth message is used to obtain resource information of the terminal device. This fourth message includes the identification information of the first network element performing the inference task and the identification information of the first model within the network device. For details regarding the identification information of the first network element and the identification information of the first model, please refer to the description above; they will not be repeated here.

[0202] When a terminal device sends a first message to a network device, the network device can be the first network element.

[0203] In this embodiment, the network device sends a fourth message to the terminal device, which enables the terminal device to send its own resource information to the network device. Based on the terminal device's resource information and the first model, the network device generates a first instruction message or a second instruction message to clarify whether the network device or the network device and the terminal device will subsequently perform the inference task of the first model. This allows the first model to be deployed entirely or partially in the network device, so that the terminal device no longer needs to obtain the inference result from the service device, thereby reducing transmission latency and improving service efficiency.

[0204] In one possible embodiment, before performing step S404 described above, the present application embodiment may further perform the following steps:

[0205] S405: The network device determines whether the first model has been stored in the network device based on the identification information of the first model; if the first model is not stored in the network device, the first model is stored; if the first model is stored in the network device, the first model is invoked.

[0206] The network device determines the first network element in the network device that performs the inference task of the first model.

[0207] This step is the same as step S304 in the embodiment shown in Figure 3 above, and will not be repeated here.

[0208] In one possible embodiment, before performing step S405 as described above, the following steps may also be performed:

[0209] S406: The service device sends a third message to the network device, and the network device receives the third message accordingly.

[0210] The third message includes the first model and its identification information.

[0211] It is understood that in this embodiment, the transfer and deployment of the first model is triggered by the network device. Therefore, after the terminal registers, the service device can send a third message to the network device, distributing the first model and its identification information. After receiving the third message, the network device can execute the above steps S405 and S404, storing the first model in the network device and obtaining the terminal device's resource information to facilitate the subsequent generation of first or second indication information, thereby deploying the first model completely or partially in the network device.

[0212] In this embodiment, the service device actively sends the first model and identification information to the network device. The network device can store or retrieve the first model, obtain resource information from the terminal device, and generate first or second indication information. This allows the first model to be partially or completely deployed in the network device, reducing the transmission latency for obtaining subsequent inference results. When the first model is deployed in both the network device and the terminal device, the resources of the terminal device can also be utilized, improving the resource utilization rate of the terminal device and further reducing transmission latency.

[0213] Optionally, after performing step S403 above, the embodiments of this application may also perform the following steps:

[0214] S407: The terminal device sends a fifth message to the service device where the first model is located, and the service device receives the fifth message accordingly.

[0215] The fifth message includes either the first instruction information or the second instruction information mentioned above.

[0216] It is understood that, in the embodiments of this application, when the second information includes the first indication information, the fifth information includes the first indication information. When the second information includes the second indication information, the fifth information includes the second indication information.

[0217] The fifth piece of information may also include the identification information of the first network element. See the description above for details regarding this identification information.

[0218] In the embodiments of this application, after receiving the second information from the network device, the terminal device sends a fifth message to the service device. This message can provide feedback to the service device that the first model has been transferred and deployed in the network device, and also provide feedback on the subject that will subsequently perform the inference task and the first network element that will actually perform the inference task. This allows the service device to obtain relevant information and eliminates the need to perform the inference task of the first model again, thereby ensuring that the inference result can be obtained without going through the service device and reducing transmission latency.

[0219] Please refer to Figure 5, which is a flowchart illustrating another communication method provided in an embodiment of this application. It is understood that the steps in the embodiments of this application can be considered reasonable variations or supplements to the embodiments in Figure 2 above; or, it is understood that the communication method in the embodiments of this application can also be considered an embodiment that can be executed independently, and this application does not limit it in this regard. In this communication method, the transfer deployment of the first model is triggered by the service device.

[0220] This communication method includes, but is not limited to, the following steps:

[0221] S501: The terminal device sends a first message to the network device, and the network device receives the first message accordingly.

[0222] S502: The network device generates a second message based on the first model and the resource information of the terminal device.

[0223] S503: The network device sends a second message to the terminal device, and the terminal device receives the second message accordingly.

[0224] The steps S501 to S503 described above are the same as steps S201 to S203 in the embodiment shown in Figure 2 above, and will not be repeated here.

[0225] In one possible embodiment, before performing step S501 described above, the following steps may also be performed:

[0226] S504: The service device sends a fourth message to the terminal device, and the terminal device receives the fourth message accordingly.

[0227] It is understandable that step S501 above can specifically be, in response to the fourth message, the terminal device sends the first message to the network device.

[0228] The fourth message is used to obtain resource information of the terminal device. This fourth message includes the identification information of the first network element performing the inference task and the identification information of the first model within the network device. For details regarding the identification information of the first network element and the identification information of the first model, please refer to the description above; they will not be repeated here.

[0229] When a terminal device sends a first message to a network device, the network device can be the aforementioned first network element.

[0230] The relevant descriptions of the identification information of the first network element and the identification information of the first model can be found in the above description, and will not be repeated here.

[0231] In this embodiment, in a scenario where the transfer deployment of the first model is triggered by a service device, the service device can proactively send the aforementioned fourth message to the terminal device, enabling the terminal device to send its resource information to the network device. This allows the network device to generate first or second instruction information. This clarifies whether the subsequent inference task of the first model will be performed by the network device or by both the network device and the terminal device, allowing the first model to be deployed entirely or partially on the network device. This eliminates the need for the terminal device to obtain inference results from the service device, reducing transmission latency and improving service efficiency.

[0232] In one possible embodiment, before performing step S504 as described above, the following steps may also be performed:

[0233] S505: The network device sends the eighth message to the service device where the first model is located, and the service device receives the eighth message accordingly.

[0234] The eighth message includes the identification information of the first network element in the network device that performs the inference task.

[0235] In this embodiment, the first network element is determined in advance by the network device. The network device feeds back the identification information of the first network element to the service device, so that the service device can know the network element in the network device that handles the deployment and inference of the first model. This allows the service device to avoid having to perform the inference task of the first model again, thereby reducing the inference transmission latency.

[0236] In one possible embodiment, before performing step S505 as described above, the following steps may also be performed:

[0237] S506: The network device determines whether the first model has been stored in the network device based on the identification information of the first model; if the first model has not been stored in the network device, the first model is stored; if the first model has been stored in the network device, the first model is invoked.

[0238] The network device determines the first network element in the network device that performs the inference task of the first model.

[0239] This step is the same as step S304 in the embodiment shown in Figure 3 above, and will not be repeated here.

[0240] In one possible embodiment, before performing step S506 described above, the following steps may also be performed:

[0241] S507: The service device sends a third message to the network device, and the network device receives the third message accordingly.

[0242] The third message includes the first model and its identification information.

[0243] The relevant descriptions of the recognition information for the first model are as described above and will not be repeated here.

[0244] It is understood that in this embodiment, the deployment of the first model to the network device is triggered by the service device. Therefore, after registering with the network, the service device can send the first model and its identification information to the network device by sending a third message. After receiving the third message, the network device can execute the above step S506 to store or retrieve the first model and allocate a first network element responsible for processing the inference task of the first model, so that the identification information of the first network element can be sent to the terminal device, resource information can be obtained from the terminal device, and first or second indication information can be generated.

[0245] In this embodiment, the service device actively sends the first model and its identification information to the network device. The network device can store or retrieve the first model, allocate a first network element as a service network element, and generate first or second indication information. This enables the first model to be partially or completely deployed in the network device, reducing the transmission latency in the subsequent process of obtaining inference results. When the first model is deployed in both the terminal device and the network device, the resources of the network device can also be utilized, improving the resource utilization rate of the network device and further reducing transmission latency.

[0246] Optionally, after performing step S503 above, the embodiments of this application may further perform the following steps:

[0247] S508: The terminal device sends a fifth message to the service device where the first model is located, and the service device receives the fifth message accordingly.

[0248] The fifth message includes either the first instruction information or the second instruction information mentioned above.

[0249] It is understood that, in the embodiments of this application, when the second information includes the first indication information, the fifth information includes the first indication information. When the second information includes the second indication information, the fifth information includes the second indication information.

[0250] In the embodiments of this application, after receiving the second information from the network device, the terminal device sends a fifth message to the service device. This message can provide feedback to the service device that the first model has been transferred and deployed in the network device, and also provide feedback on the subject that will subsequently perform the inference task. This allows the service device to obtain relevant information and eliminates the need to perform the inference task of the first model again, thereby ensuring that the inference result can be obtained without going through the service device and reducing transmission latency.

[0251] This application also provides a communication method. It is understood that the steps in this application's embodiments can be executed after any of the embodiments illustrated in Figures 2 to 5, or the communication method in this application's embodiments can also be considered as an embodiment that can be executed independently; this application does not impose any limitations on this. This communication method illustrates the process by which a terminal device obtains the inference results of the first model after the first model is transferred and deployed.

[0252] This communication method includes, but is not limited to, the following steps:

[0253] S601: If the second message includes the first instruction information, the terminal device sends a sixth message to the network device, and the network device receives the sixth message accordingly.

[0254] S602: The network device independently executes the inference task of the first model based on the first data and obtains the first inference result.

[0255] S603: The network device sends the first inference result to the terminal device, and the terminal device receives the first inference result accordingly.

[0256] The sixth message requests the network device to independently execute the inference task. This sixth message includes the first data required to execute the inference task. The first inference result is obtained based on the first data. The network device mentioned in this communication method can be the aforementioned first network element.

[0257] Understandably, since the first instruction information instructs the network device to independently execute the inference task of the first model, the terminal device can send the first data to the network device. The network device then performs inference based on the first data and the first model to obtain the first inference result. This process is completed independently by the network device, without needing to send data to the service device or obtain the inference result from the service device, thus reducing the transmission latency of the entire inference process.

[0258] In one possible embodiment, the present application embodiment may further perform the following steps:

[0259] S604: If the second message includes the second instruction information, the terminal device sends a seventh message to the network device, and the network device receives the seventh message accordingly.

[0260] The seventh message is used to request the terminal device and the network device to jointly perform the inference task. The seventh message includes the second data. The second data is obtained by the terminal device inferring the first data based on the first sub-model, or the second data is obtained by preprocessing the first data. The first data is the data required to perform the inference task.

[0261] It is understood that the above steps S601 to S604 can be executed before step S604 or after step S604, or steps S601 to S604 can be executed multiple times in the communication method, or step S604 can be executed multiple times in the communication method, etc. This application does not limit this.

[0262] In one possible embodiment, after performing step S604 as described above, the present application embodiment may further perform the following steps:

[0263] S605: When the second instruction information indicates that the network device obtains the inference result based on the second sub-model as an intermediate layer feature, the terminal device infers the second data based on the first sub-model to obtain the third data.

[0264] S606: The network device infers the fourth data from the second data based on the second sub-model.

[0265] S607: The network device sends the fourth data to the terminal device. Correspondingly, the terminal device receives the fourth data from the network device.

[0266] S608: The terminal device obtains the second inference result based on the third and fourth data.

[0267] The fourth data is obtained by the network device through reasoning on the second data based on the second sub-model.

[0268] It is understandable that if the second indication information indicates that the inference result obtained by the network device based on the second sub-model is an intermediate layer feature, then the inference result obtained by the network device based on the second sub-model is not the final inference result and needs to be further processed by the terminal device. This indicates that the deployment method of the first model is a parallel deployment method. This parallel deployment method means that the first sub-model and the second sub-model in the first model are deployed in the terminal device and the network device respectively, and the terminal device obtains the final inference result after the terminal device and the network device each perform part of the inference.

[0269] At this point, the second data is obtained by preprocessing the first data. The third and fourth data belong to the intermediate layer features.

[0270] In this embodiment, the terminal device and the network device determine the specific process for jointly executing the inference task based on the second instruction information. The terminal device and the network device respectively use the first sub-model and the second sub-model to perform partial inference and obtain the final inference result. This eliminates the need to send data to the service device and obtain the inference result from the service device, reducing the transmission latency of the entire inference process. Since the terminal device participates in the inference, its resources can be utilized, improving resource utilization. The independent inference performed by the terminal device and the network device reduces the inference time required and improves inference efficiency.

[0271] Optionally, after performing step S604 above, the embodiments of this application may further perform the following steps:

[0272] S609: When the second instruction information indicates that the inference result obtained by the network device based on the second sub-model is the final inference result, the terminal device sends the second data to the network device, and the network device receives the second data accordingly.

[0273] S610: The network device infers the second data based on the second sub-model to obtain the third inference result.

[0274] S611: The network device sends the third inference result to the terminal device, and the terminal device receives the third inference result accordingly.

[0275] It is understandable that the second instruction information indicates that the inference result obtained by the network device based on the second sub-model is the final inference result, indicating that the deployment method of the first model is a serial deployment method. This serial deployment method means that the first sub-model and the second sub-model in the first model are deployed in the terminal device and the network device respectively, and the terminal device performs partial inference and sends the inference result to the network device, and the network device performs partial inference to obtain the final inference result.

[0276] At this point, the second data is obtained by reasoning from the first data. The second data belongs to the intermediate layer features.

[0277] Optionally, the second data can also be obtained by preprocessing the first data and then inferring based on the first sub-model.

[0278] In this embodiment, the terminal device and the network device determine the specific process for jointly executing the inference task based on the second instruction information. After performing partial inference, the terminal device sends the inference result to the network device for further partial inference to obtain the final inference result. This eliminates the need to send data to the service device and obtain the inference result from the service device, reducing the transmission latency of the entire inference process. Since the terminal device participates in inference, its resources can be utilized, improving resource utilization.

[0279] Referring to Figures 6a, 6b, and 6c, Figure 6a shows a flowchart illustrating the process by which the terminal device obtains the inference result after the first model is transferred and deployed; Figure 6b shows another flowchart illustrating the process by which the terminal device obtains the inference result after the first model is transferred and deployed; and Figure 6c shows yet another flowchart illustrating the process by which the terminal device obtains the inference result after the first model is transferred and deployed. Figures 6a, 6b, and 6c correspond to the inference processes of the network device independently executing the inference task of the first model, the terminal device and network device jointly executing the inference task of the first model in a parallel deployment mode, and the terminal device and network device jointly executing the inference task of the first model in a serial deployment mode, respectively.

[0280] The methods of the embodiments of this application have been described in detail above. The following provides an apparatus for implementing any one of the methods in the embodiments of this application. For example, an apparatus is provided that includes a unit (or means) for implementing the steps performed by the device in any of the above methods.

[0281] Please refer to Figure 7, which is a schematic diagram of the structure of a communication device provided in an embodiment of this application.

[0282] As shown in Figure 7, the communication device 70 may include a communication unit 701 and a processing unit 702. The communication unit 701 and the processing unit 702 may be software, hardware, or a combination of software and hardware.

[0283] The communication unit 701 can implement sending and / or receiving functions, and can also be described as a transceiver unit. The communication unit 701 can also be a unit integrating an acquisition unit and a sending unit, wherein the acquisition unit is used to implement the receiving function, and the sending unit is used to implement the sending function. Optionally, the communication unit 701 can be used to receive information sent by other devices, and can also be used to send information to other devices.

[0284] In one possible design, the communication device 70 may correspond to the terminal device in the method embodiments shown in Figures 2 to 6c. For example, the communication device 70 may be a terminal device or a chip within the terminal device. The communication device 70 may include units for performing the operations performed by the terminal device in the method embodiments shown in Figures 2 to 6c, and each unit in the communication device 70 is for implementing the operations performed by the terminal device in the method embodiments shown in Figures 2 to 6c. The descriptions of each unit are as follows:

[0285] The communication unit 701 is used to send a first message to the network device. The first message is used to request the network device to deploy a first model. The first message includes the identification information of the first model and the resource information of the terminal device.

[0286] The communication unit 701 is further configured to receive a second message from the network device, the second message including a first instruction message or a second instruction message, the first instruction message indicating that the network device independently executes the inference task of the first model, the second instruction message indicating that the terminal device and the network device jointly execute the inference task, and indicating that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model.

[0287] In one possible implementation, the device further includes:

[0288] Processing unit 702 is used to generate the first message.

[0289] Regarding the communication unit 701 and processing unit 702 described in this design, the steps they perform can be referred to the implementation methods corresponding to the terminal devices in the method embodiments shown in Figures 2 to 6c above.

[0290] Regarding the technical effects of the implementation methods performed by the communication unit 701 and processing unit 702 described in this design, please refer to the description of the technical effects corresponding to the method embodiments shown in Figures 2 to 6c above.

[0291] In another possible design, the communication device 70 may correspond to the network device in the method embodiments shown in Figures 2 to 6c. For example, the communication device 70 may be a network device or a chip within a network device. The communication device 70 may include units for performing the operations performed by the network device in the method embodiments shown in Figures 2 to 6c, and each unit in the communication device 70 is for implementing the operations performed by the network device in the method embodiments shown in Figures 2 to 6c. The descriptions of each unit are as follows:

[0292] The communication unit 701 is configured to receive a first message from a terminal device, the first message being a request for the network device to deploy a first model, the first message including identification information of the first model and resource information of the terminal device;

[0293] The communication unit 701 is also used to send the second message to the terminal device;

[0294] Processing unit 702 is configured to generate a second message based on the resource information of the first model and the terminal device. The second message includes either the first instruction information or the second instruction information. The first instruction information indicates that the network device independently executes the inference task of the first model, and the second instruction information indicates that the terminal device and the network device jointly execute the inference task. It also indicates that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model.

[0295] Regarding the communication unit 701 and processing unit 702 described in this design, the steps they perform can be referred to the implementation methods corresponding to the network devices in the method embodiments shown in Figures 2 to 6c above.

[0296] Regarding the technical effects of the implementation methods performed by the communication unit 701 and processing unit 702 described in this design, please refer to the description of the technical effects corresponding to the method embodiments shown in Figures 2 to 6c above.

[0297] According to embodiments of this application, the various units in the device shown in FIG7 can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effect of the embodiments of this application. The above units are based on logical function division. In practical applications, the function of one unit can also be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the electronic device may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0298] It should be noted that the implementation of each unit can also refer to the corresponding description of the method embodiments shown in Figures 2 to 6c above.

[0299] In the communication device 70 described in Figure 7, the network device can generate first instruction information or second instruction information based on the resource information of the first model and the terminal device, so that the first model can be fully or partially deployed in the network device, and the network device can independently execute the inference task of the first model or the terminal device and the network device can jointly execute the inference task of the first model without obtaining the inference result of the first model from the service device, which can reduce transmission latency.

[0300] Please refer to Figure 8, which is a schematic diagram of the structure of a communication device provided in an embodiment of this application.

[0301] It should be understood that the communication device 80 shown in FIG8 is only an example. The communication device in the embodiments of this application may also include other components, or include components with functions similar to the various components in FIG8, or may not include all the components in FIG8.

[0302] The communication device 80 includes a communication interface 801 and at least one processor 802.

[0303] The communication device 80 can correspond to any network element or device among terminal devices and network devices. The communication interface 801 is used for sending and receiving signals, and at least one processor 802 executes program instructions, causing the communication device 80 to implement the corresponding process of the method executed by the corresponding device in the above method embodiments.

[0304] In one possible design, the communication device 80 may correspond to the terminal device in the method embodiments shown in Figures 2 to 6c. For example, the communication device 80 may be a terminal device or a chip within the terminal device. The communication device 80 may include components for performing the operations performed by the terminal device in the above method embodiments, and each component in the communication device 80 is specifically designed to implement the operations performed by the terminal device in the above method embodiments. Specifically, it may be as follows:

[0305] Send a first message to the network device, the first message being used to request the network device to deploy a first model, the first message including the identification information of the first model and the resource information of the terminal device;

[0306] The system receives a second message from the network device, the second message including a first instruction or a second instruction, the first instruction indicating that the network device independently performs the inference task of the first model, the second instruction indicating that the terminal device and the network device jointly perform the inference task, and instructing the terminal device and the network device to respectively deploy a first sub-model and a second sub-model in the first model.

[0307] In another possible design, the communication device 80 may correspond to the network device in the method embodiments shown in Figures 2 to 6c. For example, the communication device 80 may be a network device or a chip within a network device. The communication device 80 may include components for performing the operations performed by the network device in the above method embodiments, and each component in the communication device 80 is specifically designed to implement the operations performed by the network device in the above method embodiments. Specifically, it may be as follows:

[0308] Receive a first message from a terminal device, the first message being used to request the network device to deploy a first model, the first message including identification information of the first model and resource information of the terminal device;

[0309] Based on the resource information of the first model and the terminal device, a second message is generated. The second message includes the first instruction information or the second instruction information. The first instruction information indicates that the network device independently executes the inference task of the first model. The second instruction information indicates that the terminal device and the network device jointly execute the inference task. It also indicates that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model.

[0310] The second message is sent to the terminal device.

[0311] In the communication device 80 described in Figure 8, the network device can generate first instruction information or second instruction information based on the resource information of the first model and the terminal device, so that the first model can be fully or partially deployed in the network device, and the network device can independently execute the inference task of the first model or the terminal device and the network device can jointly execute the inference task of the first model without obtaining the inference result of the first model from the service device, which can reduce transmission latency.

[0312] For cases where the communication device can be a chip or a chip system, please refer to the schematic diagram of the chip structure shown in Figure 9.

[0313] As shown in Figure 9, chip 90 includes processor 901 and interface 902. There can be one or more processors 901, and multiple interfaces 902. It should be noted that the functions of processor 901 and interface 902 can be implemented through hardware design, software design, or a combination of both; no restrictions are placed here.

[0314] Optionally, chip 90 may also include memory 903, which is used to store necessary program instructions and data.

[0315] In this application, processor 901 can be used to call implementation programs of communication methods provided in one or more embodiments of this application in one or more devices or network elements of a terminal device or network device from memory 903, and execute the instructions contained in the program. Interface 902 can be used to output the execution results of processor 901. In this application, interface 902 can specifically be used to output various messages or information of processor 901.

[0316] The communication methods provided by one or more embodiments of this application can be referred to the various embodiments shown in Figures 2 to 6c above, and will not be repeated here.

[0317] The processor in this embodiment can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0318] The memory in this application embodiment is used to provide storage space, in which data such as operating system and computer programs can be stored. The memory includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact disc read-only memory (CD-ROM).

[0319] According to the method provided in the embodiments of this application, the embodiments of this application also provide a computer-readable storage medium storing a computer program. When the computer program is run on one or more processors, it can implement the method shown in Figures 2 to 6c.

[0320] According to the method provided in the embodiments of this application, the embodiments of this application also provide a computer program product, which includes a computer program. When the computer program runs on a processor, it can implement the methods shown in Figures 2 to 6c.

[0321] This application also provides a system comprising at least one communication device 70, communication device 80, or chip 90 as described above, for performing the steps performed by the corresponding device in any of the embodiments of FIG2 to FIG6c.

[0322] This application also provides a system comprising a terminal device and a network device. The terminal device is used to execute the steps performed by the terminal device in any of the embodiments of Figures 2 to 6c, and the network device is used to execute the steps performed by the network device in any of the embodiments of Figures 2 to 6c.

[0323] This application also provides a processing apparatus, including a processor and an interface; the processor is used to execute the method in any of the above method embodiments.

[0324] It should be understood that the aforementioned processing device can be a chip. For example, the processing device can be a field-programmable gate array (FPGA), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf FPGA, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, a system-on-chip (SoC), a central processing unit (CPU), a network processor (NP), a digital signal processor (DSP), a microcontroller unit (MCU), a programmable logic device (PLD), or other integrated chips. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0325] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0326] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0327] The units in the above-described device embodiments and the electronic devices in the method embodiments completely correspond to each other, with corresponding modules or units performing corresponding steps. For example, the communication unit (transceiver) performs the receiving or sending steps in the method embodiments, while other steps besides sending and receiving can be performed by the processing unit (processor). The functions of specific units can be found in the corresponding method embodiments. There can be one or more processors.

[0328] It is understood that in the embodiments of this application, the electronic device may perform some or all of the steps in the embodiments of this application. These steps or operations are merely examples, and the embodiments of this application may also perform other operations or variations thereof. Furthermore, the steps may be performed in different orders as presented in the embodiments of this application, and it is not necessarily necessary to perform all the operations in the embodiments of this application.

[0329] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0330] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0331] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0332] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0333] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0334] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the contributing part, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0335] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A communication method, characterized in that, Applied to a terminal device, the method includes: Send a first message to the network device, the first message being used to request the network device to deploy a first model, the first message including the identification information of the first model and the resource information of the terminal device; The system receives a second message from the network device, the second message including a first instruction or a second instruction, the first instruction indicating that the network device independently performs the inference task of the first model, the second instruction indicating that the terminal device and the network device jointly perform the inference task, and instructing the terminal device and the network device to respectively deploy a first sub-model and a second sub-model in the first model.

2. The method according to claim 1, characterized in that, Before sending the first message to the network device, it also includes: Based on the resource information of the first model and the terminal device, determine whether the terminal device can independently execute the inference task; Sending the first message to the network device includes: If the terminal device cannot perform the inference task independently, the first message is sent to the network device.

3. The method according to claim 2, characterized in that, Before determining whether the terminal device can independently execute the inference task based on the first model and the resource information of the terminal device, the method further includes: A third message is received, the third message including the first model and the identification information of the first model.

4. The method according to claim 1, characterized in that, Before sending the first message to the network device, it also includes: Receive a fourth message, the fourth message including the identification information of the first network element performing the inference task in the network device and the identification information of the first model; Sending the first message to the network device includes: In response to the fourth message, the first message is sent to the network device, where the network device is the first network element.

5. The method according to claim 3 or 4, characterized in that, The method further includes: A fifth message is sent to the service device where the first model is located, the fifth message including the first indication information or the second indication information.

6. The method according to claim 2 or 3, characterized in that, The second message also includes the identification information of the first network element in the network device that performs the inference task.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: If the second message includes the first indication information, a sixth message is sent to the network device, the sixth message being used to request the network device to independently execute the inference task, the sixth message including first data required to execute the inference task; Receive the first inference result, which is obtained based on the first data.

8. The method according to any one of claims 1-7, characterized in that, The method further includes: If the second message includes the second indication information, a seventh message is sent to the network device. The seventh message is used to request the terminal device to jointly execute the inference task with the network device. The seventh message includes second data. The second data is obtained by the terminal device inferring the first data based on the first sub-model, or the second data is obtained by preprocessing the first data. The first data is the data required to execute the inference task.

9. The method according to claim 8, characterized in that, The method further includes: When the second indication information indicates that the network device obtains an intermediate layer feature based on the second sub-model, the third data is obtained by reasoning the second data based on the first sub-model. Receive fourth data from the network device, the fourth data being obtained by the network device from reasoning about the second data based on the second sub-model; The second reasoning result is obtained based on the third and fourth data.

10. The method according to any one of claims 1-9, characterized in that, The first model includes an artificial intelligence model or a machine learning model.

11. A communication method, characterized in that, Applied to network devices, the method includes: Receive a first message from a terminal device, the first message being used to request the network device to deploy a first model, the first message including identification information of the first model and resource information of the terminal device; Based on the resource information of the first model and the terminal device, a second message is generated. The second message includes the first instruction information or the second instruction information. The first instruction information indicates that the network device independently executes the inference task of the first model. The second instruction information indicates that the terminal device and the network device jointly execute the inference task. It also indicates that the terminal device and the network device respectively deploy the first sub-model and the second sub-model in the first model. The second message is sent to the terminal device.

12. The method according to claim 11, characterized in that, Before receiving the first message from the terminal device, it also includes: A fourth message is sent to the terminal device. The fourth message is used to obtain the resource information of the terminal device. The fourth message includes the identification information of the first network element in the network device that performs the inference task and the identification information of the first model.

13. The method according to claim 11, characterized in that, Before receiving the first message from the terminal device, the method further includes: sending an eighth message to the service device where the first model is located, the eighth message including the identification information of the first network element in the network device that performs the inference task.

14. The method according to claim 12 or 13, characterized in that, The method further includes receiving a third message, the third message including the first model and the identification information of the first model.

15. The method according to claim 11, characterized in that, The first message also includes the first model.

16. The method according to claim 14 or 15, characterized in that, The method further includes: Based on the identification information of the first model, determine whether the first model has been stored in the network device; If the first model is not stored in the network device, the first model is stored. If the first model has been stored in the network device, the first model is invoked. Identify the first network element in the network device that performs the inference task of the first model.

17. The method according to any one of claims 11-16, characterized in that, The method further includes: If the second message includes the first indication information, a sixth message is received from the terminal device, the sixth message being used to request the network device to independently execute the inference task, the sixth message including first data required to execute the inference task; Based on the first data, the reasoning task is executed independently to obtain the first reasoning result; The first inference result is sent to the terminal device.

18. The method according to any one of claims 11-17, characterized in that, The method further includes: If the second message includes the second indication information, a seventh message is received from the terminal device. The seventh message is used to request the terminal device and the network device to jointly execute the inference task. The seventh message includes second data. The second data is obtained by the terminal device from the first data based on the first sub-model, or the second data is obtained by the terminal device from the first data after preprocessing. The first data is the data required to execute the inference task.

19. The method according to claim 18, characterized in that, The method further includes: The fourth data is obtained by reasoning from the second data based on the second sub-model; The fourth data is sent to the terminal device.

20. A communication device, characterized in that, Includes units for performing the method as described in any one of claims 1 to 10 or 11 to 19.

21. A communication device, characterized in that, Includes a processor for performing the method as claimed in any one of claims 1 to 10 or 11 to 19.

22. A communication device, characterized in that, It includes logic circuits and interfaces, wherein the logic circuits and the interfaces are coupled; The interface is used for inputting and / or outputting information, and the logic circuit is used for performing the method as claimed in any one of claims 1 to 10 or 11 to 19.

23. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, which, when executed, performs the method as claimed in any one of claims 1 to 10 or 11 to 19.

24. A computer program product, characterized in that, The computer program product includes a computer program that, when executed, performs the method as claimed in any one of claims 1 to 10 or claims 11 to 19.

25. A communication system, characterized in that, Includes at least one of the following: Terminal equipment, network equipment; The terminal device is used to perform the method as described in any one of claims 1 to 10, and the network device is used to perform the method as described in any one of claims 11 to 19.

Images