Method and device for wireless communication

Abstract

The present application provides a method and device for wireless communication. The method comprises: a first device acquires first information, wherein the first information is used for managing models associated with the first device, and the models associated with the first device comprise a first model; and the first device manages the first model on the basis of the first information. According to embodiments of the present application, information (i.e., the first information) used for managing the models associated with the first device is acquired, thereby implementing management of the models, for example, management of parallel processing conflicts of model operation under the condition that the capability of a terminal device is limited.

Description

Method and apparatus for wireless communication Technical Field

[0001] The present application relates to the field of communication technology, and more particularly, to a method and device for wireless communication. Background Art

[0002] In the near future, communication systems may introduce a large number of models for different functions. This large number of models may lead to some problems, such as the need for terminal devices to perform inference and training on multiple models simultaneously. In this case, there is currently no suitable solution for managing these models.

[0003] Summary of the Invention

[0004] The present application provides a method and device for wireless communication. The following introduces various aspects involved in the present application.

[0005] In a first aspect, a method for wireless communication is provided, including: a first device obtains first information, where the first information is used to manage a model associated with the first device, the model associated with the first device includes a first model; and the first device manages the first model based on the first information.

[0006] In a second aspect, a method for wireless communication is provided, comprising: a second device sending first information to a first device, wherein the first information is used to manage a model associated with the first device, and the model associated with the first device includes a first model.

[0007] According to a third aspect, a device for wireless communication is provided, which is a first device. The device includes: an acquisition unit for acquiring first information, wherein the first information is used to manage a model associated with the first device, and the model associated with the first device includes a first model; and a management unit for managing the first model based on the first information.

[0008] In a fourth aspect, a device for wireless communication is provided, wherein the device for wireless communication is a second device, and the device includes: a sending unit for sending first information to a first device, wherein the first information is used to manage a model associated with the first device, and the model associated with the first device includes a first model.

[0009] In a fifth aspect, a wireless communication device is provided, comprising a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory so that the device executes part or all of the steps in the method of the first aspect or the second aspect.

[0010] In a sixth aspect, a wireless communication device is provided, comprising a processor for calling a program from a memory so that the device executes part or all of the steps in the method of the first aspect or the second aspect.

[0011] In a seventh aspect, a chip is provided, comprising a processor for calling a program from a memory so that the chip executes part or all of the steps in the method of the first aspect or the second aspect.

[0012] In an eighth aspect, a computer-readable storage medium is provided, on which a program is stored, wherein the program enables a wireless communication device to execute part or all of the steps in the method of the first aspect or the second aspect.

[0013] In a ninth aspect, a computer program product is provided, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a communication device to perform some or all of the steps of the methods of the various aspects described above. In some implementations, the computer program product may be a software installation package.

[0014] In a tenth aspect, a computer program is provided, which enables a wireless communication device to perform part or all of the steps in the method of the first aspect or the second aspect.

[0015] The embodiment of the present application helps to implement model management by obtaining information (i.e., first information) used to manage the model associated with the first device, such as the management of parallel processing conflicts in model operation when the terminal device capabilities are limited. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG1 is a wireless communication system used in an embodiment of the present application.

[0017] FIG2 is a schematic diagram of a neural network applicable to an embodiment of the present application.

[0018] Figure 3 is a schematic diagram of a convolutional neural network applicable to an embodiment of the present application.

[0019] FIG4 is a schematic diagram of an AI / ML air interface model framework applicable to an embodiment of the present application.

[0020] FIG5 is a flow chart of a method for wireless communication provided in an embodiment of the present application.

[0021] FIG6 is a schematic diagram of a device for wireless communication according to an embodiment of the present application.

[0022] FIG7 is a schematic diagram of another device for wireless communication according to an embodiment of the present application.

[0023] FIG8 is a schematic structural diagram of a communication device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solution in this application will be described below with reference to the accompanying drawings.

[0025] Figure 1 illustrates a wireless communication system 100 used in an embodiment of the present application. The wireless communication system 100 may include a network device 110 and a terminal device 120. The network device 110 may be a device that communicates with the terminal device 120. The network device 110 may provide communication coverage for a specific geographic area and may communicate with the terminal device 120 within the coverage area.

[0026] FIG1 exemplarily shows a network device and two terminals. Optionally, the wireless communication system 100 may include multiple network devices and each network device may include other numbers of terminal devices within its coverage area, which is not limited in the embodiments of the present application.

[0027] Optionally, the wireless communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiment of the present application.

[0028] It should be understood that the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: fifth generation (5G) system or new radio (NR), long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), etc. The technical solutions provided in this application can also be applied to future communication systems, such as the sixth generation mobile communication system, satellite communication system, etc.

[0029] The terminal device in the embodiments of the present application may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station (MS), mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device. The terminal device in the embodiments of the present application may refer to a device that provides voice and / or data connectivity to a user and can be used to connect people, objects and machines, such as a handheld device with wireless connection function, a vehicle-mounted device, etc. The terminal device in the embodiments of the present application can be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Optionally, the UE can be used to act as a base station. For example, the UE can act as a scheduling entity that provides sidelink signals between UEs in V2X or D2D, etc. For example, a cellular phone and a car communicate with each other using sidelink signals. The cellular phone and smart home devices communicate without relaying the communication signal through the base station.

[0030] The network device in the embodiments of the present application may be a device for communicating with a terminal device, and the network device may also be referred to as an access network device or a radio access network device, such as a base station. The network device in the embodiments of the present application may refer to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network, a core network device, a model monitoring and management device, or an operation administration and maintenance (OAM) device. The base station can broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station MeNB, secondary station SeNB, multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), centralized unit-control plane (CU-CP), centralized unit-user plane (CU-UP) positioning node, etc.Core network equipment can broadly cover the following various names, or be replaced with the following names, such as: location management function (LMF) network element, network slice selection function (NSSF), authentication server function (AUSF), unified data management (UDM), access and mobility management function (AMF), session management function (SMF), policy control function (PCF), user plane function (UPF), sensing control function (SF), network data analysis function (NWDAF) network element.

[0031] The base station may be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station may also refer to a communication module, a modem or a chip for being set in the aforementioned device or apparatus. The base station may also be a mobile switching center and a device that performs the base station function in device-to-device D2D, vehicle-to-everything (V2X), machine-to-machine (M2M) communications, a network side device in a 6G network, a device that performs the base station function in future communication systems, and the like. The base station may support networks with the same or different access technologies. The embodiments of the present application do not limit the specific technology and specific device form adopted by the network equipment. The base station may be fixed or mobile. For example, a helicopter or a drone may be configured to act as a mobile base station, and one or more cells may move according to the location of the mobile base station. In other examples, a helicopter or a drone may be configured to be used as a device for communicating with another base station.

[0032] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0033] In some deployments, the network device in the embodiments of the present application may refer to a CU or a DU, or the network device may include a CU and a DU. The gNB may also include an AAU.

[0034] The network equipment and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; they can also be deployed in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the scenarios in which the network equipment and terminal devices are located.

[0035] It should be understood that all or part of the functions of the communication device in this application can also be implemented through software functions running on hardware, or through virtualization functions instantiated on a platform (such as a cloud platform).

[0036] Artificial intelligence (AI)

[0037] In recent years, artificial intelligence research represented by neural networks has achieved great results in many fields, and it will play an important role in people's production and life for a long time to come.

[0038] FIG2 is a schematic diagram of a neural network applicable to an embodiment of the present application. The neural network shown in FIG2 can be divided into three categories according to the positions of different layers: an input layer 210, a hidden layer 220, and an output layer 230. Generally speaking, the first layer is the input layer 210, the last layer is the output layer 230, and the intermediate layers between the first and last layers are all hidden layers 220. Samples can be input from the input layer 210, processed by the hidden layer 220, and the final result is generated in the output layer 230. Among them, each node represents a processing unit, which can be considered to simulate a neuron. Multiple neurons form a layer of a neural network, and multiple layers of information transmission and processing construct an overall neural network.

[0039] With the continuous development of neural network research, deep learning algorithms have been proposed in recent years, introducing a large number of hidden layers. This layer-by-layer training of multi-hidden neural networks for feature learning has greatly improved the learning and processing capabilities of neural networks and has been widely applied in pattern recognition, signal processing, optimization and combination, anomaly detection, and other fields.

[0040] Similarly, with the development of deep learning, convolutional neural networks have also been further studied. Figure 3 is a schematic diagram of a convolutional neural network to which an embodiment of the present application is applicable. The basic structure of a convolutional neural network may include: an input layer 310, multiple convolutional layers 320, multiple pooling layers 330, a fully connected layer 340, and an output layer 350. The introduction of the convolutional layer 320 and the pooling layer 330 effectively controls the sharp increase in network parameters, limits the number of parameters, and exploits the characteristics of the local structure, thereby improving the robustness of the algorithm.

[0041] Model Management

[0042] The AI / ML air interface model framework shown in Figure 4 was adopted during relevant technical discussions, and model management has become an important part of the air interface model architecture.

[0043] Referring to Figure 4, the AI / ML air interface model framework may include functional modules such as data collection 410, model training 420, model management 430, inference 440, and model storage 450. Among them, the data collection function can collect training data, monitoring data, and inference data, and transmit these data to the model training function, model management function, and inference function respectively. The model training function can store the trained model or the updated model (trained / updated model) in the model storage function. The model storage function can transfer / deliver the model to the inference function to realize model inference.

[0044] Model management functions, shown in Figure 4, include overseeing AI / ML model execution or monitoring AI / ML models and AI / ML-based functions. Model management is also responsible for making decisions based on data received from data collection and inference functions to ensure correct inference operations. Model management functions specifically include managing instructions, requesting model transfers, and monitoring outputs.

[0045] These management instructions can serve as input to the inference function to fine-tune its operation. Information related to these management instructions can include information about the selection, (de)activation, or switching of AI / ML models or AI / ML-based functions, as well as fallback to non-AI / ML operations (i.e., those not relying on the inference process).

[0046] The above-mentioned model transfer / delivery request can be used to request a model from the model storage function.

[0047] The above monitoring outputs can be monitored for (re)training purposes.

[0048] Model Transfer

[0049] In some scenarios, such as when the model management function requests the transmission of a model, the model can be transmitted between communication devices. For example, the model can be transmitted over the air interface between communication devices. Currently, the communication protocol specifies seven ways of transmitting models from a network device to a terminal device. The following is an introduction to model transmission methods 1 to 7, where the network device can be any of the following: access network equipment; core network equipment other than the location management function (LMF); LMF; server (for example, an over-the-top (OTT) server, information management equipment, or operation administration and maintenance (OAM) equipment).

[0050] Model transmission mode 1: The access network device transmits the model to the terminal device through a radio resource control (RRC) message.

[0051] Model transmission mode 2: The core network device (except LMF) transmits the model to the terminal device through NAS signaling.

[0052] Model transmission mode 3: LMF transmits the model to the terminal device through the long term evolution positioning protocol (LPP) message.

[0053] Model transmission method 4: The access network device transmits the model to the terminal device through the user.

[0054] Model transmission mode 5: The core network device (except LMF) transmits the model to the terminal device through the user.

[0055] Model transmission mode 6: LMF transmits the model to the terminal device through the user. The message used to transmit the model is not defined in 3GPP.

[0056] Model transmission mode 7: The server transmits the model to the terminal device, where the message used to transmit the model is not defined in 3GPP.

[0057] In other implementations, if the communication system supports model transmission, the model identifier (such as model ID) can be used for lifecycle management of the model.

[0058] Current technical discussions have proposed a variety of use cases for applying AI to wireless communications. For example, AI / machine learning (ML) technologies can be used to compress and decompress channel state information (CSI), reducing air interface transmission overhead and improving the accuracy of CSI feedback information. AI / ML technologies can also be used to predict beam information in the time and spatial domains to improve the accuracy of beam selection. AI / ML technologies can also be used to predict UE location information to improve the accuracy of terminal device location information in non-line of sight (NLOS) scenarios.

[0059] In the future, new AI-based use cases may be added, such as those related to mobility. Mobility-related use cases may include target cell / beam decision, RRM measurement prediction, failure prediction and repair, and other sub-use cases.

[0060] When a large number of models are introduced, some problems may arise. For example, the current application scenarios of AI / ML models are relatively limited, and a separate model needs to be designed for each sub-use case or sub-scenario, which also means that the terminal device may need to perform reasoning and training processes of multiple models at the same time. As an example, the terminal device may need to perform reasoning of AI models for beam management and positioning services at the same time. As another example, the terminal device may need to perform RRM measurement prediction for the source cell and the target cell based on the AI ​​model at the same time. As another example, the terminal device may need to perform CSI compression model training and positioning model reasoning at the same time. However, due to the limitations of the terminal device's capabilities, such as computing power, battery capacity, and running memory size, the simultaneous operation of multiple models may be restricted, such as the easy occurrence of model parallel processing conflicts (also known as model parallel conflicts).

[0061] In this case, there is currently no suitable solution for managing the model.

[0062] In order to solve the above problems, an embodiment of the present application provides a method for wireless communication, which obtains information (i.e., first information) used to manage the model associated with the first device, which helps to implement model management, such as the management of parallel processing conflicts in model operation when the terminal device capabilities are limited.

[0063] The following describes a method for wireless communication provided by an embodiment of the present application in conjunction with Figure 5. The method shown in Figure 5 can be applied to the communication scenario shown in Figure 1 (e.g., the first device can be terminal device 120, and the second device can be network device 110); the method shown in Figure 5 can also be applied to a sideline communication scenario (e.g., the first device and the second device can both be terminal devices), which is not limited in this application.

[0064] The method shown in FIG. 5 may include step S510 and step S520 .

[0065] In step S510 , the first device obtains first information.

[0066] In step S520 , the first device manages the first model based on the first information.

[0067] The first model may be a model associated with the first device, such as the AI / ML model mentioned above, or the model associated with the first device may include the first model. For example, the first model uses resources of the first device for training / inference, the first model is a model run by the first device (or the first device performs inference / training of the first model), or the first device communicates based on the first model.

[0068] The above-mentioned first information can be used to manage the model associated with the first device. For example, the above-mentioned first information can be used to determine the management policy of the first model. In some embodiments, the first information may include one or more of the following information: second information for indicating a model identifier; third information associated with a model priority; fourth information associated with a pause in model inference / training and / or an adjustment of the inference cycle of the model; fifth information associated with model deactivation; sixth information for indicating the inference order of the model; seventh information for indicating performance gain information of the model; eighth information for indicating a model that must be inferred / trained by the first device; and ninth information associated with a replacement model.

[0069] The following will introduce the above information that may be included in the first information respectively.

[0070] The second information is used to indicate the model identifier

[0071] In some embodiments, the first information may include second information, which may be used to indicate a model identifier. The model identifier may be a symbol, code, or name that uniquely identifies and recognizes a specific model. For example, the model identifier may be the version number of the model, the name of the model, or the ID of the model. The model identifier may be used to identify and distinguish when using, managing, and tracking the model. In an embodiment of the present application, the model identifier may be used to uniquely identify the model to be managed. In some embodiments, the model identifier may also be used to indicate relevant information of the model.

[0072] The third information is associated with the model priority

[0073] In some embodiments, the first information may include third information associated with the model priority. The third information may, for example, include one or more of the following: information on whether the configuration of the model priority is supported; information on whether the priority processing function is enabled; and model priority information.

[0074] In some embodiments, the model priority information can be determined based on a variety of information. For example, the model priority information can be determined based on the service type of the model. As an example, among the beam management service, positioning service and mobility service, the beam management service has the highest priority, the positioning service has the second highest priority (that is, the priority of the positioning service is lower than the priority of the beam management service and higher than the priority of the mobility service), and the mobility service has the lowest priority. For another example, the priority information of the model can be determined based on the current service of the first device. As an example, the model associated with the current service of the first device has the highest priority.

[0075] In some embodiments, the model priority information is related to the second information and / or the business type associated with the model. That is, the configuration granularity of the model priority information can be model granularity or business type granularity. For example, the model priority information at the model granularity can indicate the model priority information of a single model, wherein the model priority information of multiple models can be the same or different. For another example, the model priority information at the business type granularity can indicate the model priority information of multiple models associated with the same type of business, wherein the model priority information of multiple models associated with the same type of business is the same.

[0076] The model priority information may indicate the priority order of the models, and there are various methods in which the model priority information indicates the priority order of the models.

[0077] In some embodiments, the model priority information may be an item in a discrete set. The priority order corresponding to each item in the discrete set may be determined or predefined according to usage requirements, and this application does not limit this. For example, the model priority information may be an item in the set {0, 1, 2, 3}, where 3 represents the highest priority, 2 represents the second highest priority (i.e., lower than the priority corresponding to 3 and higher than the priority corresponding to 0), and 0 represents the lowest priority. Taking the model priority information associated with the service type mentioned above as an example, the model priority information may be represented as a discrete set {BM, Pos, Mob}, where the elements in the set respectively represent that the beam management (BM) service has the highest priority, the positioning (Pos) service has the second highest priority, and the mobility service (Mob) service has the lowest priority.

[0078] In some embodiments, the model priority information may be a continuous value. The correspondence between the continuous value and the priority order may be determined or predefined based on usage requirements, and this application does not impose any limitation thereto. For example, if the model priority information of the first model may be 3.4 and the model priority information of the second model may be 2.1, then the priority of the first model is higher than that of the second model.

[0079] In some embodiments, the model priority information may be an item from a predefined set with a default order. For example, the model priority information may be an item from {p0, p1, p2}, where the correspondence between p0, p1, and p2 and the priority order may be predefined, such as increasing from the front to the back in the set, that is, p0 has the lowest priority and p2 has the highest priority.

[0080] In some embodiments, the third information may include information on whether model priority configuration is supported and / or information on whether a priority processing function is enabled. For example, if the model does not support information on model priority configuration, this information may help ensure that the necessary models of the first device can operate normally. For another example, if the model supports information on model priority configuration, the model may determine whether to enable the priority processing function based on usage requirements, thereby helping to improve the flexibility of model management.

[0081] Whether the model supports the configuration of model priority, and / or whether the model has the priority processing function enabled can be indicated in a variety of ways, such as explicit indication or implicit indication. For example, one or more bits can be used to indicate whether the model supports the configuration of model priority, such as "0" indicating that the configuration of model priority is not supported, "1" indicating that the configuration of model priority is supported, or "1" indicating that the configuration of model priority is not supported, and "0" indicating that the configuration of model priority is supported. For another example, a bit can be used to indicate whether the model has the priority processing function enabled, such as "0" indicating that the priority processing function is not enabled, "1" indicating that the priority processing function is enabled, or "1" indicating that the priority processing function is not enabled, and "0" indicating that the priority processing function is enabled. The method of indicating whether the model supports the configuration of model priority can include, for example, implicit indication. As an example, if the model priority information is configured, it indicates that the model supports the configuration of model priority; if the model priority information is not configured, it indicates that the model does not support the configuration of model priority.

[0082] Whether the model supports configuring model priority and whether the model has priority processing enabled can be indicated separately or in combination. For example, "10" can be used to indicate that the model supports priority processing but the priority processing function is not currently enabled, and "11" can be used to indicate that the model supports priority processing and the priority processing function is currently enabled. For another example, "1" can be used to indicate that the model supports configuring model priority and the priority processing function is enabled, and "0" can be used to indicate that the model does not support configuring model priority and the priority processing function is not enabled.

[0083] It should be noted that the indication meaning of the above bits is only given for example and is not limited in this application.

[0084] The fourth information is associated with the suspension of model inference / training and / or the adjustment of the model's inference cycle

[0085] In some embodiments, the first information may include fourth information associated with the suspension of model reasoning / training and / or the adjustment of the model's reasoning cycle. In other words, the fourth information may be information associated with model backoff. The model backoff mentioned here may refer to pausing the model's reasoning / training and / or adjusting the model's reasoning cycle. Based on this, the resources of the first device can be deployed to resolve conflicts in the parallel execution of multiple models by pausing the reasoning / training of one or part of the model, or adjusting the reasoning cycle of one or part of the model. In other words, the problem of conflicts in the parallel execution of multiple models can be resolved by model backoff.

[0086] In some embodiments, the fourth information may include a maximum number of times, which may be used to indicate the maximum value of the model reasoning / training backoff number. The model reasoning / training backoff number mentioned here may be the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted.

[0087] In some embodiments, the maximum backoff number may be the maximum backoff number within the first time period, for example, a maximum of 5 backoffs may be allowed within 10 seconds, where the first time period is 10 seconds and the maximum backoff number is 5. In other embodiments, the maximum backoff number may be a ratio of the number of model backoffs to the number of model parallel conflicts, for example, the maximum backoff number is 50%.

[0088] In some embodiments, if the number of model inference / training backoffs for the first model is less than the maximum number, the first model may perform model backoff. For example, if the maximum backoff number is 50%, and the ratio of the current model backoff number to the number of model parallel conflicts is 20%, then model backoff may be performed.

[0089] In some embodiments, if the number of model inference / training backoffs for the first model is greater than or equal to the maximum number, the first device deactivates the first model to save resources. Still taking the above-mentioned maximum backoff number of 50% as an example, if the ratio of the current model backoff number to the number of model parallel conflicts is 60%, the model can be deactivated. In other embodiments, if the number of model inference / training backoffs for the first model is greater than or equal to the maximum number, the first device performs inference / training of the first model to avoid performance loss caused by the inability of the model to run.

[0090] When the number of model inference / training backoffs is greater than or equal to the maximum number, which of the above processing methods (deactivation or execution) is performed on the first model can be determined based on the importance of the first model, such as priority, whether it is associated with the current business of the first device, and other information.

[0091] In some embodiments, the fourth information may include a first duration for indicating a backoff duration for model inference and / or training to improve the flexibility of model management. The backoff duration may be a duration for suspending model inference / training and / or adjusting the model inference cycle within the backoff duration.

[0092] For example, the first duration can be a fixed value for ease of implementation. As an example, the first duration can be 2 seconds. If a model parallelism conflict occurs, the first model can be suspended for inference / training for 2 seconds, or the inference cycle of the model can be adjusted within 2 seconds.

[0093] For another example, the first duration can be associated with multiple values, such as a fixed list or sequence, and the value of the first duration can be determined based on the number of backoffs, or in other words, the first duration can be related to the number of model inference / training backoffs. As an example, the first duration can be {1s, 2s, 4s}. After the first model parallel conflict occurs (or the first model backoff is executed, that is, the number of model backoffs is 1), the model suspends inference / training for 1s, or adjusts the model's inference cycle within 1s. After the second model parallel conflict occurs (or the number of model backoffs is 2 times), the model suspends inference / training for 2s, or adjusts the model's inference cycle within 2s. After the third model parallel conflict occurs (or the number of model backoffs is 3 times), the model suspends inference / training for 4s, or adjusts the model's inference cycle within 4s.

[0094] The more times model parallel conflicts occur, the more strained the resources of the first device are. Based on this, the first duration can be positively correlated with the number of backoffs for model reasoning / training, that is, by increasing the duration of model backoff, model parallel conflicts can be avoided, or the frequency of model parallel conflicts can be reduced. For example, the first duration can be a sequence with an initial value of 2s and an increment of 1s. The first time a model parallel conflict occurs, or the first time a model backoff is executed, the model suspends reasoning / training for 2s, or the reasoning cycle of the model is adjusted within 2s. The second time a model parallel conflict occurs, or the second time a model backoff is executed, the model suspends reasoning / training for 3s, or the reasoning cycle of the model is adjusted within 3s. The third time a model parallel conflict occurs, or the third time a model backoff is executed, the model suspends reasoning / training for 4s, or the reasoning cycle of the model is adjusted within 4s.

[0095] In order to improve the accuracy of model management, the first duration can be positively correlated with the number of backoffs of model reasoning / training within the second time period to eliminate the impact of model backoffs with a longer interval with the current model conflict on the model management strategy. For example, the first duration can be positively correlated with the number of backoffs of model reasoning / training within 1 minute, that is, the second time period is 1 minute. In other words, if the time interval between two model parallel conflicts is greater than a certain threshold, such as in the second time period, the number of model parallel conflicts is recounted, or the previous number of model parallel conflicts is cleared. If the time interval between two model parallel conflicts is less than or equal to a certain threshold, the model backoff duration can be determined according to the above sequence associated with the first duration, such as the incremental sequence.

[0096] As mentioned above, model backoff may include adjusting the inference cycle of the model. Therefore, the fourth information may include model inference cycle information. For example, the model inference cycle information may include a first cycle and / or a second cycle, the first cycle being applied to the model inference process when the resources of the first device are sufficient, and the second cycle being applied to the model inference process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle. The situation where the resources of the first device are sufficient mentioned here may, for example, refer to a situation where no model parallel conflict occurs, and the situation where the resources of the first device are insufficient may, for example, refer to a situation where a model parallel conflict occurs. That is to say, when a model parallel conflict occurs, the inference cycle of the model may be adjusted from the first cycle to the second cycle.

[0097] In some embodiments, the fourth information may indicate information about the second period in a variety of ways. For example, the fourth information may directly indicate the second period, such as indicating that the second period is 5s. For another example, the fourth information may indicate an association relationship between the second period and the first period, such as indicating a multiple relationship between the second period and the first period, or referred to as a period multiple. As an example, the fourth information may indicate that the second period is a multiple of the first period by 2 (i.e., the period multiple is 2). If the first period is 2s, then the second period is 4s. For another example, the fourth information may indicate the second period through a fixed list or sequence. As an example, the model period sequence may be an incremental sequence, such as {2s, 4s, 6s}. When the model reasoning period is adjusted for the first time, the second period is 2s. When the model reasoning period is adjusted for the second time, the second period is 4s. When the model reasoning period is adjusted for the third time, the second period is 6s.

[0098] In an embodiment of the present application, when a model parallel conflict occurs, adjusting the model's reasoning cycle, such as increasing the model's reasoning cycle (i.e., the second cycle is greater than the first cycle), can reduce the model's demand for resources per unit time, thereby helping to resolve the problem of model parallel conflict.

[0099] The aforementioned adjustment of the model inference cycle within the first duration may refer to the adjusted inference cycle (e.g., the second cycle) being applied for the first duration. In other words, the model inference is performed using the second cycle within the first duration, and the model inference is performed using the first cycle after the first duration.

[0100] In order to avoid the situation where the model continuously backs off and cannot be reasoned / trained, the fourth information may include information related to the fairness of the model.

[0101] In some embodiments, the information associated with the fairness of the model may include one or more of the following information: information indicating an initial value of the model fairness; information indicating an increment of the model fairness; and information indicating a termination condition of the model fairness.

[0102] The initial value of the model fairness can be, for example, a fixed value or a fixed level. As an example, the initial value of the model fairness can be determined based on model priority information, such as the initial value of the model fairness being equal to the model priority. Alternatively, the initial value of the model fairness can be an element at a specific position in a model fairness set, such as the first element. For example, if the model fairness set is {0, 1, 3, 5}, the initial value of the model fairness is the first element in the model fairness set, 0.

[0103] The increment of model fairness indicates the amount of increase in model fairness after each model backoff. The increment of model fairness can be the same or different each time a model backoff occurs. For example, the increment of model fairness is 1 each time a model backoff occurs. In another example, the increment of model fairness can be in the form of an increment list or sequence, and for different numbers of model backoffs, the increment of model fairness corresponds to different elements in the increment list or sequence.

[0104] In some embodiments, the model fairness increment may be an incremental level, or a list of incremental levels (which will be described later with reference to specific examples and will not be repeated here).

[0105] In some embodiments, the above-mentioned model fairness can be determined based on the initial value of the model fairness and the increment of the model fairness. If the increment of the model fairness is the same each time the model backoff occurs, then the model fairness = the initial value of the model fairness + the increment of the model fairness × the number of model backoffs. Taking the initial value of the model fairness as 2.1 and the increment of the model fairness as 1 as an example, if the number of model backoffs is 1, the model fairness is 3.1, and if the number of model backoffs is 2, the model fairness is 4.1. Still taking the initial value of the model fairness as 2.1 as an example, if the increment of the model fairness is an incremental list, such as {1,2,3}, when the number of model backoffs is 1, the model fairness is 3.1 (that is, the increment of the model fairness is the first element 1 in the incremental list), and when the number of model backoffs is 2, the model fairness is 5.1 (that is, the increment of the model fairness is the second element 2 in the incremental list).

[0106] As mentioned above, the model fairness increment can be an incremental level, or an incremental level list. In this case, the fourth information can indicate the fairness of the model each time a model backoff occurs through the model fairness set and / or the model fairness incremental level (or incremental level list). For example, the model fairness set is {0, 1, 3, 5}, and the initial value of the model fairness is the nth element in the model fairness set (taking n = 1 as an example). If the incremental level of the model fairness increment is m (taking m = 1 as an example), then when the number of model backoffs is 1, the model fairness is 1 (i.e., the n+mth element in the model fairness set); when the number of model backoffs is 2, the model fairness is 3 (i.e., the n+m+mth element in the model fairness set). For example, the model fairness set is {0, 1, 3, 5}, the initial value of the model fairness is the nth element in the model fairness set (taking n=1 as an example), and the incremental level list is {1, 0, 2}. Then when the model backoff number is 1, the model fairness is 1 (that is, the n+1th element in the model fairness set, where 1 is the first element in the incremental level list); when the model backoff number is 2, the model fairness is 1 (that is, the n+1+0th element in the model fairness set, where 0 is the second element in the incremental level list); when the model backoff number is 3, the model fairness is 5 (that is, the n+1+0+2th element in the model fairness set, where 2 is the third element in the incremental level list).

[0107] In some embodiments, if the model fairness of the first model satisfies the termination condition for model fairness, the first device executes inference / training of the first model or deactivates the first model. If the initial value of the model fairness of the first model is 0, the model fairness increment is 1, and the termination condition for model fairness is 4, then when the model fairness is 4, the first device executes inference / training of the first model or deactivates the first model. In other words, when a model parallelism conflict occurs, model backoff is performed at most four times.

[0108] It should be noted that at regular intervals or when preset conditions are met, the above-mentioned information such as the number of model parallel conflicts, the number of model backoffs, the number of adjusted inference cycles, and the model fairness increment will be cleared or reset to the initial value to avoid the impact of historical information of model parallel conflicts on the model management decision for the current model parallel conflict, thereby helping to improve the accuracy of model management.

[0109] It should be noted that the fourth information may include one or more of the above information, that is, the fourth information may include the maximum number of times; the first duration; model reasoning cycle information; and one or more of the information associated with the fairness of the model.

[0110] The fifth information is associated with model deactivation

[0111] In some embodiments, the first information may include fifth information associated with model deactivation. The fifth information may include, for example, one or more of: information indicating whether model deactivation is supported; a second duration indicating the duration of model deactivation; and a first timer indicating the moment of model deactivation.

[0112] In some embodiments, the fifth information may include information on whether model deactivation is supported. The model deactivation information may, for example, indicate whether model deactivation is supported by one or more bits. For example, "0" may be used to indicate that the model does not support model deactivation, while "1" may be used to indicate that the model supports model deactivation. Conversely, "1" may be used to indicate that the model does not support model deactivation, while "0" may be used to indicate that the model supports model deactivation. The meanings of the above bits are given for illustrative purposes only and are not limited in this application.

[0113] In some embodiments, the fifth information may include a second duration to indicate the duration of model deactivation. For example, the second duration may be 1 minute. In another example, the second duration may be a fixed list or an incremental sequence. As an example, the second duration may be {1 min, 2 min, 3 min}. When the first model is deactivated for the first time, the second duration may be 1 minute, that is, the deactivation duration of the first model is 1 minute. When the first model is deactivated for the third time, the second duration may be 3 minutes, that is, the deactivation duration of the first model is 3 minutes.

[0114] If the deactivation time of the first model reaches the second time, the first device may reactivate the first model. In some embodiments, the first device performs an activation judgment before reactivating the first model. The activation judgment mentioned here may refer to determining whether there is a model parallel conflict at the current moment. If there is no model parallel conflict at present, the first model may be reactivated. If there is a model parallel conflict at present, the first model may not be reactivated, or the first model may remain deactivated to avoid invalid activation and aggravation of the model parallel conflict. In this case, in order to ensure the performance gain of the first model to the system, the activation judgment may be performed again after a certain interval.

[0115] In some embodiments, the fifth information may include a first timer to indicate the time when the model is deactivated. For example, the model is deactivated when the first timer expires. The duration of the first timer may be determined based on factors such as the model type and usage requirements.

[0116] The sixth information is used to indicate the inference order of the model

[0117] In the process of completing a certain task, the first device may need to process multiple models serially. The serial processing mentioned here may, for example, refer to the reasoning result of the first model as the input of the second model. Therefore, the management of the model needs to take into account the reasoning order of the model, that is, the first information may include sixth information for indicating the reasoning order of the model. In other words, the sixth information may be used to indicate the reasoning order relationship of multiple models, such as the sixth information may indicate that the reasoning of the first model is executed first, and then the reasoning of the second model is executed, or the reasoning of the first model is executed before the reasoning of the second model is executed. Among them, multiple models, such as the first model and the second model, are models that need to be processed serially.

[0118] In some embodiments, the inference order of the models may be indicated by one or more bits. For example, 001 may be used to indicate the first model processed in series, 010 may indicate the second model processed in series, and 100 may indicate the third model processed in series.

[0119] In some embodiments, the order of model inference can be indicated by a model identifier list. For example, the model identifier list can be {2, 1, 3}, indicating that the first model processed in series is model 2, the second model processed in series is model 1, and the third model processed in series is model 3. In other words, the order of model inference is model 2, model 1, and model 3.

[0120] The seventh information is used to indicate the performance gain information of the model

[0121] For models with small performance gains, pausing the model's inference / training, adjusting the model's inference cycle, or deactivating the model may have a minor impact on system performance. However, for models with large performance gains, pausing the model's inference / training, adjusting the model's inference cycle, or deactivating the model may have a significant impact on system performance. Therefore, when managing models, it is necessary to consider the model's performance gain to ensure or improve system performance.

[0122] Based on this, the first information may further include seventh information indicating the performance gain of the model. That is, the seventh information may be used to indicate the communication performance gain of communication based on the model compared to communication without the model. In some embodiments, the seventh information may indicate the performance gain of the model by an absolute value of the model performance gain, a percentage of the model performance gain to the performance of communication without the model, or a performance multiple.

[0123] In some embodiments, when the model is managed based on the seventh information, if a model parallel conflict occurs, the model backoff described above can be performed preferentially on the model with smaller performance gain, such as pausing the model's reasoning / training and / or adjusting the model's reasoning cycle, or deactivating the model with smaller performance gain; giving priority to maintaining the operation of the model with larger performance gain, or giving priority to maintaining the reasoning / training of the model with larger performance gain.

[0124] The eighth information is used to indicate the model that the first device must reason / train

[0125] In some embodiments, the first information may include eighth information indicating a model that the first device must infer / train.

[0126] The aforementioned models that must be inferred / trained may, for example, include models of high importance, models that need to be trained as quickly as possible, or models that require continuous inference. For example, when the first device executes a handover decision, the model associated with the handover decision is a model that must be inferred / trained to ensure the communication performance of the system.

[0127] When managing models, the first device must ensure the operation of models that require inference / training. For example, if the models that the first device requires inference / training include the first model, the first device will not activate the first model, will not suspend inference / training of the first model, and / or will not adjust the inference cycle of the first model. For models that do not require inference / training, the first device can manage them based on other information.

[0128] In some embodiments, the models that must be inferred / trained can be indicated by a required model identifier list. For example, the required model identifier list can be {1, 3, 7}, indicating that Model 1, Model 3, and Model 7 are the models that must be inferred / trained by the first device.

[0129] Ninth information, associated with the replacement model

[0130] In some embodiments, the first information may include ninth information associated with a replacement model. The replacement model may be associated with the first model. For example, the replacement model may be a model that has lower requirements on the capabilities of the first device, such as less resource usage, and has the same functions as the first model. When a model parallel conflict occurs, replacing the first model with the replacement model (i.e., deactivating the first model and activating the replacement model) helps resolve the problem of model parallel conflict. The resources mentioned here may include computing resources, energy consumption resources, running memory resources, etc.

[0131] In this way, when multiple models associated with the first device conflict with parallel processing, model replacement can increase the number of models that the first device can process in parallel, thereby helping to resolve the model parallel conflict. For example, if the first device has 30MB of RAM and each model's inference / training requires 10MB of RAM, the first device can process three models in parallel. If the model is replaced with a model that requires 5MB of RAM for inference / training, the number of models that the first device can process in parallel increases to six.

[0132] In some embodiments, the ninth information may include one or more of the following information: information on whether the replacement model is supported; model information of the replacement model; a method for obtaining the replacement model; and a timing for using the replacement model.

[0133] For example, the ninth information may include information on whether a replacement model is supported. Although the replacement model consumes fewer resources than the original model, the performance of the replacement model may also be reduced to a certain extent. Therefore, some models may support the replacement model, but some models may not support the replacement model. In the embodiment of the present application, model management based on information on whether the replacement model is supported helps meet the performance requirements of the model.

[0134] For another example, the ninth information may include model information of the replacement model, such as a model identifier.

[0135] If the first device or the second device locally includes a replacement model, the first device may perform model replacement based on the model information of the replacement model, or receive the replacement model from the second device (or transmit the replacement model from the second device to the first device), thereby performing model replacement. The second device mentioned here may be a device that communicates with the first device.

[0136] If the first device and / or the second device does not locally include the replacement model, the original model may be processed to obtain the replacement model. Therefore, the ninth information may include a method for obtaining the replacement model, such as a model compression method such as knowledge distillation or network pruning.

[0137] It should be noted that the first device can process the original model to obtain the replacement model, or the second device can process the original model and transmit the obtained replacement model to the first device. The device that processes the original model can be determined based on the resource capabilities of the device, such as a device with sufficient resources to process the original model.

[0138] For another example, the ninth information may include a timing for using the replacement model. For example, the timing for using the replacement model may include time T1, after time T2, or after the expiration of the second timer. In other words, the ninth information may include one or more of time T1, time T2, and the second timer.

[0139] It should be noted that the second to ninth information can be used independently or in combination. For example, when determining whether to deactivate the first model, the performance gain of the first model can be considered. For another example, when a model conflict occurs and model backoff is required, whether to perform model backoff on the first model can be determined based on the priority of multiple models or the order in which the models are inferred.

[0140] In an embodiment of the present application, the first device manages the model associated with the first device based on the first information, such as deactivating the model, pausing model reasoning / training, and adjusting the reasoning cycle of the model, so as to alleviate the contradiction between insufficient resources of the first device and large demand for model resources, thereby helping to achieve the management of the model associated with the first device.

[0141] The first information is introduced in detail above. The configuration method and configuration or transmission timing of the first information will be introduced below.

[0142] In some embodiments, the first device may obtain the first information in one or more of the following ways: obtaining predefined first information; obtaining preconfigured first information; and obtaining dynamic indication information sent by the second device, wherein the dynamic indication information includes the first information.

[0143] For example, the first information may be predefined information, or the first information is carried in predefined information, such as protocol predefined information. For example, the protocol may predefine model priority information, such as model priority information based on service type.

[0144] For another example, the first information may be pre-configured information, or in other words, the first information is carried in pre-configured information, such as information pre-configured by the network device for the terminal device. For example, the first information may be carried in an RRC message.

[0145] For another example, the first information may be information carried in dynamic indication information sent by the second device. That is, the first device receives dynamic indication information sent by the second device, and the first information is carried in that information. In this embodiment of the present application, the dynamic indication first information can dynamically adjust the model management policy based on business needs and terminal device status.

[0146] In some embodiments, the first information associated with different models may be carried in one piece of information or in multiple pieces of information; the first information associated with the first model may be configured in more than one manner or in more than one piece of information.

[0147] It should be noted that the first information can be configured in one or more of the above-mentioned ways. For example, the first information can be configured in a predefined, preconfigured, and dynamic indication way. For another example, the first information can be configured in a preconfigured and dynamic indication way.

[0148] In the case where there are multiple first information, the priority of the first information satisfies one or more of the following: the priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the preconfigured first information; the priority of the predefined first information is the highest; the priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; wherein, the difference between the current moment and the second moment is smaller than the difference between the current moment and the third moment.

[0149] If the first information associated with the first model is configured by way of pre-configuration and dynamic indication, the priority of the first information carried in the dynamic indication information is higher than the priority of the first information carried in the pre-configuration information.

[0150] If the first information associated with the first model is configured by predefinition or other means, the first information carried in the predefined information has the highest priority.

[0151] If the first information associated with the first model is configured in a single manner and there are multiple first information pieces, the priority of the first information pieces is determined based on the time they are received / decoded. For example, the priority of the first information received at the second time is higher than the priority of the first information received at the third time; where the difference between the current time and the second time is less than the difference between the current time and the third time. In other words, the priority of the first information received closest to the current time is the highest, or in other words, the first information received or decoded most recently is the highest.

[0152] In some embodiments, if the configuration method of the first information includes pre-configuration and dynamic indication, the first information is configured when the first device receives the model transmitted by the second device, or the first information is configured when the trigger condition is met; wherein the trigger condition includes the first device reporting the device status information of the first device to the second device, and / or updating the model configuration information of the second device.

[0153] Among them, if the first information is pre-configured information, the first information can be configured (such as initial configuration and / or reconfiguration) when the first device receives the model transmitted by the second device, and the first information can be reconfigured (or updated) when the first device reports the device status information of the first device to the second device; and the first information can be reconfigured in response to the update of the model configuration information of the second device, such as the update of the model configuration information of the second device after cell switching. It should be noted that the reporting method of the device status information of the first device includes periodic reporting, and the reporting of the device status information mentioned here can refer to periodic reporting.

[0154] If the first information is dynamic indication information, the triggering condition for transmitting the first information includes the first device reporting the device status information of the first device to the second device. It should be noted that the reporting method of the device status information of the first device may also include triggered reporting, and the reporting of device status information mentioned here may refer to triggered reporting. The triggering condition for triggered reporting may include the occurrence of a model parallel conflict. That is, when a model parallel conflict occurs, the first device reports the device status information of the first device to the second device.

[0155] In some embodiments, the device status information of the first device may include one or more of capability information of the first device, information associated with handling of model parallel conflicts of the first device, and wireless environment information of the model.

[0156] In some embodiments, the device status information of the first device may include part or all of the information in the first information, such as any one or more of the contents of the first information mentioned above, such as model priority information, model reasoning order information, etc.

[0157] In some embodiments, the device status information of the first device may include status information of the first model, such as whether the first model is deactivated, the time when the first model is reactivated, whether the reasoning / training of the first model is suspended, whether the reasoning cycle of the first model is adjusted, the reasoning cycle of the first model, the number of model backoffs of the first model (which may include the number of times the first model suspends reasoning and / or the number of times the model reasoning cycle is adjusted), the backoff duration of the first model, the fairness information of the first model, and the model identifier of the model stored locally on the first device, etc.

[0158] In some embodiments, the device status information of the first device may include whether the first device has a capability for accelerating model reasoning. Whether the first device has a capability for accelerating model reasoning may, for example, refer to whether the first device has dedicated hardware (such as a graphics processing unit (GPU) and / or an application specific integrated circuit (ASIC)) that can be used to accelerate model reasoning.

[0159] In some embodiments, the device status information of the first device may include storage space available for model inference in the first device. The storage space may, for example, refer to the runtime memory space of the first device. The larger the runtime memory space, the more models that can be inferred / trained by the first device, or the larger the models that can be inferred / trained. If part of the runtime memory space allocated to model storage by the first device is occupied by the model, the remaining runtime memory space can be used by the new model. Therefore, the storage space available for model inference in the first device includes the remaining runtime memory space that can be used for model inference.

[0160] In some embodiments, the device state information of the first device may include the computing power and / or remaining computing power available for model inference in the first device. Generally, the stronger the computing power of the first device, the more models that can be inferred / trained, or the larger the models that can be inferred / trained.

[0161] In some embodiments, the device status information of the first device may include the amount of power available for model inference in the first device.

[0162] In some embodiments, the device status information of the first device may include information associated with models stored on the first device, such as identifiers (e.g., model IDs) and / or quantities of models stored on the first device. For example, a model identifier corresponds to a model of a specific function, which is used to determine whether an existing model structure can be reused.

[0163] In some embodiments, the device status information of the first device may include performance requirements for a model associated with the first device, such as a requirement for a positioning error of no more than 0.5 m in a positioning service. If the first model is deactivated or a replacement model for the first model fails to meet the model performance requirements, the first model may not be deactivated or the replacement model may not be adopted in model management.

[0164] In some embodiments, the device status information of the first device may include whether the first device supports model compression or whether the first device supports the use of a replacement model. As mentioned above, the use of a replacement model may result in a certain performance degradation. If the first device does not support model compression or does not support the use of a replacement model, then the model compression or model replacement operation may not be performed on the first model during the model management process.

[0165] In some embodiments, the device status information of the first device may include the location information of the first device. As mentioned above, the application scenarios of a single model are relatively limited. Therefore, the first device may need to use different models in different locations. In this way, the model management strategy can be dynamically determined based on the location information of the first device.

[0166] The location information of the first device can be used to indicate the location of the first device, such as an absolute location (such as coordinates) or a relative location (such as a relative location to a network device). It can also be used to indicate the movement of the first device, such as by indicating the movement of the first device through multiple pieces of location information. For example, the location information of the first device may include one or more of the following: absolute location information of the first device at a first moment; relative location information of the first device at a first moment; information about one or more locations of the first device within a second time window before the first moment; information about N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and movement speed of the first device, where N is a positive integer.

[0167] In some embodiments, the device status information of the first device may include data arrival information for the first device. For example, the data arrival information for the first device includes one or more of the following: the amount of data arriving at a first moment; the amount and / or arrival time of data arriving at M consecutive sampling points before the first moment; and the amount and / or arrival time of data arriving within a fourth time window before the first moment; where M is a positive integer. In this way, the first device can manage the first model based on information such as the data to be processed or the changing trend of the amount of data to be processed, thereby helping to improve the adaptability of the model management strategy to the data processing business of the first device.

[0168] In some embodiments, the device status information of the first device may include information associated with the serving cell and / or information associated with the neighboring cell. In the scenario of cell mobility management, when the first device accesses a different serving cell, the model configuration of the first device may be updated (such as including an identifier of a model that the first device can use). In this way, the model management strategy can be dynamically determined based on the information associated with the serving cell and / or the information associated with the neighboring cell of the first device.

[0169] The information associated with the serving cell and / or neighboring cell may include a cell identifier and / or a cell signal quality. The cell identifier may include, for example, a physical cell identifier (PCI). The cell signal quality may include reference signal receiving power (RSRP), reference signal receiving quality (RSRQ), signal-to-interference-plus-noise ratio (SINR), etc.

[0170] For example, the signal quality of a cell includes one or more of the following: signal quality at a first moment; signal quality at P consecutive sampling points before the first moment; and signal quality within a first time window before the first moment; where P is a positive integer.

[0171] In some embodiments, the device status information of the first device may include information related to a buffer status report (BSR) of the first device. For example, the information related to the BSR of the first device includes one or more of the following: a BSR at a first moment; BSRs at R consecutive sampling points before the first moment; and a BSR within a fifth time window before the first moment, where R is a positive integer.

[0172] It should be noted that the first moment mentioned above can be the current moment and / or a historical moment. For example, the location information of the first device includes the absolute location information of the first device at the first moment, which can mean that the location information of the first device includes the absolute location information of the first device at the current moment and / or a historical moment. For another example, the cell signal quality includes the signal quality at the first moment, which can mean that the cell signal quality includes the cell signal quality at the current moment and / or the cell signal quality at a historical moment.

[0173] It should be noted that the method provided in the embodiment of the present application can be applied to the model management of the first device, such as the model management when a model parallel conflict occurs on the first device. For example, in response to the model parallel conflict on the first device, the first device manages the first model based on the first information.

[0174] The method for wireless communication provided by the embodiment of the present application has good backward compatibility and a wide range of application scenarios. With the evolution of technology, the resources of the first device required for the model may become increasingly large, and the hardware capabilities of the old terminal devices may become even more stretched. The method provided by the embodiment of the present application can enable old terminal devices or devices with lower capabilities to still have the possibility of applying multiple models. In addition, the method provided by the embodiment of the present application can enable all types of existing devices to run multiple models, with a huge potential user group and market.

[0175] It should be noted that the reasoning / training of the model mentioned in the embodiments of the present application may refer to the reasoning of the model and / or the training of the model.

[0176] The method embodiment of the present application is described in detail above in conjunction with Figures 1 to 5 , and the device embodiment of the present application is described in detail below in conjunction with Figures 6 to 8 . It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, and therefore, for portions not described in detail, reference can be made to the above method embodiment.

[0177] FIG6 is a schematic diagram of a device for wireless communication according to an embodiment of the present application, and the device 600 may be the first device described above.

[0178] 6 , a device 600 may include an acquiring unit 610 and a managing unit 620 .

[0179] The acquisition unit 610 is configured to acquire first information, where the first information is used to manage the model associated with the first device, and the model associated with the first device includes a first model.

[0180] The management unit 620 is configured to manage the first model based on the first information.

[0181] In some embodiments, the first information includes one or more of the following: second information for indicating a model identifier; third information associated with a model priority; fourth information associated with a pause in model inference / training and / or an adjustment of a model's inference cycle; fifth information associated with model deactivation; sixth information for indicating a model's inference order; seventh information for indicating a gain in communication performance based on a model compared to communication without a model; eighth information for indicating a model that the first device must infer / train; and ninth information associated with a replacement model that is related to the first model.

[0182] In some embodiments, the third information includes one or more of the following: information on whether configuration of model priority is supported; information on whether a priority processing function is enabled; and model priority information.

[0183] In some embodiments, the model priority information is related to the second information and / or a business type associated with the model.

[0184] In some embodiments, the fourth information includes one or more of the following: a maximum number of times, used to indicate the maximum value of the model inference / training backoff number, the model inference / training backoff number is the number of times the model inference / training is suspended and / or the model inference cycle is adjusted; a first duration, used to indicate the backoff duration of the model inference and / or training, the backoff duration is the duration of suspending the model inference / training and / or adjusting the model inference cycle within the backoff duration; model inference cycle information; and information associated with the fairness of the model.

[0185] In some embodiments, the management unit is configured to: if the model inference / training backoff number of the first model is greater than or equal to the maximum number, the first device deactivating the first model.

[0186] In some embodiments, the first duration is related to the number of model inference / training backoffs.

[0187] In some embodiments, the model inference cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model inference process when the resources of the first device are sufficient, and the second cycle is applied to the model inference process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

[0188] In some embodiments, the information associated with the fairness of the model includes one or more of the following information: information indicating an initial value of the fairness of the model; information indicating an increment of the fairness of the model; and information indicating a termination condition of the fairness of the model.

[0189] In some embodiments, the management unit is configured to: if the model fairness of the first model satisfies the termination condition, the first device performs inference / training of the first model or deactivates the first model.

[0190] In some embodiments, the model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

[0191] In some embodiments, the increment of the model fairness is associated with the number of model inference / training backoffs, wherein the number of model inference / training backoffs is the number of times model inference / training is paused and / or the model inference cycle is adjusted.

[0192] In some embodiments, the initial value of the model fairness is determined based on model priority information.

[0193] In some embodiments, the fifth information includes one or more of the following: information on whether model deactivation is supported; a second duration for indicating the duration of model deactivation; and a first timer for indicating the moment of model deactivation.

[0194] In some embodiments, the sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

[0195] In some embodiments, the multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

[0196] In some embodiments, the management unit is used to: if the model that the first device must infer / train includes the first model, the first device does not deactivate the first model or suspend the inference / training of the first model, and / or the first device does not adjust the inference cycle of the first model.

[0197] In some embodiments, the ninth information includes one or more of the following information: information on whether the replacement model is supported; model information of the replacement model; a method for obtaining the replacement model; and a timing for using the replacement model.

[0198] In some embodiments, the first device obtains the first information in one or more of the following ways: obtaining predefined first information; obtaining preconfigured first information; and obtaining dynamic indication information sent by the second device, wherein the dynamic indication information includes the first information.

[0199] In some embodiments, the first information is configured when the first device receives the model transmitted by the second device, or the first information is configured when a trigger condition is met; wherein the trigger condition includes the first device reporting the device status information of the first device to the second device, and / or updating the model configuration information of the second device.

[0200] In some embodiments, the device status information of the first device includes one or more of the following: part or all of the information in the first information; status information of the first model; whether the first device has the ability to accelerate model reasoning; storage space available for model reasoning in the first device; computing power available for model reasoning in the first device; amount of power available for model reasoning in the first device; information associated with the model stored in the first device; performance requirements of the model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; information associated with the serving cell; information associated with the neighboring cell; and relevant information of the cache status report BSR of the first device.

[0201] In some embodiments, the location information of the first device includes one or more of the following: absolute location information of the first device at a first moment; relative location information of the first device at the first moment; information on one or more locations of the first device within a second time window before the first moment; information on N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and moving speed of the first device; wherein N is a positive integer.

[0202] In some embodiments, the data arrival information of the first device includes one or more of the following: the amount of data arriving at the first moment; the amount of data and / or arrival time of the arrival data at M consecutive sampling points before the first moment; and the amount of data and / or arrival time of the arrival data within a fourth time window before the first moment; where M is a positive integer.

[0203] In some embodiments, the information associated with the serving cell and / or the neighboring cell includes a cell identity and / or a cell signal quality.

[0204] In some embodiments, the signal quality of the cell includes one or more of the following: signal quality at a first moment; signal quality of P consecutive sampling points before the first moment; and signal quality within a first time window before the first moment; where P is a positive integer.

[0205] In some embodiments, the BSR-related information of the first device includes one or more of the following: the BSR at the first moment; the BSR of R consecutive sampling points before the first moment; and the BSR in the fifth time window before the first moment; where R is a positive integer.

[0206] In some embodiments, the first moment is a current moment and / or a historical moment.

[0207] In some embodiments, the reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

[0208] In some embodiments, the priority of the first information satisfies one or more of the following: the priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the preconfigured first information; the predefined first information has the highest priority; the priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; wherein, the difference between the current moment and the second moment is less than the difference between the current moment and the third moment.

[0209] In some embodiments, the management unit is configured to: in response to a model parallelism conflict of the first device, the first device manages the first model based on the first information.

[0210] FIG7 is a schematic diagram of another device for wireless communication according to an embodiment of the present application. Device 700 may be the second device described above.

[0211] 7 , a device 700 may include a sending unit 710 .

[0212] The sending unit 710 is configured to send first information to a first device, where the first information is used to manage models associated with the first device, and the models associated with the first device include a first model.

[0213] In some embodiments, the first information includes one or more of the following: second information for indicating a model identifier; third information associated with a model priority; fourth information associated with a pause in model inference / training and / or an adjustment of a model's inference cycle; fifth information associated with model deactivation; sixth information for indicating a model's inference order; seventh information for indicating a gain in communication performance based on a model compared to communication without a model; eighth information for indicating a model that the first device must infer / train; and ninth information associated with a replacement model that is related to the first model.

[0214] In some embodiments, the third information includes one or more of the following: information on whether configuration of model priority is supported; information on whether a priority processing function is enabled; and model priority information.

[0215] In some embodiments, the model priority information is related to the second information and / or a business type associated with the model.

[0216] In some embodiments, the fourth information includes one or more of the following: a maximum number of times, used to indicate the maximum value of the model inference / training backoff number, the model inference / training backoff number is the number of times the model inference / training is suspended and / or the model inference cycle is adjusted; a first duration, used to indicate the backoff duration of the model inference and / or training, the backoff duration is the duration of suspending the model inference / training and / or adjusting the model inference cycle within the backoff duration; model inference cycle information; and information associated with the fairness of the model.

[0217] In some embodiments, the first duration is related to the number of model inference / training backoffs.

[0218] In some embodiments, the model inference cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model inference process when the resources of the first device are sufficient, and the second cycle is applied to the model inference process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

[0219] In some embodiments, the information associated with the fairness of the model includes one or more of the following information: information indicating an initial value of the fairness of the model; information indicating an increment of the fairness of the model; and information indicating a termination condition of the fairness of the model.

[0220] In some embodiments, the model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

[0221] In some embodiments, the increment of the model fairness is associated with the number of model inference / training backoffs, wherein the number of model inference / training backoffs is the number of times model inference / training is paused and / or the model inference cycle is adjusted.

[0222] In some embodiments, the initial value of the model fairness is determined based on model priority information.

[0223] In some embodiments, the fifth information includes one or more of the following: information on whether model deactivation is supported; a second duration for indicating the duration of model deactivation; and a first timer for indicating the moment of model deactivation.

[0224] In some embodiments, the sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

[0225] In some embodiments, the multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

[0226] In some embodiments, the seventh information is used to indicate a gain in communication performance when communicating based on the model compared to communicating without adopting the model.

[0227] In some embodiments, the ninth information includes one or more of the following information: information on whether the replacement model is supported; model information of the replacement model; a method for obtaining the replacement model; and a timing for using the replacement model.

[0228] In some embodiments, the first information is configured when the first device receives the model transmitted by the second device, or the first information is configured when a trigger condition is met; wherein the trigger condition includes the first device reporting the device status information of the first device to the second device, and / or updating the model configuration information of the second device.

[0229] In some embodiments, the device status information of the first device includes one or more of the following: part or all of the information in the first information; status information of the first model; whether the first device has the ability to accelerate model reasoning; storage space available for model reasoning in the first device; computing power available for model reasoning in the first device; amount of power available for model reasoning in the first device; information associated with the model stored in the first device; performance requirements of the model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; information associated with the serving cell; information associated with the neighboring cell; and relevant information of the cache status report BSR of the first device.

[0230] In some embodiments, the location information of the first device includes one or more of the following: absolute location information of the first device at a first moment; relative location information of the first device at the first moment; information on one or more locations of the first device within a second time window before the first moment; information on N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and moving speed of the first device; wherein N is a positive integer.

[0231] In some embodiments, the data arrival information of the first device includes one or more of the following: the amount of data arriving at the first moment; the amount of data and / or arrival time of the arrival data at M consecutive sampling points before the first moment; and the amount of data and / or arrival time of the arrival data within a fourth time window before the first moment; where M is a positive integer.

[0232] In some embodiments, the information associated with the serving cell and / or the neighboring cell includes a cell identity and / or a cell signal quality.

[0233] In some embodiments, the signal quality of the cell includes one or more of the following: signal quality at a first moment; signal quality of P consecutive sampling points before the first moment; and signal quality within a first time window before the first moment; where P is a positive integer.

[0234] In some embodiments, the BSR-related information of the first device includes one or more of the following: the BSR at the first moment; the BSR of R consecutive sampling points before the first moment; and the BSR in the fifth time window before the first moment; where R is a positive integer.

[0235] In some embodiments, the first moment is a current moment and / or a historical moment.

[0236] In some embodiments, the reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

[0237] In some embodiments, the priority of the first information satisfies one or more of the following: the priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the preconfigured first information; the predefined first information has the highest priority; the priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; wherein, the difference between the current moment and the second moment is less than the difference between the current moment and the third moment.

[0238] In an optional embodiment, the transmitting unit and the receiving unit mentioned above may be a transceiver 830. The device 600 and the device 700 may further include a processor 810 and a memory 820, as specifically shown in FIG8 .

[0239] Figure 8 is a schematic block diagram of a communication device according to an embodiment of the present application. The dashed lines in Figure 8 indicate that the unit or module is optional. The device 800 may be used to implement the method described in the above method embodiment. The device 800 may be a chip, a first device, or a second device.

[0240] The device 800 may include one or more processors 810. The processor 810 may support the device 800 to implement the method described in the method embodiment above. The processor 810 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0241] The apparatus 800 may further include one or more memories 820. The memories 820 store programs that can be executed by the processor 810, causing the processor 810 to perform the methods described in the above method embodiments. The memories 820 may be independent of the processor 810 or integrated into the processor 810.

[0242] The apparatus 800 may further include a transceiver 830. The processor 810 may communicate with other devices or chips via the transceiver 830. For example, the processor 810 may transmit and receive data with other devices or chips via the transceiver 830.

[0243] The present application also provides a computer-readable storage medium for storing a program. The computer-readable storage medium can be applied to the first device or the second device provided in the present application, and the program causes a computer to execute the method performed by the device for wireless communication in each embodiment of the present application.

[0244] The present application also provides a computer program product. The computer program product includes a program. The computer program product can be applied to the first device or the second device provided in the present application, and the program causes a computer to execute the method performed by the device for wireless communication in each embodiment of the present application.

[0245] The embodiments of the present application also provide a computer program. The computer program can be applied to the wireless communication device provided in the embodiments of the present application, and the computer program enables a computer to execute the method performed by the wireless communication device in each embodiment of the present application.

[0246] It should be understood that the terms "system" and "network" in this application can be used interchangeably. In addition, the terms used in this application are only used to explain the specific embodiments of this application and are not intended to limit this application. The terms "first", "second", "third", and "fourth" in the specification and claims of this application and the accompanying drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0247] In the embodiments of this application, the term "indication" may refer to a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" may refer to a direct indication of B, e.g., B can obtain information through A; it may refer to an indirect indication of B, e.g., A indicates C, e.g., B can obtain information through C; or it may refer to an association between A and B.

[0248] In the embodiment of the present application, "B corresponding to A" means that B is associated with A and B can be determined based on A. However, it should be understood that determining B based on A does not mean determining B based solely on A, but B can also be determined based on A and / or other information.

[0249] In the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and indication, configuration and configuration, etc.

[0250] In the embodiments of the present application, "pre-definition" or "pre-configuration" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., a terminal device and a network device). The present application does not limit the specific implementation method. For example, pre-definition may refer to information defined in a protocol.

[0251] In the embodiments of the present application, the “protocol” may refer to a standard protocol in the communications field, for example, it may include an LTE protocol, an NR protocol, and related protocols used in future communication systems, and the present application does not limit this.

[0252] In the embodiments of this application, the term "and / or" is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0253] In various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0254] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0255] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0256] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0257] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. 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 computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be read by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0258] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for wireless communication, characterized in that: include: The first device acquires first information, where the first information is used to manage models associated with the first device, and the models associated with the first device include a first model; The first device manages the first model based on the first information.

2. The method according to claim 1, characterized in that The first information includes one or more of the following: The second information is used to indicate a model identifier; The third information is associated with the model priority; Fourth information, associated with the suspension of model reasoning / training and / or the adjustment of the reasoning cycle of the model; The fifth information is associated with model deactivation; The sixth information is used to indicate the reasoning order of the model; seventh information, used to indicate a gain in communication performance of communication based on the model compared to communication without the model; Eighth information, used to indicate a model that must be inferred / trained by the first device; as well as Ninth information is associated with a replacement model, and the replacement model is related to the first model.

3. The method according to claim 2, characterized in that The third information includes one or more of the following: Whether to support configuration model priority information; Information on whether the priority processing function is enabled; and Model priority information.

4. The method according to claim 3, characterized in that The model priority information is related to the second information and / or the business type associated with the model.

5. The method according to any one of claims 2 to 4, characterized in that: The fourth information includes one or more of the following: Maximum number of times, used to indicate the maximum value of the model reasoning / training backoff number, where the model reasoning / training backoff number is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted; A first duration is used to indicate a backoff duration for model reasoning and / or training, where the backoff duration is a duration for suspending model reasoning / training and / or adjusting a model reasoning cycle within the backoff duration; Model reasoning cycle information; as well as Information associated with the fairness of the model.

6. The method according to claim 5, characterized in that The first device manages the first model based on the first information, including: If the model inference / training backoff number of the first model is greater than or equal to the maximum number, the first device deactivates the first model.

7. The method according to claim 5 or 6, characterized in that: The first duration is related to the model inference / training backoff times.

8. The method according to any one of claims 5 to 7, characterized in that: The model reasoning cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model reasoning process when the resources of the first device are sufficient, and the second cycle is applied to the model reasoning process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

9. The method according to any one of claims 5 to 8, characterized in that: The information associated with the fairness of the model includes one or more of the following information: Information indicating the initial value of the model's fairness; Information indicating an increase in model fairness; and Information indicating the termination criteria for model fairness.

10. The method according to claim 9, characterized in that The first device manages the first model based on the first information, including: If the model fairness of the first model satisfies the termination condition, the first device performs inference / training of the first model or deactivates the first model.

11. The method according to claim 9 or 10, characterized in that: The model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

12. The method according to any one of claims 9 to 11, characterized in that: The increment of the model fairness is associated with the number of model reasoning / training backoffs, wherein the number of model reasoning / training backoffs is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted.

13. The method according to any one of claims 9 to 12, characterized in that: The initial value of the model fairness is determined based on the model priority information.

14. The method according to any one of claims 2 to 13, characterized in that The fifth information includes one or more of the following: Information on whether model deactivation is supported; The second duration is used to indicate the duration of model deactivation; as well as The first timer is used to indicate the time when the model is deactivated.

15. The method according to any one of claims 2 to 14, characterized in that The sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

16. The method according to claim 15, characterized in that The multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

17. The method according to any one of claims 2 to 16, characterized in that The first device manages the first model based on the first information, including: If the model that the first device must infer / train includes the first model, the first device does not deactivate the first model or suspend inference / training of the first model, and / or the first device does not adjust the inference cycle of the first model.

18. The method according to any one of claims 2 to 17, characterized in that The ninth information includes one or more of the following information: Information on whether the replacement model is supported; Model information of the replacement model; a method for obtaining the replacement model; and When to use the replacement model.

19. The method according to any one of claims 1 to 18, characterized in that The first device acquires the first information in one or more of the following ways: Acquire the predefined first information; Acquire the preconfigured first information; and Dynamic indication information sent by the second device is obtained, where the dynamic indication information includes the first information.

20. The method according to claim 19, characterized in that The first information is configured when the first device receives the model transmitted by the second device, or when a trigger condition is met; The trigger condition includes the first device reporting device status information of the first device to the second device, and / or updating model configuration information of the second device.

21. The method according to claim 20, characterized in that The device status information of the first device includes one or more of the following: Part or all of the first information; state information of the first model; Whether the first device has the capability of accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; information associated with the model stored in the first device; performance requirements of a model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; Information associated with the serving cell; Information associated with neighboring cells; as well as The first device has a buffer status report (BSR) related information.

22. The method according to claim 21, characterized in that The information associated with the serving cell and / or the neighboring cell includes a cell identifier and / or a signal quality of the cell.

23. The method according to claim 22, characterized in that The signal quality of the cell includes one or more of the following: Signal quality at the first moment; The signal quality of P consecutive sampling points before the first moment; as well as signal quality in a first time window before a first moment; Wherein, P is a positive integer.

24. The method according to any one of claims 21 to 23, characterized in that The location information of the first device includes one or more of the following: Absolute position information of the first device at a first moment; relative position information of the first device at the first moment; Information about one or more locations of the first device within a second time window before the first moment; Information about N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and a moving speed of the first device; Wherein, N is a positive integer.

25. The method according to any one of claims 21 to 24, characterized in that The data arrival information of the first device includes one or more of the following: The amount of data arriving at the first moment; The data volume and / or arrival time of the arrival data of M consecutive sampling points before the first moment; and The amount of data and / or the arrival time of the data arriving in a fourth time window before the first moment; Wherein, M is a positive integer.

26. The method according to any one of claims 21 to 25, characterized in that The BSR related information of the first device includes one or more of the following: BSR at the first moment; The BSR of R consecutive sampling points before the first moment; and BSR in the fifth time window before the first moment; Wherein, R is a positive integer.

27. The method according to any one of claims 23 to 26, characterized in that The first moment is the current moment and / or a historical moment.

28. The method according to any one of claims 20 to 27, characterized in that The reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

29. The method according to any one of claims 1 to 28, characterized in that The priority of the first information satisfies one or more of the following: The priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the pre-configured first information; The predefined first information has the highest priority; The priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; The difference between the current moment and the second moment is smaller than the difference between the current moment and the third moment.

30. The method according to any one of claims 1 to 29, characterized in that The first device manages the first model based on the first information, including: In response to a model parallelism conflict of the first device, the first device manages the first model based on the first information.

31. A method for wireless communication, characterized in that: include: The second device sends first information to the first device, where the first information is used to manage models associated with the first device, and the models associated with the first device include the first model.

32. The method according to claim 31, characterized in that The first information includes one or more of the following: The second information is used to indicate a model identifier; The third information is associated with the model priority; Fourth information, associated with the suspension of model reasoning / training and / or the adjustment of the reasoning cycle of the model; The fifth information is associated with model deactivation; The sixth information is used to indicate the reasoning order of the model; seventh information, used to indicate a gain in communication performance of communication based on the model compared to communication without the model; Eighth information, used to indicate a model that must be inferred / trained by the first device; as well as Ninth information is associated with a replacement model, and the replacement model is related to the first model.

33. The method according to claim 32, characterized in that The third information includes one or more of the following: Whether to support configuration model priority information; Information on whether the priority processing function is enabled; and Model priority information.

34. The method according to claim 33, characterized in that The model priority information is related to the second information and / or the business type associated with the model.

35. The method according to any one of claims 32 to 34, characterized in that The fourth information includes one or more of the following: Maximum number of times, used to indicate the maximum value of the model reasoning / training backoff number, where the model reasoning / training backoff number is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted; A first duration is used to indicate a backoff duration for model reasoning and / or training, where the backoff duration is a duration for suspending model reasoning / training and / or adjusting a model reasoning cycle within the backoff duration; Model reasoning cycle information; as well as Information associated with the fairness of the model.

36. The method according to claim 35, characterized in that The first duration is related to the model inference / training backoff times.

37. The method according to claim 35 or 36, characterized in that The model reasoning cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model reasoning process when the resources of the first device are sufficient, and the second cycle is applied to the model reasoning process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

38. The method according to any one of claims 35 to 37, characterized in that The information associated with the fairness of the model includes one or more of the following information: Information indicating the initial value of the model's fairness; Information indicating an increase in model fairness; and Information indicating the termination criteria for model fairness.

39. The method according to claim 38, characterized in that The model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

40. The method according to claim 38 or 39, characterized in that The increment of the model fairness is associated with the number of model reasoning / training backoffs, wherein the number of model reasoning / training backoffs is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted.

41. The method according to any one of claims 38 to 40, characterized in that The initial value of the model fairness is determined based on the model priority information.

42. The method according to any one of claims 32 to 41, characterized in that The fifth information includes one or more of the following: Information on whether model deactivation is supported; The second duration is used to indicate the duration of model deactivation; as well as The first timer is used to indicate the time when the model is deactivated.

43. The method according to any one of claims 32 to 42, characterized in that The sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

44. The method according to claim 43, characterized in that The multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

45. The method according to any one of claims 32 to 44, characterized in that The ninth information includes one or more of the following information: Information on whether the replacement model is supported; Model information of the replacement model; a method for obtaining the replacement model; and When to use the replacement model.

46. ​​The method according to any one of claims 31 to 45, characterized in that The first information is configured when the first device receives the model transmitted by the second device, or when a trigger condition is met; The trigger condition includes the first device reporting device status information of the first device to the second device, and / or updating model configuration information of the second device.

47. The method according to claim 46, characterized in that The device status information of the first device includes one or more of the following: Part or all of the first information; state information of the first model; Whether the first device has the capability of accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; information associated with the model stored in the first device; performance requirements of a model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; Information associated with the serving cell; Information associated with neighboring cells; as well as The first device has a buffer status report (BSR) related information.

48. The method according to claim 47, characterized in that The information associated with the serving cell and / or the neighboring cell includes a cell identifier and / or a signal quality of the cell.

49. The method according to claim 48, characterized in that The signal quality of the cell includes one or more of the following: Signal quality at the first moment; The signal quality of P consecutive sampling points before the first moment; as well as signal quality in a first time window before a first moment; Wherein, P is a positive integer.

50. The method according to any one of claims 47 to 49, characterized in that The location information of the first device includes one or more of the following: Absolute position information of the first device at a first moment; relative position information of the first device at the first moment; Information about one or more locations of the first device within a second time window before the first moment; Information about N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and a moving speed of the first device; Wherein, N is a positive integer.

51. The method according to any one of claims 47 to 50, characterized in that The data arrival information of the first device includes one or more of the following: The amount of data arriving at the first moment; The data volume and / or arrival time of the arrival data of M consecutive sampling points before the first moment; and The amount of data and / or the arrival time of the data arriving in a fourth time window before the first moment; Wherein, M is a positive integer.

52. The method according to any one of claims 47 to 51, characterized in that The BSR related information of the first device includes one or more of the following: BSR at the first moment; The BSR of R consecutive sampling points before the first moment; and BSR in the fifth time window before the first moment; Wherein, R is a positive integer.

53. The method according to any one of claims 49 to 52, characterized in that The first moment is the current moment and / or a historical moment.

54. The method according to any one of claims 46 to 53, characterized in that The reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

55. The method according to any one of claims 31 to 54, characterized in that The priority of the first information satisfies one or more of the following: The priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the pre-configured first information; The predefined first information has the highest priority; The priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; The difference between the current moment and the second moment is smaller than the difference between the current moment and the third moment.

56. A device for wireless communication, characterized in that: The device is a first device, and the device includes: an acquiring unit, configured to acquire first information, where the first information is used to manage a model associated with the first device, where the model associated with the first device includes a first model; A management unit is used to manage the first model based on the first information.

57. The device according to claim 56, characterized in that The first information includes one or more of the following: The second information is used to indicate a model identifier; The third information is associated with the model priority; Fourth information, associated with the suspension of model reasoning / training and / or the adjustment of the reasoning cycle of the model; The fifth information is associated with model deactivation; The sixth information is used to indicate the reasoning order of the model; seventh information, used to indicate a gain in communication performance of communication based on the model compared to communication without the model; Eighth information, used to indicate a model that must be inferred / trained by the first device; as well as Ninth information is associated with a replacement model, and the replacement model is related to the first model.

58. The device according to claim 57, characterized in that The third information includes one or more of the following: Whether to support configuration model priority information; Information on whether the priority processing function is enabled; and Model priority information.

59. The device according to claim 58, characterized in that The model priority information is related to the second information and / or the business type associated with the model.

60. The apparatus according to any one of claims 57 to 59, characterized in that The fourth information includes one or more of the following: Maximum number of times, used to indicate the maximum value of the model reasoning / training backoff number, where the model reasoning / training backoff number is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted; A first duration is used to indicate a backoff duration for model reasoning and / or training, where the backoff duration is a duration for suspending model reasoning / training and / or adjusting a model reasoning cycle within the backoff duration; Model reasoning cycle information; as well as Information associated with the fairness of the model.

61. The device according to claim 60, characterized in that The management unit is used to: If the model inference / training backoff number of the first model is greater than or equal to the maximum number, the first device deactivates the first model.

62. The device according to claim 60 or 61, characterized in that The first duration is related to the model inference / training backoff times.

63. The apparatus according to any one of claims 60 to 62, characterized in that The model reasoning cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model reasoning process when the resources of the first device are sufficient, and the second cycle is applied to the model reasoning process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

64. The apparatus according to any one of claims 60 to 63, characterized in that The information associated with the fairness of the model includes one or more of the following information: Information indicating the initial value of the model's fairness; Information indicating an increase in model fairness; and Information indicating the termination criteria for model fairness.

65. The device according to claim 64, characterized in that The management unit is used to: If the model fairness of the first model satisfies the termination condition, the first device performs inference / training of the first model or deactivates the first model.

66. Apparatus according to claim 64 or 65, characterised in that The model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

67. The apparatus according to any one of claims 64-65, characterized in that The increment of the model fairness is associated with the number of model reasoning / training backoffs, wherein the number of model reasoning / training backoffs is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted.

68. The apparatus according to any one of claims 64 to 67, characterized in that The initial value of the model fairness is determined based on the model priority information.

69. The apparatus according to any one of claims 57 to 68, characterized in that The fifth information includes one or more of the following: Information on whether model deactivation is supported; The second duration is used to indicate the duration of model deactivation; as well as The first timer is used to indicate the time when the model is deactivated.

70. The apparatus according to any one of claims 57 to 69, characterized in that The sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

71. The device according to claim 70, characterized in that The multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

72. The apparatus according to any one of claims 57 to 71, characterized in that The management unit is used to: If the model that the first device must infer / train includes the first model, the first device does not deactivate the first model or suspend inference / training of the first model, and / or the first device does not adjust the inference cycle of the first model.

73. The apparatus according to any one of claims 57 to 72, characterized in that The ninth information includes one or more of the following information: Information on whether the replacement model is supported; Model information of the replacement model; a method for obtaining the replacement model; and When to use the replacement model.

74. The apparatus according to any one of claims 56 to 73, characterized in that The first device acquires the first information in one or more of the following ways: Acquire the predefined first information; Acquire the preconfigured first information; and Dynamic indication information sent by the second device is obtained, where the dynamic indication information includes the first information.

75. The device according to claim 74, characterized in that The first information is configured when the first device receives the model transmitted by the second device, or when a trigger condition is met; The trigger condition includes the first device reporting device status information of the first device to the second device, and / or updating model configuration information of the second device.

76. The device according to claim 75, characterized in that The device status information of the first device includes one or more of the following: Part or all of the first information; state information of the first model; Whether the first device has the capability of accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; information associated with the model stored in the first device; performance requirements of a model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; Information associated with the serving cell; Information associated with neighboring cells; as well as The first device has a buffer status report (BSR) related information.

77. The device according to claim 76, characterized in that The information associated with the serving cell and / or the neighboring cell includes a cell identifier and / or a signal quality of the cell.

78. The device according to claim 77, characterized in that The signal quality of the cell includes one or more of the following: Signal quality at the first moment; The signal quality of P consecutive sampling points before the first moment; as well as signal quality in a first time window before a first moment; Wherein, P is a positive integer.

79. The apparatus according to any one of claims 76 to 78, characterized in that The location information of the first device includes one or more of the following: Absolute position information of the first device at a first moment; relative position information of the first device at the first moment; Information about one or more locations of the first device within a second time window before the first moment; Information about N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and a moving speed of the first device; Wherein, N is a positive integer.

80. The apparatus according to any one of claims 76 to 79, characterized in that The data arrival information of the first device includes one or more of the following: The amount of data arriving at the first moment; The data volume and / or arrival time of the arrival data of M consecutive sampling points before the first moment; and The amount of data and / or the arrival time of the data arriving in a fourth time window before the first moment; Wherein, M is a positive integer.

81. The apparatus according to any one of claims 76 to 80, characterized in that The BSR related information of the first device includes one or more of the following: BSR at the first moment; The BSR of R consecutive sampling points before the first moment; and BSR in the fifth time window before the first moment; Wherein, R is a positive integer.

82. The apparatus according to any one of claims 78 to 81, characterized in that The first moment is the current moment and / or a historical moment.

83. The apparatus according to any one of claims 75 to 82, characterized in that The reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

84. The apparatus according to any one of claims 56 to 83, characterized in that The priority of the first information satisfies one or more of the following: The priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the pre-configured first information; The predefined first information has the highest priority; The priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; The difference between the current moment and the second moment is smaller than the difference between the current moment and the third moment.

85. The apparatus according to any one of claims 56 to 84, characterized in that The management unit is used to: In response to a model parallelism conflict of the first device, the first device manages the first model based on the first information.

86. A device for wireless communication, characterized in that: The device is a second device, and the device includes: A sending unit is used to send first information to a first device, where the first information is used to manage a model associated with the first device, and the model associated with the first device includes a first model.

87. The device according to claim 86, characterized in that The first information includes one or more of the following: The second information is used to indicate a model identifier; The third information is associated with the model priority; Fourth information, associated with the suspension of model reasoning / training and / or the adjustment of the reasoning cycle of the model; The fifth information is associated with model deactivation; The sixth information is used to indicate the reasoning order of the model; seventh information, used to indicate a gain in communication performance of communication based on the model compared to communication without the model; Eighth information, used to indicate a model that must be inferred / trained by the first device; as well as Ninth information is associated with a replacement model, and the replacement model is related to the first model.

88. The device according to claim 87, characterized in that The third information includes one or more of the following: Whether to support configuration model priority information; Information on whether the priority processing function is enabled; and Model priority information.

89. The device according to claim 88, characterized in that The model priority information is related to the second information and / or the business type associated with the model.

90. The apparatus according to any one of claims 87 to 89, characterized in that The fourth information includes one or more of the following: Maximum number of times, used to indicate the maximum value of the model reasoning / training backoff number, where the model reasoning / training backoff number is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted; A first duration is used to indicate a backoff duration for model reasoning and / or training, where the backoff duration is a duration for suspending model reasoning / training and / or adjusting a model reasoning cycle within the backoff duration; Model reasoning cycle information; as well as Information associated with the fairness of the model.

91. The device according to claim 90, characterized in that The first duration is related to the model inference / training backoff times.

92. The device according to claim 90 or 91, characterized in that The model reasoning cycle information includes a first cycle and / or a second cycle, wherein the first cycle is applied to the model reasoning process when the resources of the first device are sufficient, and the second cycle is applied to the model reasoning process when the resources of the first device are insufficient, wherein the first cycle is different from the second cycle.

93. The apparatus according to any one of claims 90 to 92, characterized in that The information associated with the fairness of the model includes one or more of the following information: Information indicating the initial value of the model's fairness; Information indicating an increase in model fairness; and Information indicating the termination criteria for model fairness.

94. The device according to claim 93, characterized in that The model fairness is determined based on an initial value of the model fairness and an increment of the model fairness.

95. The apparatus according to claim 93 or 94, characterized in that The increment of the model fairness is associated with the number of model reasoning / training backoffs, wherein the number of model reasoning / training backoffs is the number of times the model reasoning / training is suspended and / or the model reasoning cycle is adjusted.

96. The apparatus according to any one of claims 93 to 95, characterized in that The initial value of the model fairness is determined based on the model priority information.

97. The apparatus according to any one of claims 87 to 96, characterized in that The fifth information includes one or more of the following: Information on whether model deactivation is supported; The second duration is used to indicate the duration of model deactivation; as well as The first timer is used to indicate the time when the model is deactivated.

98. The apparatus according to any one of claims 87 to 97, characterized in that The sixth information is used to indicate the reasoning order relationship of multiple models, wherein the multiple models are models that need to be processed serially.

99. The device according to claim 98, characterized in that The multiple models include the first model and the second model. If the inference result of the first model is used as the input of the second model, the sixth information is used to indicate that the first model should be inferred before the second model is inferred.

100. The apparatus according to any one of claims 87 to 99, characterized in that The ninth information includes one or more of the following information: Information on whether the replacement model is supported; Model information of the replacement model; a method for obtaining the replacement model; and When to use the replacement model.

101. The apparatus according to any one of claims 87 to 100, characterized in that The first information is configured when the first device receives the model transmitted by the second device, or when a trigger condition is met; The trigger condition includes the first device reporting device status information of the first device to the second device, and / or updating model configuration information of the second device.

102. The device according to claim 101, characterized in that The device status information of the first device includes one or more of the following: Part or all of the first information; state information of the first model; Whether the first device has the capability of accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; information associated with the model stored in the first device; performance requirements of a model associated with the first device; whether the first device supports model compression; location information of the first device; data arrival information of the first device; Information associated with the serving cell; Information associated with neighboring cells; as well as The first device has a buffer status report (BSR) related information.

103. The device according to claim 102, characterized in that The information associated with the serving cell and / or the neighboring cell includes a cell identifier and / or a signal quality of the cell.

104. The device according to claim 103, characterized in that The signal quality of the cell includes one or more of the following: Signal quality at the first moment; The signal quality of P consecutive sampling points before the first moment; as well as signal quality in a first time window before a first moment; Wherein, P is a positive integer.

105. The apparatus according to any one of claims 102-104, characterized in that The location information of the first device includes one or more of the following: Absolute position information of the first device at a first moment; relative position information of the first device at the first moment; Information about one or more locations of the first device within a second time window before the first moment; Information about N consecutive locations of the first device before the first moment; displacement information of the first device within a third time window before the first moment; and a moving speed of the first device; Wherein, N is a positive integer.

106. The apparatus according to any one of claims 102-105, characterized in that The data arrival information of the first device includes one or more of the following: The amount of data arriving at the first moment; The data volume and / or arrival time of the arrival data of M consecutive sampling points before the first moment; and The amount of data and / or the arrival time of the data arriving in a fourth time window before the first moment; Wherein, M is a positive integer.

107. The apparatus according to any one of claims 102 to 106, characterized in that The BSR related information of the first device includes one or more of the following: BSR at the first moment; The BSR of R consecutive sampling points before the first moment; and BSR in the fifth time window before the first moment; Wherein, R is a positive integer.

108. The apparatus according to any one of claims 104 to 107, characterized in that The first moment is the current moment and / or a historical moment.

109. The apparatus according to any one of claims 101 to 108, characterized in that The reporting method of the device status information of the first device includes periodic reporting and / or triggered reporting, wherein the triggering condition of the triggered reporting includes the occurrence of model parallel conflict.

110. The apparatus according to any one of claims 86 to 109, characterized in that The priority of the first information satisfies one or more of the following: The priority of the first information carried in the dynamic indication information sent by the second device is higher than the priority of the pre-configured first information; The predefined first information has the highest priority; The priority of the first information received at the second moment is higher than the priority of the first information received at the third moment; The difference between the current moment and the second moment is smaller than the difference between the current moment and the third moment.

111. A device for wireless communication, characterized in that: The invention comprises a memory and a processor, wherein the memory is used to store a program, and the processor is used to call the program in the memory to execute the method as claimed in any one of claims 1-30 or 31-55.

112. A device for wireless communication, characterized in that: Comprising a processor, configured to call a program from a memory to execute a method as claimed in any one of claims 1-30 or 31-55.

113. A chip, characterized in that: It comprises a processor, which is used to call a program from a memory, so that a device equipped with the chip executes the method as described in any one of claims 1-30 or 31-55.

114. A computer-readable storage medium, characterized in that A program is stored thereon, the program causing a computer to execute the method according to any one of claims 1-30 or 31-55.

115. A computer program product, characterized in that It comprises a program which causes a computer to execute the method as claimed in any one of claims 1-30 or 31-55.

116. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1-30 or 31-55.

Images