Management method and device, electronic equipment and storage medium

Abstract

The application discloses a management and control method and device, electronic equipment and storage medium, and belongs to the field of communication. The management and control method comprises the following steps: determining a target AI unit and / or updating a target report based on first information, wherein the first information comprises at least one of the following: first configuration information, the first configuration information comprising at least one of configuration information related to the first AI unit and configuration information related to the target report; and first capability information of the first device, the first capability information comprising capability information related to the first unit AI resource. By using the technical scheme, unified configuration management of the AI unit can be realized, flexible resource scheduling and allocation can be realized, the applicability of the AI technology in the communication system can be improved, and the service capability and service quality of the communication network can be improved.

Description

Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to a control method, device, electronic device, and storage medium. Background Technology Artificial Intelligence (AI) has been widely applied in various fields. In the field of communications, AI models are also being used more and more extensively. For example, in wireless communications, electronic devices (servers, network-side devices, or terminals) can collect training data from associated devices and / or their own devices for AI modeling. After the AI ​​modeling is completed, the electronic device can deploy the AI ​​unit (also called the AI ​​model) in itself or other electronic devices for model inference. The electronic device running the AI ​​model can also send (report) the results of the model inference to other electronic devices in the communication network, thereby improving the service capabilities and service quality of the communication network.

[0002] How to effectively manage AI units and / or inference results is an urgent problem to be solved. Summary of the Invention

[0003] This application provides a management and control method that enables unified management of AI units, flexible resource scheduling and allocation, and solves the problem of lacking effective management and control of AI units and / or inference results in related technologies.

[0004] In a first aspect, a control method is provided, applied to a first device, the method comprising: determining a target AI unit based on first information and / or updating a target report; The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The AI ​​units applicable to the first device; the AI ​​units activated by the first device; The AI ​​unit that the first device will run; AI unit used to obtain the target report; Reference AI unit used to obtain the target report; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report; The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

[0005] Secondly, a control device is provided for use in the first device, the device comprising: The processing module is used to determine the target AI unit and / or update the target report based on the first information; The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The AI ​​units that can be applied to the first device; The AI ​​unit activated by the first device; The AI ​​unit that the first device will run; AI unit used to obtain the target report; Reference AI unit used to obtain the target report; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report; The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

[0006] Thirdly, an electronic device is provided, comprising a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect.

[0007] Fourthly, a readable storage medium is provided, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0008] One of the above technical solutions has the following advantages or beneficial effects: by determining the target AI unit and / or updating the target report based on the first information, it is possible to achieve flexible resource scheduling and allocation through unified configuration management of the AI ​​unit, thereby improving the applicability of AI technology in the communication system and enhancing the service capabilities and service quality of the communication network.

[0009] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0010] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This diagram illustrates a block diagram of a wireless communication system to which embodiments of this application may be applied; Figure 2 This diagram illustrates a flow chart of a control method provided in an embodiment of the present disclosure. Figure 3 This diagram illustrates the processing mode of the AI ​​unit in the first device provided in this embodiment of the present disclosure. Figure 4 This diagram illustrates another flow chart of the control method provided in an embodiment of the present disclosure; Figure 5 This illustration shows another flowchart of the control method provided in an embodiment of the present disclosure; Figure 6 This illustration shows another flowchart of the control method provided in an embodiment of the present disclosure; Figure 7 This illustration shows another flowchart of the control method provided in an embodiment of the present disclosure; Figure 8 This illustration shows another flowchart of the control method provided in an embodiment of the present disclosure; Figure 9 This diagram shows a block diagram of a control device provided in an embodiment of this application; Figure 10 A schematic diagram of the hardware structure of an electronic device for implementing the control method provided in the embodiments of this disclosure; Figure 11 A schematic diagram of the hardware structure of a terminal to implement an embodiment of this application; Figure 12 A schematic diagram of the hardware structure of a network-side device to implement an embodiment of this application. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in this disclosure, the technical solutions in the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0014] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, the scope of protection for "A or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. In addition, the terms "A and / or B," "at least one of A and B," and "at least one of A or B" also cover at least the above three scenarios. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0015] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as the sender explicitly informing the receiver of specific information, the required operation, or the requested result in the instruction sent. An indirect instruction can be understood as the receiver determining the corresponding information based on the instruction sent by the sender, or making a judgment and determining the required operation or requested result based on the judgment result.

[0016] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G communication system.

[0017] Figure 1This diagram illustrates a block diagram of a wireless communication system applicable to embodiments of this application. The wireless communication system includes a terminal 11, a network-side device 12, and a server 13. The terminal 11 can also be referred to as User Equipment (UE), and can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home devices (home appliances with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game consoles, personal computers (PCs), ATMs, or self-service machines, etc. Wearable devices include: smartwatches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in this application embodiment. Network-side equipment 12 may include access network equipment or core network equipment, and server 13 may be a single server or server cluster connected to the core network equipment. Access network equipment can also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (APs), or Wireless Fidelity (WiFi) nodes, etc.Among them, base stations can be referred to as Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit / Receive Point (TRP), and Non-Terrestrial Network (NTN) equipment (such as satellite or high altitude platform). The term "base station" can be any suitable term in the field, such as "station" or any other appropriate term in the relevant field, as long as the same technical effect is achieved. The term "base station" is not limited to specific technical terms. It should be noted that the embodiments of this application only use the base station in the NR system as an example for introduction, and do not limit the specific type of base station.

[0018] Core network equipment, also known as core network nodes, core network functions, or core network elements, includes, but is not limited to, at least one of the following: Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (L-NEF), and Binding Support. Functions include BSF, Application Function (AF), Location Management Function (LMF), Gateway Mobile Location Centre (GMLC), Network Data Analytics Function (NWDAF), and Non-Terrestrial Network (NTN) devices (such as satellite or high altitude platform stations).It should be noted that the embodiments of this application only use the core network equipment in the NR system as an example for introduction, and do not limit the specific type of core network equipment. If the name of the core network equipment mentioned in the embodiments of this application changes in subsequent protocol versions (such as 6G), it is also within the scope of protection of this application.

[0019] Optionally, the core network equipment can be implemented by one or more functional modules in a single device, or by multiple devices working together; this application does not specifically limit this. It is understood that the aforementioned functional modules can be network elements in hardware devices, software functional modules running on dedicated hardware, or virtualized functional modules instantiated on a platform (e.g., a cloud platform).

[0020] The control method provided in this application will be described in detail below with reference to the accompanying drawings, through some embodiments and application scenarios.

[0021] Figure 2 This illustration shows a flowchart of a control method provided in an embodiment of the present disclosure, such as... Figure 2 As shown, the method may include the following steps: In step S101, the target AI unit is determined and / or the target report is updated based on the first information.

[0022] The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The first device can utilize AI units; The AI ​​unit activated by the first device; The first device will be the AI ​​unit that runs; The AI ​​unit that acquires the target report; A reference AI unit for obtaining the target report; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report. The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

[0023] This method can be applied to a first device, which can be a terminal, a network-side device, or a server. The network-side device can be an access network device, such as a base station, or a core network device, such as 5GC (5G Core Network). This application does not impose any restrictions on this.

[0024] The first information may be received by the first device from at least one other connected electronic device (e.g., the terminal as the first device may obtain the first configuration information from the connected base station), or it may be information configured by the user for the first device (e.g., the first device may obtain the first capability information in response to the user's configuration information or by reading a preset configuration file). Of course, it may also be any combination of the two acquisition methods (e.g., some of the first information is received by the first device from at least one other electronic device, and the other first information is information configured by the user for the first device). This disclosure does not limit this.

[0025] In the embodiments of this disclosure, the AI ​​unit may also be referred to as an AI model, ML (Machine Learning) model, ML unit, AI structure, AI function, or AI characteristic. The specific functions implemented by the AI ​​unit may also be referred to as a neural network unit, neural network model, or neural network function. An AI unit refers to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc., related to AI. Alternatively, an AI unit may be a processing method, algorithm, function, module, or unit for a specific dataset (which may include the input dataset and / or output dataset of the AI ​​unit). Alternatively, an AI unit may be a processing method, algorithm, function, module, or unit running on AI / ML-related hardware, such as a GPU (Graphics Processing Unit), NPU (Neural Processing Unit), TPU (Tensor Processing Unit), or ASIC (Application Specific Integrated Circuit). This disclosure does not specifically limit the definition of AI / ML-related hardware.

[0026] The state of the AI ​​unit corresponding to the first device can include three states: not loaded into the first device, loaded into the first device but not running, and loaded into the first device and running. Alternatively, it can include four states: not loaded into the first device, loaded into the first device but not activated, activated into the first device but not running, and activated into the first device and running.

[0027] In an optional embodiment, the target AI unit can be configured as any of the following: Inactive AI units; Activated AI unit; The running AI unit; The AI ​​unit has stopped operating; Deactivated AI units; Released AI units.

[0028] In some embodiments, the first device may determine the target AI unit from the first AI units based on first information. Specifically, this may include: A1. The first AI unit can be an AI unit applicable to the first device, that is, an AI unit that can be loaded and run on the first device. Specifically, the input corresponding to the AI ​​unit applicable to the first device is something that the first device can collect or acquire, and the output corresponding to the AI ​​unit applicable to the first device can be applied to the first device or to other electronic devices connected to the first device. The first device meets the network conditions corresponding to the AI ​​unit applicable to the first device, for example, the first device meets the environmental dependencies required for the operation of the AI ​​unit. The AI ​​unit applicable to the first device can be an AI unit already loaded in the first device, or it can be an AI unit not loaded into the first device; this application does not impose any restrictions on this.

[0029] A2. The first AI unit can be an AI unit activated in the first device. For example, the target AI unit can be an AI unit to be run, an AI unit to be released, or an AI unit to be deactivated. The first device can determine the target AI unit based on the AI ​​units that have been activated in the first device.

[0030] A3. The first AI unit can be an AI unit that the first device will run. This AI unit can be an AI unit already activated and / or an AI unit already loaded in the first device, or it can be a subset of AI units determined from the already activated and / or loaded AI units in the first device according to preset conditions. For example, the target AI unit can be an AI unit to be run, an AI unit to be deactivated, or an AI unit to be released. The first device can determine the target AI unit based on the AI ​​units that the first device will run.

[0031] A4. The first AI unit may be an AI unit that acquires a target report, meaning an AI unit capable of acquiring the target report. For example, the target AI unit may be an AI unit that acquires a target report. The first device may determine and run the target AI unit from among the AI ​​units that acquire the target report based on first information, and update the target report.

[0032] A5. The first AI unit may be a reference AI unit for obtaining the target report. The reference AI unit for obtaining the target report refers to a preset AI unit that has the ability to obtain the target report and can be used to evaluate the performance of other AI units.

[0033] It is understood that the first AI unit can include any combination of multiple items in A1-A5 above. That is, the first device can determine the target AI unit based on multiple items in A1-A5 above. For example, if the target AI unit is an AI unit to be released, the first device can determine the AI ​​unit to be released based on the union of the AI ​​units activated by the first device and the AI ​​units that the first device will run. As another example, the target AI unit can be an AI unit that obtains a target report. The first device can determine the target AI unit based on the intersection of the AI ​​units that the first device can apply and the AI ​​unit that obtains the target report, and can further update the target report.

[0034] In an optional embodiment, the target AI unit may also be configured with at least one of runtime or priority for the control of the target AI unit.

[0035] In an optional embodiment, determining the target AI unit based on the first information includes the AI ​​management unit or layer in the first device, and determining the target AI unit to run / activate / release based on the first information. Optionally, the target AI unit to run / activate / release can be determined based on the first capability information of the first device, or based on the configuration information related to the first AI unit (e.g., the priority associated with the first AI unit), or based on the AI ​​resources required to run the target AI unit, or based on the configuration information related to the target report, or further, the target AI unit to run / activate / release can be determined based on the configuration information related to the first AI unit, the configuration information related to the target report, and the first capability information of the first device.

[0036] In an optional embodiment, determining whether to update the target report based on the first information includes determining the time, order, or whether to update the target report based on the first information. Optionally, the time, order, or whether to update the target report may be determined based on the first capability information of the first device; or, at least one of the following may be determined based on configuration information related to the target report (e.g., the priority of the target report); or, at least one of the following may be determined based on configuration information related to the target AI unit corresponding to the target report (e.g., the AI ​​resources required by the target AI unit corresponding to the target report); or, the time, order, or whether to update the target report may be determined based on multiple of the first capability information of the first device, the configuration information related to the target report, and the configuration information related to the target AI unit.

[0037] In some embodiments, the configuration information related to the target report includes at least one of the following: The configuration information of the first signal, wherein if the configuration information related to the target report includes the configuration information of the first signal, the target report is obtained by the first device based on the first signal; The content of the target report; The time-domain location where the target report was sent; The time-domain type of the target report sent; The frequency of target report submissions; Prioritize target reports; The target report contains instructions and information.

[0038] Configuration information related to the target report may include: B1. Configuration information of the first signal; for example, the first device can be a terminal, and the target report can be a report obtained by the terminal during channel estimation or signal measurement, such as a channel state report, beam information report, positioning information report, or sensing information report. The configuration information of the first signal can be the relevant configuration information of the RS (Reference Signal) used by the terminal for channel estimation or measurement, such as the relevant configuration information of CSI-RS (Channel Status Information-Reference Signal), DMRS (Demodulation Reference Signal), or PRS (Positioning Reference Signal). Of course, the first signal can include multiple reference signals among the above-mentioned different reference signals, and this application does not limit this. Optionally, the configuration information of the first signal may specifically include one or more of the following: the identification information of the first signal, the time-frequency position of the first signal, the period of the first signal, the time-domain type of the first signal, the use case application embodiment associated with the first signal, and the number of the first signal.

[0039] B2. Content of the target report; For example, the content of the target report in the configuration information related to the target report can be configuration information containing specific content in the target report. Multiple different types of target reports can also be predefined (e.g., CSI reporting type, beam reporting type, positioning information type), where the specific content of different types of target reports differs. The configuration information related to the target report can be configuration information for the type of target report to determine the specific content of the corresponding target report. The configuration information related to the target report can also be indication information for the AI ​​unit output information.

[0040] B3. The time-domain location of the target report transmission; for example, the configuration information related to the target report may include at least one of the frame, subframe, and time slot for target report transmission, thereby enabling the first device to determine the timing of target report transmission based on the time-domain location of target report transmission. Furthermore, the time-domain location of target report transmission can also be used to determine whether the terminal can obtain the target report from the target AI unit before the target report is transmitted or before a reference time. Optionally, it can be further determined whether to update the target report.

[0041] B4. Time domain type of target report transmission; for example, the above time domain type can be one of periodic transmission, non-periodic transmission, or semi-static transmission in the time domain.

[0042] B5. Target report sending cycle; e.g., 20ms, 40ms, 80ms, 160ms, 320ms, 640ms, etc. B6. Priority of Target Reports; Optionally, if multiple reports need to be sent and there is a time-domain location conflict, the first device can determine the target report to be sent first based on the priority of the target reports. Alternatively, in the case of AI resource conflicts, such as multiple target reports corresponding to target AI units that need to run, but the AI ​​resources are insufficient to support the operation of the target AI units or their simultaneous operation, the target AI unit can be determined based on the priority of the multiple target reports, for example, the target AI unit to run can be determined according to the priority of the target reports and the AI ​​resources available to the terminal.

[0043] B7. Instructions for the target report; for example, the instructions for the target report may be, for instance, the ID of the target report.

[0044] It is understood that the configuration information related to the target report can include any combination of multiple items in B1-B7 above. For example, the configuration information related to the target report can include the content of the target report, the time domain type of the target report, the time domain type of the target report, and the period of the target report. The first device can determine the content of the target report and how to send the target report.

[0045] Optionally, in some embodiments, the configuration information of the target report may further include at least one of the following: Configuration information related to the first AI unit.

[0046] Indication information is used to indicate the first AI unit capable of obtaining the target report. Optionally, the indication information may be at least one of the following: identification information of the AI ​​unit, model indication information of the AI ​​unit, indication information of the dataset associated with the AI ​​unit, and indication information enabling the AI ​​unit.

[0047] In some embodiments, the configuration information related to the first AI unit includes at least one of the following: The instruction information associated with the first AI unit; Configuration information of the second signal associated with the first AI unit; The complexity of the first AI unit; AI resources occupied by the first AI unit; The first AI unit occupies the first unit's AI resources; The input information associated with the first AI unit; The output information associated with the first AI unit; The priority of the first AI unit.

[0048] The first device can determine the target AI unit and / or update the target report based on the configuration information related to the first AI unit and / or the configuration information related to the target report.

[0049] Specifically, the configuration information related to the first AI unit may include: C1. Indication information associated with the first AI unit; The indication information associated with the first AI unit can be indication information unrelated to the model structure of the first AI unit. For example, the indication information associated with the first AI unit can be the identifier of a specific dataset associated with the AI ​​unit, or the identifier of a specific scenario, environment, channel characteristics, or device related to the AI ​​unit, or the identifier of a function, characteristic, capability, or module related to the AI ​​unit. This application does not impose any restrictions on this. For example, the indication information associated with the first AI unit can be the unique identifier of the AI ​​unit, or the identifier of the training dataset corresponding to the AI ​​unit, or the identifier of the validation dataset corresponding to the AI ​​unit. For example, if the identifier information (data_id) of the training datasets corresponding to AI unit A, AI unit B, and AI unit C are 1, 1, and 2 respectively, and the indication information associated with the first AI unit is the identifier information of the training dataset (data_id=1), then the first AI unit is the aforementioned AI unit A and AI unit B.

[0050] Optionally, the indication information associated with the first AI unit can also be indication information related to the model structure of the first AI unit. For example, the indication information related to the model structure can be used to indicate the model structure corresponding to the AI ​​unit. For instance, the model structure corresponding to AI unit A is a convolutional neural network (model_id=1), and the model structures corresponding to AI units B and C are LSTM (Long Short-Term Memory) networks (model_id=2). The indication information related to the model structure of the first AI unit (model_id=2) then indicates that the first AI unit is either AI unit B or AI unit C. For example, the model structure corresponding to AI unit A is model structure 1 as specified in the protocol, and the model structure corresponding to AI unit B is model structure 2 as specified in the protocol, wherein model structure 1 and model structure 2 differ in at least one of their model composition, architecture, or hyperparameters.

[0051] C2. Configuration information of the second signal associated with the first AI unit; for example, the AI ​​unit may also be associated with a second signal. For instance, if the AI ​​unit is used for localization, the second signal associated with the AI ​​unit could be, for example, a PRS (Pressure Signal Representation System), and the configuration information of the second signal associated with the AI ​​unit could be the configuration information of the PRS. As another example, if the AI ​​unit is used for beam training, the second signal associated with the AI ​​unit could be, for example, a CSI-RS (Content Specific Information Representation System), and the configuration information of the second signal associated with the AI ​​unit could be the configuration information of the CSI-RS.

[0052] C3. Complexity of the first AI unit; The complexity of an AI unit refers to the complexity of the model or the complexity of the AI ​​unit's operations. It depends on the number of parameters, network structure, scale of training data, computational load, and other indicators of the AI ​​unit. The first device can determine the target AI unit based on the complexity of the first AI unit.

[0053] C4. AI resources occupied by the first AI unit; the AI ​​resources occupied by the first AI unit may include AI storage resources and / or AI computing resources occupied by the first AI unit. The first device may determine the target AI unit based on at least one of the AI ​​resources occupied by the first AI unit, such as FLOPs (Floating Point Operations Per Second), memory usage, storage usage, or AI computing resources. For example, the terminal determines whether the first AI unit can be run based on the remaining AI resources and the AI ​​resources occupied by the first AI unit. If the remaining AI resources of the terminal are greater than the AI ​​resources occupied by the first AI unit, then the first AI unit can be run. For example, the terminal determines the target AI unit based on the remaining AI resources, the AI ​​resources occupied by multiple first AI units to be run, and the priority of the multiple first AI units to be run. The target AI unit is one or more of the aforementioned multiple first AI units.

[0054] C5, the AI ​​resources occupied by the first AI unit; In some embodiments, a first unit AI resource is introduced as an abstract unit of measurement for the AI ​​resources occupied by the AI ​​unit, also known as an APU (Artificial Intelligence Process Unit). This unit can be used to evaluate the capabilities of electronic devices running AI units, determine the AI ​​resources occupied by different AI units, and serve as the basis for unified and flexible scheduling and allocation of AI units.

[0055] In some embodiments, a first unit AI resource, also known as an APU (Artificial intelligence Process Unit), is defined, which can be used to evaluate the capabilities of electronic devices running AI units, determine the AI ​​resources occupied by different AI units, and thus enable the management and scheduling of AI resources of electronic devices.

[0056] For example, the remaining AI resources of the first device can be m units of AI resources, and AI unit A and AI unit B occupy n and k units of AI resources respectively. In the case of n+k, the first device can run AI unit A and AI unit B.

[0057] For example, the remaining AI resources of the first device can be m units of AI resources, and AI unit A and AI unit B occupy n and k units of AI resources respectively. In the case of n + k and no higher-priority AI unit needs to run, the first device can run AI unit A and AI unit B.

[0058] For example, the remaining AI resources of the first device can be m first-unit AI resources. The AI resources occupied by AI unit A and AI unit B are n first-unit AI resources and k first-unit AI resources respectively. In the case of m < n + k, and m > n and / or m > k, the first device can run the one with the higher priority among AI unit A and AI unit B.

[0059] C6. The input information associated with the first AI unit; C7. The output information associated with the first AI unit; It can be understood that different first AI units can be associated with the same or different input information, and different first AI units can be associated with the same or different output information. Therefore, the first device can determine the corresponding target AI unit from the first AI units according to the input information and / or output information in the configuration information related to the first AI unit. The input information associated with the first AI unit can be, for example, the recognition information of one or more signals, or the recognition information of other reports. The output information associated with the first AI unit can be, for example, compressed CSI information, the strongest beam information, positioning information, or sensing information.

[0060] C8. The priority of the first AI unit; It can be understood that the priorities of different first AI units can be the same or different. In the case of conflicts in the AI resources of multiple first AI units, the first device can determine the target AI unit based on the priority of the first AI unit. Or, when the remaining AI resources of the first device are not sufficient to run the first AI units waiting to run, the target AI unit to run can be determined based on the priority of the first AI unit. For example, select the N first AI units with the highest priority less than the remaining AI resources of the first device as the target AI units to run.

[0061] In an optional embodiment, the first AI unit to be run (e.g., the AI ​​unit that the first device will run) can queue up and wait for the first device to have AI resources (e.g., idle AI resources) to run the first AI unit. Once the first AI unit is in the queue, it can be considered a target AI unit and can be run by the first device. Optionally, the first AI unit to be run can queue up and wait for the first device to have AI resources (e.g., idle AI resources) to run the first AI unit before a first reference time. Once the first AI unit is in the queue, it can be considered a target AI unit and can be run by the first device. If no available AI resources are available for the first AI unit by the first reference time, the first device can instruct higher layers or other electronic devices, e.g., instruct that no AI resources are available for the first AI unit or target report. Optionally, the first AI unit to be run can queue up and wait for the first device to have AI resources (e.g., idle AI resources) to run the first AI unit before a first timer expires. Once the first AI unit is in the queue, it can be considered a target AI unit and can be run by the first device. If no AI resources are available for the first AI unit by the time the first timer expires, the first device instructs higher-level or other electronic devices, such as indicating that no AI resources are available for the first AI unit or target report.

[0062] In some embodiments, determining the target AI unit and / or updating the target report based on the first information includes: determining the target AI unit and / or updating the target report based on the first information and the third information.

[0063] The third piece of information includes any one of the following: AI resource utilization rate; busy rate.

[0064] The AI ​​resource occupancy ratio of the first device can characterize the utilization of AI resources on the first device. In some possible implementations, the AI ​​resource occupancy ratio of the first device can include the utilization ratio of computing resources and / or storage resources. The AI ​​resource occupancy ratio of the first device can be the higher of the utilization ratios of computing resources and storage resources, or it can be a weighted resource occupancy ratio obtained by weighting the utilization ratios of computing resources and storage resources according to corresponding preset weights.

[0065] The Busy Ratio of the first device can characterize the utilization rate of AI resources of the first device within a time period. AI resources can include computing resources and storage resources. The Busy Ratio of the first device can also be the utilization rate of components that process AI resources (such as GPUs), or the average utilization rate within a time period.

[0066] In some embodiments, determining the target AI unit and / or updating the target report based on the first information and the third information includes: If the AI ​​resource utilization rate is greater than or equal to a first threshold, or if the busy rate is greater than or equal to a second threshold, send a first indication message regarding the AI ​​resource utilization rate; or... If the AI ​​resource occupancy rate is greater than or equal to the third threshold, or if the busy rate is greater than or equal to the fourth threshold, a second indication message indicating that the AI ​​resource is busy or there is no available AI resource is sent.

[0067] Among them, the third threshold can be greater than the first threshold, or the third threshold can be equal to the first threshold; the fourth threshold can be greater than the second threshold, or the fourth threshold can be equal to the second threshold.

[0068] For example, once the occupancy rate of AI resources is greater than or equal to a first threshold, or the busy rate is greater than or equal to a second threshold, the first device (such as the AI ​​management unit, AI resource management unit, MAC layer, AI layer) can send the first indication information of the occupancy rate of AI resources to higher layers or other electronic devices. When the occupancy rate of AI resources is greater than or equal to the third threshold, or the busy rate is greater than or equal to the fourth threshold, the first device (such as the AI ​​management unit, AI resource management unit, MAC layer, AI layer) can send a second indication message to the higher layer or other electronic devices indicating that the AI ​​resources are busy or there are no available AI resources.

[0069] In some possible implementations, the first device may periodically send a first indication message to higher layers or other electronic devices when the occupancy rate of AI resources is greater than or equal to a first threshold, or when the busy rate is greater than or equal to a second threshold. When the occupancy rate of AI resources is less than a fifth threshold and the busy rate is less than a sixth threshold, the first device may stop sending the first indication message. The fifth threshold is less than the first threshold and the sixth threshold is less than the second threshold.

[0070] Taking the first device as an example, the terminal can report the first indication information of the AI ​​resource occupancy rate to the base station when the AI ​​resource occupancy rate is greater than or equal to the first threshold.

[0071] In some optional embodiments, determining the target AI unit and / or updating the target report based on the first and third information may include: determining the processing time of the target AI unit based on a first range of AI resource occupancy or a second range of busy rate. Optionally, if the AI ​​resource occupancy or busy rate falls within the first range, the first device may determine the processing time of the target AI unit as a second processing time. If the busy rate falls within the second range, the first device may determine the processing time of the target AI unit as a third processing time. If the AI ​​resource occupancy rate falls within the third range, the first device may determine the processing time of the target AI unit as a fourth processing time. If the busy rate falls within the fourth range, the first device may determine the processing time of the target AI unit as a fifth processing time. That is, for the same target AI unit, the processing time of the target AI unit can be determined based on the AI ​​resource occupancy or busy rate, and the processing time of the target AI unit may be different for different AI resource occupancy or busy rates.

[0072] It is understandable that the configuration information related to the first AI unit can include any combination of multiple items from C1-C8 mentioned above. For example, the configuration information related to the first AI unit can include the AI ​​resources occupied by the first AI unit and the first unit AI resources occupied by the first AI unit. The first device can evaluate the resource satisfaction of the AI ​​unit from two dimensions based on the AI ​​resources occupied by the first AI unit and the first unit AI resources occupied by the first AI unit to determine the target AI unit. As another example, the configuration information related to the first AI unit can include the AI ​​resources occupied by the first AI unit, the first unit AI resources occupied by the first AI unit, and the priority of the first AI unit. The first device can determine the target AI unit based on priority in the event of resource conflict, based on resource satisfaction.

[0073] In some embodiments, the first unit AI resource is determined according to any one of the following: D1, the preset first AI resource; wherein, the first AI resource is related to the first reference report; For example, the first reference report can be a report with a preset format and / or content. The first AI resource is the AI ​​resource corresponding to the first reference report, that is, the AI ​​resource corresponding to the first reference report is used as the first unit AI resource (e.g., one APU) for unified APU resource scheduling and allocation. For instance, if the specified reference AI beam report is one APU, the number of APUs required to obtain the target report is determined by comparing the complexity of the target report with that of the AI ​​beam report. If the number of cycles / number of signals / subband range / bandwidth / number of ports of the first signal associated with the target report is K times that of the reference report, i.e., K APUs. Or, if two of the numbers of cycles / number of signals / subband range / bandwidth / number of ports of the first signal associated with the target report are K*L times that of the reference report, i.e., K*L APUs, then the number of APUs corresponding to the target report is K*L times that of the reference report.

[0074] D2, a preset second AI resource; wherein, the second AI resource is related to the function of the first reference AI; For example, the first reference AI function can be different AI modules / AI functions (e.g., preset input and / or output counts, preset fully connected modules, such as preset transform blocks, such as preset transform complexity). That is, the AI ​​resources corresponding to the first reference AI function are used as the first unit AI resources for unified resource scheduling and allocation. For instance, the first reference function can be specified as 1 APU. The number of APUs corresponding to the target report is obtained by comparing the complexity of the target report or target AI unit with that of the first reference function. If the transform associated with the target report or target AI unit is K times that of the reference report, then the number of APUs is K times or O(K) times that of the reference report. If the input is K times that of the first reference function, then the number of APUs is K^2 (or O(K^2)) APUs of the reference report.

[0075] D3, a preset third AI resource, wherein the third AI resource is related to the first reference AI unit; For example, the first reference AI unit may be a preset reference model. That is, the AI ​​resources corresponding to the first reference AI unit are used as the AI ​​resources of the first unit for unified resource scheduling and allocation. For instance, by comparing the first AI unit with the first reference AI unit to determine the APU of the first AI unit, it can be used to determine whether there are sufficient AI resources available for the first AI unit, thereby determining the target unit.

[0076] D4. The first preset number of floating-point operations (FLOPs); A first preset number of floating-point operations (FLOPs) can be used as the AI ​​resource of the first unit for unified resource scheduling and allocation. For example, the APU of the first AI unit is determined by comparing the number of FLOPs of the first AI unit with that of the first reference AI unit, thereby determining whether there are sufficient AI resources available for the first AI unit, and thus identifying the target unit.

[0077] D5, the second preset number of AI parameters.

[0078] AI parameters can be model parameters or hyperparameters within an AI unit, such as one or more parameters including input node parameters, output node parameters, hidden node parameters, the number of transform blocks, the number of transform heads, and dimensions. The number of AI parameters can be used to determine the AI ​​resources of the first AI unit for unified resource scheduling and allocation. For example, the number of APUs in the first AI unit can be determined by comparing the parameters of the first AI unit with the aforementioned second preset number of AI parameters.

[0079] In some optional embodiments, if the number of AI resources (i.e., the number of APUs) of the first AI unit determined by any one of D1-D5 is not an integer, it can be rounded up. For example, the first AI unit may need 1.5 first AI resource APUs, but the first device can determine whether to run and when to run the target AI unit based on 2 APUs.

[0080] The first unit of AI resources includes at least one of the following: Unit AI storage resources; Unit AI computing resources.

[0081] In some embodiments, the first capability information is used to characterize the processing capability of the AI ​​unit of the first device, and specifically the first capability information includes at least one of the following: Storage capacity information; Processing capacity information; Processing efficiency information; Among them, the storage capacity information includes that the first device can store up to M1 first unit AI resources, and / or that the first device can store up to N1 base AI units; The processing capacity information includes at least one of the following: The first device can process a maximum of M2 first-unit AI resources; The first device can process a maximum of N2 baseline AI units; The first device can process a maximum of M3 first-unit AI resources in the first time. The first device can process a maximum of N3 baseline AI units in the first time. The first device can process a maximum of M4 first unit AI resources at the same time; The first device can process a maximum of N4 baseline AI units at the same time; The first device has already occupied M5 first-unit AI resources; The first device has already occupied N5 baseline AI units; The first device has already occupied M6 first-unit AI resources at the first reference time; The first device has already occupied N6 reference AI units at the first reference time; The processing efficiency information includes the first processing time for processing a third preset number of first unit AI resources, and / or the second processing time for processing a fourth preset number of AI units.

[0082] In some optional embodiments, the processing time for the terminal to process a third preset number of first unit AI resources in the first state is different from the processing time for processing the third preset number of first unit AI resources in the second state. For example, the first capability information may include different processing efficiency information associated with different states. For instance, if the first device is a terminal, the first state may be a power-saving state of the terminal, while the second state may be a non-power-saving state of the terminal. Furthermore, the first state and the second state may also be different states representing the busy level of the first device.

[0083] It is understood that the baseline AI unit in the first capability information can be an AI unit that is assumed to require a fifth preset number (e.g., Z) of unit AI resources. For example, the fifth preset number (e.g., Z) of unit AI resources can be the maximum number of unit AI resources required to run the AI ​​unit, or the fifth preset number (e.g., Z) of unit AI resources can be the unit AI resources required to process the maximum configuration.

[0084] It is understandable that when the first device stores the AI ​​unit, it also needs to store the environment and dependencies necessary for the AI ​​unit to activate and / or run. The above N1*Z can be less than or equal to M1 or M2.

[0085] Optionally, the first device needs to consider that it can process a maximum of M2 first unit AI resources and a maximum of N2 baseline AI units. If either of these conditions exceeds the first device's capacity, the requirement cannot be met. For example, N2*Z can be less than or equal to M2.

[0086] Depending on the different implementations of the first device, the first device can have a variety of different AI unit processing modes. Figure 3 This diagram illustrates the processing mode of the AI ​​unit in the first device provided in this embodiment of the present disclosure, such as... Figure 3As shown in (a), the first device can be in a mode of loading first and then selecting to run. Figure 3 In the AI ​​units shown in (a), AI units 1-6 are AI units already recorded in the first device, where AI units 1-4 are loaded (or stored) but not running, while AI units 5 and 6 are running AI units. Figure 3 As shown in (b), the first device can also be in a load-and-run mode. Figure 3 In the AI ​​units shown in (b), AI units 7-8 are loaded and running AI units. It should be noted that a running AI unit can be an AI unit that is currently using AI resources to perform calculations, or an AI unit that needs to run periodically.

[0087] It should be noted that the "load first, then select to run" mode can be further divided into more states. For example, the AI ​​unit can be made active by loading the necessary environment and dependencies so that it can be run at any time. For AI units that will not run for a short period of time, the necessary environment and dependencies can be unloaded to save storage resources and keep the AI ​​unit in an inactive state.

[0088] It is understandable that when the first device is in the mode of loading and running, M1 can be equal to M2 and N1 can be equal to N2. When the first device is in the mode of loading first and then selecting to run, M2 can be less than M1 and N2 can be less than N1.

[0089] It is understandable that the maximum number of M3 first-unit AI resources and N3 baseline AI units that the first device can process in the first time period, and the maximum number of M4 first-unit AI resources and N4 baseline AI units that the first device can process simultaneously, represent the first capability information of the first device from the perspective of parallel processing capability. The M5 units of AI resources and N5 baseline AI units that the first device has already occupied represent the first capability information of the first device from the perspective of the AI ​​resources that the first device has already occupied.

[0090] The first reference time is a point in time determined based on the target report time corresponding to the target AI unit and the estimated runtime of the AI ​​unit. It is used to determine whether the AI ​​unit can run at the latest by that time. If it is determined that the AI ​​unit cannot run by that time, the first device can send an indication message indicating insufficient AI resources or that the target AI unit should not run. For example, the first reference time can be the latest time that satisfies the following formula.

[0091] T ref <T report -Trun (Formula 1) Among them, T ref As the first reference time, T report T represents the sending time of the target report corresponding to the AI ​​unit. run This refers to the estimated runtime of the AI ​​unit or the processing time determined based on its capabilities.

[0092] In some embodiments, the fifth preset quantity may be less than or equal to at least one of M2, M3, and M4. The third preset quantity may be equal to M2, M3, or M4, and the fourth preset quantity may be equal to N2, N3, or N4.

[0093] In some embodiments, the first device may determine the target AI unit and / or update the target report based on at least one of the first information, namely the configuration information related to the first AI unit, the configuration information related to the target report, and the first capability information of the first device.

[0094] For example, the first device can respond to receiving multiple different service requests, which can be associated with different AI units. For instance, different service requests can be associated with a beam management AI unit, a terminal positioning AI unit, and a channel estimation AI unit, respectively. Based on first information, the first device can determine the AI ​​units corresponding to the multiple different service requests as target AI units from among the first AI units. It can also further update the target report corresponding to the target AI unit and send the target report proactively or in response to a report sending request.

[0095] For example, the first device may also respond to receiving multiple different service requests, which may be associated with different target reports. For example, different service requests may be associated with beam management reports, terminal positioning reports, and channel estimation reports, respectively. The first device may determine the target AI unit corresponding to the beam management report, terminal positioning report, and channel estimation report based on the first information, and may further update the target report corresponding to the target AI unit, and actively or in response to a report sending request to send the target report.

[0096] It should be noted that the above-mentioned determination of the target AI unit and / or update of the target report based on the first information includes a variety of parallel technical solutions. Those skilled in the art can obtain the technical solutions not described in this application for the sake of brevity by logical reasoning based on the above description in the specification and knowledge in the field.

[0097] By adopting the above technical solution, the target AI unit is determined and / or the target report is updated based on the first information. This enables flexible resource scheduling and allocation through unified configuration management of the AI ​​unit, thereby improving the applicability of AI technology in communication systems and enhancing the service capabilities and quality of communication networks.

[0098] It is understandable that the primary capability information of the first device may change with changes in the state of the first device. For example, if the first device is a terminal, the terminal's primary capability information may change with changes in the terminal's remaining battery power; when the terminal's battery power is below a threshold, the primary capability information will decrease accordingly. As another example, if the first device is a foldable screen terminal, the change in the foldable screen shape may cause changes in the number of antennas in the terminal, and the terminal's primary capability information may change accordingly with the change in the foldable screen terminal's shape. Furthermore, if the first device is a cloud server, the cloud server's primary capability information will also change accordingly with the dynamic adjustment of cloud resources or AI resources. Secondly, the first device can also be capability information corresponding to different busy states or AI resource occupancy states; for example, the processing time required to process the same AI unit may differ under different busy states.

[0099] In some embodiments, the first device includes K state information corresponding to different AI capabilities, wherein the first capability information includes at least one of the following: The storage capacity information corresponding to the K status information; The processing capacity information corresponding to the K status information; The processing efficiency information corresponds to each of the K state information.

[0100] Figure 4 This illustration shows another flowchart of the control method provided in an embodiment of the present disclosure, such as... Figure 4 As shown, step S101 may include the following steps: In step S1011, the target AI unit is determined based on the first information and the status information of the first device.

[0101] In some embodiments, the corresponding first capability information can be determined from the first capability information corresponding to K status information based on the status information (e.g., current status information) of the first device, and the target AI unit can be determined based on the first information of the corresponding first capability information.

[0102] In some embodiments, the corresponding first capability information can be determined from the first capability information corresponding to K status information based on the status information (e.g., current status information) of the first device, and whether to run the target AI unit can be determined based on the first information of the corresponding first capability information. For example, if the first device is in a power-saving state, or the occupancy rate of AI resources is greater than or equal to a preset third occupancy rate threshold, the first device may not run the target AI unit.

[0103] In some embodiments, the corresponding first capability information can be determined from the first capability information corresponding to the K state information based on the estimated state information of the first device at the first reference time, and the target AI unit can be determined based on the first information that has determined the corresponding first capability information.

[0104] By adopting the above technical solution, the corresponding first capability information can be determined according to the different states of the first device, and the target AI unit can be determined according to the first information that has determined the corresponding first capability information. This can adapt to the state changes of the first device during the dynamic scheduling of the AI ​​unit, and further improve the accuracy of the dynamic scheduling of the AI ​​unit.

[0105] In some embodiments, the first capability information includes the support capabilities of the AI ​​unit. If an AI unit supported by the first device exists, the first capability information further includes at least one of the following: At least one of the inputs, outputs, and network conditions of the supported AI unit; Supported AI unit association configuration information; Supported AI unit associated reference signal information. Among them, E1. AI unit support capability; used to characterize whether the first device has the capability to run an AI unit. It is understood that the AI ​​unit support capability can be explicitly indicated in the first capability information (e.g., by indicating the support capability of different AI units through a preset bitmap), or implicitly indicated (e.g., if the first capability information includes any one of the input, output, network conditions, associated configuration information, and associated reference signal information of AI unit A, it implicitly indicates that the first device has the capability to support AI unit A). This application does not limit this.

[0106] E2, at least one of the inputs, outputs, and network conditions of the supported AI unit; it is understood that if the inputs, outputs, and network conditions of the AI ​​unit are inconsistent with those of the first device, it may affect the accuracy of the target report of the AI ​​unit. For example, if the channel estimation AI unit only supports a 4-antenna configuration, but the first device has an 8-antenna configuration, the difference in input may cause the channel estimation AI unit to fail to operate on the first device or reduce the accuracy of the target report.

[0107] E3. Supported AI unit association configuration information; for details, please refer to the configuration information related to the first AI unit, which will not be elaborated here.

[0108] E4. Supported AI unit associated reference signal information. For details, please refer to the configuration information of the second signal associated with the first AI unit, which will not be elaborated here.

[0109] It is understandable that the first capability information may include any combination of multiple items in E1-E4 above, which facilitates more flexible determination of the target AI unit.

[0110] In some embodiments, the first capability information further includes at least one of the following: First-hand information; First state information; The first time information is used to indicate the time information associated with the first capability information or the time information when the first capability information changes; the first status information is used to indicate the status information associated with the first capability information or the status information of the first device.

[0111] Since the first capability information of the first device can change with the state of the first device, that is, the first capability information of the first device will change accordingly with time. In some embodiments, the first capability information can be abstracted as a function of the change of first time information and / or first state information, so that the target AI unit can be determined based on the first capability information corresponding to the time or state.

[0112] By adopting the above technical solution, the corresponding first capability information can be predicted based on the first time information and / or first state information of the first device, and the target AI unit can be determined based on the first information of the predicted corresponding first capability information. This can adapt to the state changes of the first device during the dynamic scheduling of the AI ​​unit, and further improve the accuracy and efficiency of the dynamic scheduling of the AI ​​unit.

[0113] In some embodiments, step S101 may specifically include at least one of the following: F1 determines whether to run the target AI unit and / or update the target report based on the priority of the target AI unit and / or the target report. F2 determines whether to run the target AI unit and / or update the target report based on the AI ​​resources used by the target AI unit.

[0114] Specifically, method F2 may include at least one of the following: F21 determines whether to run the target AI unit and / or update the target report based on the AI ​​storage resources occupied by the target AI unit; F22 determines whether to run the target AI unit and / or update the target report based on the AI ​​computing resources occupied by the target AI unit.

[0115] Specifically, method F21 may include at least one of the following: F211, determine whether to run the target AI unit and / or update the target report based on the AI ​​resources already occupied and the AI ​​resources occupied by the target AI unit; F212, determine whether to run the target AI unit and / or update the target report based on the unused AI resources and the AI ​​resources occupied by the target AI unit; F213, determine whether to run the target AI unit and / or update the target report based on the AI ​​resources already occupied as determined at the first reference time and the AI ​​resources occupied by the target AI unit; F214, determine whether to run the target AI unit and / or update the target report based on the unoccupied AI resources determined at the first reference time and the AI ​​resources occupied by the target AI unit.

[0116] In some embodiments, if the AI ​​resources occupied by the target AI units running simultaneously meet the first capability information supported by the first device (e.g., the AI ​​resources occupied by the target AI units running simultaneously are less than or equal to the M4 first unit AI resources that the first device can process at most at the same time), all target AI units can be run, that is, the first device can determine to run all target AI units.

[0117] If the AI ​​resources occupied by the simultaneous operation of the target AI units do not meet the first capability information supported by the first device (e.g., the AI ​​resources occupied by the simultaneous operation of the target AI units are greater than the maximum M4 first unit AI resources that the first device can process at the same time), the first device can determine the target AI units that will run simultaneously based on priority. For example, if the currently occupied resources are L1 first unit AI resources and there are Y target AI units, the first device can select X AI units that satisfy the following Formula 2 based on priority, where X is the largest integer that satisfies the following Formula 2.

[0118] (Formula 2) In the above formula, the unit is the first unit of AI resources, i.e., APU. M4 represents the number of first-unit AI resources required for the nth target AI unit, M4 represents the maximum number of first-unit AI resources that the first device can process simultaneously, and L1 represents the number of first-unit AI resources currently occupied.

[0119] The aforementioned X higher-priority target AI units can be executed, while the YX lower-priority target AI units cannot be scheduled to run. Alternatively, this can be understood as the aforementioned X higher-priority target reports being able to be updated, while the YX lower-priority target reports being unable to be updated.

[0120] In some embodiments, if the storage resources required by the target AI units to be run by the first device are less than or equal to the available storage resources of the first device, then all target AI units can be run, that is, the first device can determine to run all target AI units.

[0121] If the storage resources required by the target AI unit to be run by the first device are greater than the available storage resources of the first device, the first device can determine the portion of the target AI units that can be run based on priority.

[0122] For example, the total AI storage resources of the terminal are S, the currently occupied AI storage resources are S1, and the number of target AI units is Y. .

[0123] In some optional embodiments, the Y target AI units may include Y1 active target AI units and Y2 inactive target AI units (Y=Y1+Y2). The Y2 inactive target AI units can be released to save AI storage resources (the Y2 inactive target AI units can be loaded and executed after the Y1 target AI units have finished running).

[0124] In some alternative embodiments, the first device may select X AI units that satisfy Formula 3 below based on priority to load and run.

[0125] (Formula 3) in, AI storage resources required for the nth target AI unit.

[0126] The aforementioned X higher-priority target AI units can be loaded, while the YX lower-priority target AI units cannot be loaded.

[0127] It is understandable that the decision to run a target AI unit can be made based on its priority, the priority of the target report, or a weighted average of the target AI unit's priority and the target report's priority.

[0128] In some embodiments, the first device may also determine whether to run the target AI unit based on the AI ​​storage resources and AI computing resources occupied by the target AI unit. The specific steps may be as follows: Step 10: Based on the priority of the target AI unit and / or the priority of the target report, determine the X1 higher priority target AI units that can be loaded from the Y target AI units according to Formula 3.

[0129] Step 11: Based on the priority of the target AI unit and / or the priority of the target report, further determine the X2 higher priority target AI units that can be loaded from the X1 target AI units according to Formula 2.

[0130] In some embodiments, the first device may also determine whether to run the target AI unit based on the number of AI units that the first device can load and / or run simultaneously.

[0131] In some possible implementations, the first device can process up to N4 AI units at the same time. Currently, L2 AI units are running, and the number of target AI units is Y. Then, the first device can select X target AI units that satisfy the following formula four according to priority.

[0132] (Formula 4) In another possible implementation, the first device can process up to N4 AI units at the same time. Currently, L2 AI units are running. The number of target AI units is Y. If the priority of the target AI units and / or the target report is higher than the priority of any AI unit among the L2 AI units, then the first device can stop running the L2 AI units and select min(N4,Y) target AI units with higher priority from the Y target AI units according to their priority.

[0133] In another possible implementation, the first device can process up to N4 AI units at the same time. Currently, L2 AI units are running. The number of target AI units is Y. The priority of the target AI units and / or the target reports is higher than the priority of any AI unit among the L2 AI units. Then, the first device can stop the low-priority W1 AI units and select min(N4-L2+W1,Y) target AI units with higher priority from the Y target AI units according to their priority.

[0134] In another possible implementation, the first device can process up to N4 AI units at the same time. Currently, L2 AI units are running. The number of target AI units is Y, and the Y target AI units and the L2 AI units have AI units with the same service functions. Assuming the number is I(Y, L2), the first device can determine whether the remaining Y - I(Y, L2) target AI units can run.

[0135] Other embodiments not detailed in the technical solutions for determining whether to run the target AI unit and / or whether to update the target report can be derived through reasonable reasoning based on the above embodiments and knowledge in the art, and will not be elaborated upon here.

[0136] By adopting the above technical solution, it is possible to determine whether to run the target AI unit and / or update the target report based on at least one of the priority of the target AI unit, the priority of the target report, and the AI ​​resources occupied by the target AI unit. This achieves unified resource scheduling of the target AI unit, improves the applicability of AI technology in communication systems, and enhances the service capabilities and service quality of communication networks.

[0137] Figure 5 This illustration shows another flowchart of the control method provided in this disclosure, such as... Figure 5 As shown, step S101 may include the following steps.

[0138] In step S1012, the target AI unit is determined and / or the target report is updated based on the first information and the second information.

[0139] The second information includes at least one of the following: G1, the processing resources for the first signal; G2, Monitoring Configuration Information; G3, Monitoring Instruction Information; G4, Indication information for candidate AI units; G5, First Reference Information.

[0140] Understandably, in order to determine the target AI unit and / or update the target report, the first device also needs to evaluate the processing resources of the first signal (if the configuration information related to the target report includes the configuration information of the first signal, the target report is obtained by the first device based on the first signal) in order to more accurately determine the target AI unit.

[0141] In some embodiments, the first device may determine at least one second AI unit from the first AI units, for example, it may determine the at least one candidate second AI unit based on the indication information of the candidate AI units, run the second AI unit, and monitor and evaluate the performance of the second AI unit (i.e., verify and evaluate the second AI unit through a preset verification dataset), and select one or more AI units with the best performance among the second AI units as the target AI unit.

[0142] The operation monitoring of the AI ​​unit can be based on the monitoring configuration information in the second information mentioned above. Alternatively, the first device can also determine the corresponding monitoring configuration information from multiple preset monitoring configuration information based on the monitoring instruction information to monitor the operation of the AI ​​unit.

[0143] In some embodiments, the first reference information may be a first reference time.

[0144] It is understood that the second information may include any combination of multiple items in G1-G5 above, and this application will not elaborate on them one by one.

[0145] By adopting the above technical solution, the target AI unit can be identified and the target report updated based on the second information assistance, which can further improve the accuracy of identifying the target AI unit and also help improve the accuracy of unified resource scheduling of the target AI unit.

[0146] Figure 6 This illustration shows another flowchart of the control method provided in this disclosure, such as... Figure 6 As shown, step S101 may include the following steps.

[0147] In step S1013, the performance of at least one second AI unit is monitored.

[0148] The second AI unit may be determined by the first device from the first AI unit based on the indication information of the candidate AI units in the second information.

[0149] The first device can monitor the performance of one or more second AI units based on a preset validation set of data. Specifically, the input data in the validation set can be input into the second AI unit, and the output of the second AI unit can be compared with the labels in the validation set of data to monitor the performance of the second AI unit.

[0150] In step S1014, the target AI unit is determined from at least one second AI unit based on the first information and the performance of the second AI unit.

[0151] After monitoring the performance of one or more second AI units, a preset number of AI units with higher performance or AI units whose performance meets the standard can be selected as target AI units from at least one second AI unit based on the first information.

[0152] By adopting the above technical solution, the performance of one or more second AI units can be monitored. Based on the first information and the performance of one or more second AI units, the target AI unit can be determined, which can further improve the accuracy of determining the target AI unit and also help improve the accuracy of unified resource scheduling of the target AI unit.

[0153] Figure 7 This illustration shows another flowchart of the control method provided in this disclosure, such as... Figure 7 As shown, step S101 may include the following steps.

[0154] In step S1015, if the performance of a third AI unit in the second AI unit is lower than the second performance threshold, the third AI unit is released from the first device, or the state of the third AI unit is changed to an unavailable state or an inactive state; and / or, if the performance of a fourth AI unit in the second AI unit is lower than the third performance threshold, the operation and / or monitoring of the fourth AI unit is stopped.

[0155] In some embodiments, to ensure the smooth operation of the second AI unit, one or more of the following may be performed: Whether to run the second AI unit is determined based on the AI ​​storage resources occupied by the second AI unit; Whether to run the second AI unit is determined based on the priority of the second AI unit and / or the target report; Whether to run the second AI unit is determined based on the AI ​​resources already occupied and the AI ​​resources occupied by the second AI unit; Whether to run the second AI unit is determined based on the unused AI resources and the AI ​​resources occupied by the second AI unit; Whether to run the second AI unit is determined based on the occupied AI resources determined at the first reference time and the AI ​​resources occupied by the second AI unit; Whether to run the second AI unit is determined based on the unoccupied AI resources and the AI ​​resources of the second AI unit as determined in the first reference time.

[0156] By adopting the above technical solution, in the process of monitoring the performance of one or more second AI units, AI resources can be released and operation / monitoring can be stopped in a timely manner for second AI units with lower performance, which can effectively reduce the resource consumption in the process of determining the target AI unit.

[0157] Figure 8 This illustration shows another flowchart of the control method provided in this disclosure, such as... Figure 8 As shown, the method may also include the following steps.

[0158] In step S102, the performance of the target AI unit is monitored.

[0159] In some embodiments, the first device can monitor the performance of the target AI unit based on a preset validation set of data. Specifically, the input data in the validation set can be input into the target AI unit, and the output of the target AI unit can be compared with the labels in the validation set of data to monitor the performance of the target AI unit.

[0160] In some embodiments, the first device may determine the performance of the target AI unit based on the performance of signal measurement or data reception. Optionally, the performance of the second AI unit may be determined by combining information from the target AI unit and information from the second AI unit.

[0161] In step S103, if the performance of the target AI unit is lower than a first performance threshold, an updated target AI unit is determined from at least one second AI unit.

[0162] By adopting the above technical solution, when the performance of the target AI unit is lower than the first performance threshold, the updated target AI unit can be determined from at least one second AI unit, thereby enabling timely updates to the target AI unit and further improving the accuracy of determining the target AI unit.

[0163] In some embodiments, the method further includes at least one of the following: The performance of at least one second AI unit is monitored within the first window; Monitor the target AI unit within the second window; Run at least one second AI unit within the third window; Run the target AI unit in the fourth window.

[0164] It is understood that each type of window, including the first, second, third, and fourth windows, can include window configuration parameters such as the window's temporal starting position, window length, and window period. Window configuration parameters for different types of windows can be the same or different. Window configuration parameters for the same type of window corresponding to different AI units can be the same or different; this application does not impose any restrictions on this.

[0165] In some embodiments, the method further includes at least one of the following: If the resources in the current first window do not meet the performance monitoring conditions, monitor the performance of the second AI unit in the next first window. If the resources in the current second window do not meet the performance monitoring conditions, monitor the performance of the target AI unit in the next second window. If the resources in the current third window do not meet the resource requirements for running at least one second AI unit, then run at least one second AI unit in the next third window. If the resources in the current fourth window do not meet the resource requirements for running the target AI unit, the target AI unit will be run in the next fourth window.

[0166] By adopting the above technical solution, the operation and / or performance monitoring of the AI ​​unit can be flexibly controlled through window parameters, further improving the flexibility of AI unit management.

[0167] Optionally, the period of the third window can be longer than the fourth period, or the period of the second window can be longer than the first window.

[0168] Optionally, the length of the first window can be greater than that of the second window, or the length of the third window can be greater than that of the fourth window.

[0169] In some embodiments, the AI ​​resources occupied by the target AI unit also include at least one of the following: AI resources occupied by the second AI unit; Monitor the performance of the target AI unit; Monitor the performance of the second AI unit and the AI ​​resources it consumes.

[0170] In some embodiments, the method may further include: The decision to run the target AI unit and / or update the target report is based on the first processing resources used by the target report; or... The decision on whether to run the target AI unit and / or update the target report is based on the second processing resources used to acquire the input information of the target AI unit.

[0171] In some embodiments, the AI ​​resources occupied by the target AI unit are related to the number of cycles of the first signal or the number of resources of the first signal. For example, the shorter the cycle of the first signal, the more AI resources the target AI unit occupies; the more resources the first signal has, the more AI resources the target AI unit occupies. When the first signal includes multiple different reference signals, the AI ​​resources occupied by the target AI unit can also be related to the number of types of reference signals. The more types of reference signals the first signal includes, the more AI resources the target AI unit occupies.

[0172] By adopting the above technical solution, when determining whether to run the target AI unit and / or whether to update the target report, the resources occupied by the candidate second AI unit, the resources required to monitor the operation of the AI ​​unit, and the resources required to process signals can be further considered.

[0173] In some embodiments, considering the resources occupied by the input information of the target AI unit, the larger the number of cycles and the number of resources of the first signal, the more AI resources are occupied. This further determines the AI ​​resources occupied by the target AI unit and / or the target report, which can further improve the accuracy of identifying the target AI unit and also helps improve the accuracy of unified resource scheduling of the target AI unit. For example, the AI ​​resources of the target AI unit may be integer multiples of the signal cycle and the number of resources.

[0174] In some embodiments, the method may further include one of the following: If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device are released in the order of inactive state to active state; or, If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device that are inactive or active are released in order of priority.

[0175] Taking a device operating in a "load first, then select run" mode as an example, the first device already contains inactive AI units 1 and 2, and active AI units 3 and 4. If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device can be released sequentially: first AI units 3 and 4, then AI units 1 and 2. In some possible implementations, AI units with the same state can be released according to their priority. Alternatively, if it is determined that the storage resource requirements of the target AI unit are not met based on storage capacity information, the AI ​​units can be released sequentially according to their priority. If two AI units have the same priority, the inactive AI unit is released first.

[0176] In some embodiments, the method may further include at least one of the following: Based on priority, stop the fifth AI unit in the first device from operating; Deactivate the sixth AI unit in the first device according to priority; Release the seventh AI unit in the first device according to priority; Release the corresponding AI unit in the first device according to the monitoring instructions.

[0177] Among them, the fifth AI unit is an AI unit in the first device that is in the running state and has a priority lower than the target AI unit; the sixth AI unit is an AI unit in the first device that is in the active state and has a priority lower than the target AI unit; and the seventh AI unit is an AI unit in the first device that is in the inactive state and has a priority lower than the target AI unit.

[0178] It is understood that AI units that need to be stopped, released, or deactivated can be determined based on their priority, the priority of the target report corresponding to the AI ​​unit, or a weighted priority of the priority of the AI ​​unit and its corresponding target report. This application does not impose any restrictions on this.

[0179] By adopting the above technical solution, it is possible to stop operation, release, or deactivate the AI ​​units with lower priority in the first device in a timely manner, or release the corresponding AI units according to the monitoring instructions, thereby releasing resources in a timely manner, facilitating the scheduling of target AI units, better realizing unified configuration management of AI units, and achieving flexible resource scheduling and allocation.

[0180] In some embodiments, the method may further include one of the following: Receive capability information of the second device from the second device; Receive the remaining AI resources of the second device from the second device; Receive the number of AI resources associated with one or more AI units from the second device; Send the first capability information to the second device; Send the remaining amount of AI resources to the second device; Send the number of AI resources associated with one or more AI units to the second device.

[0181] It is understandable that an AI unit can be a two-sided model involving two devices, with interaction between the two devices. The two devices can interact through at least one of the following: indication information associated with the AI ​​unit (e.g., data_ID) and indication information related to the model structure (e.g., model_ID). On this basis, the first and second devices can also interact with capability information, the number of remaining AI resources, and at least one of the following: the number of AI resources associated with the AI ​​unit. This facilitates the first device to further consider the resource status of related devices when performing unified configuration management of the AI ​​unit and achieving flexible resource scheduling and allocation.

[0182] It is understood that the first device can be a device that is logically directly connected to the second device, such as a terminal and a base station, or it can be a device that is not logically directly connected, such as a terminal and a core network device (the corresponding information can be transparently transmitted through the base station), or a base station and a server (the corresponding information can be transparently transmitted through the core network). This application does not impose any restrictions on this.

[0183] By adopting the above technical solution, the resource status of related devices can be further considered when performing unified configuration management of AI units and achieving flexible resource scheduling and allocation, thus better realizing unified configuration management of AI units and achieving flexible resource scheduling and allocation.

[0184] The control method provided in this application can be executed by a control device. This application uses the example of a control device executing the control method to illustrate the control device provided in this application.

[0185] Figure 9 This diagram illustrates a control device according to an embodiment of this application. As an example, the control device may be a communication device or a component within a communication device, such as a chip. The communication device may be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal may include, but is not limited to, the type of terminal 11 listed above, and the network-side device may include, but is not limited to, the type of network-side device 12 listed above. This embodiment of the application does not impose specific limitations.

[0186] For details, see Figure 9 The control device 900 includes a processing module 901, used to determine the target AI unit and / or update the target report based on the first information; The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The first device can utilize AI units; The AI ​​unit activated by the first device; The first device will be the AI ​​unit that runs; AI unit used to obtain target reports; Reference AI unit used to obtain target reports; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report; The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

[0187] Optionally, the configuration information related to the target report includes at least one of the following: The configuration information of the first signal, wherein if the configuration information related to the target report includes the configuration information of the first signal, the target report is obtained by the first device based on the first signal; The content of the target report; The time-domain location where the target report was sent; The time-domain type of the target report sent; The frequency of target report submissions; Prioritize target reports; The target report contains instructions and information.

[0188] Optionally, the configuration information related to the first AI unit includes at least one of the following: The instruction information associated with the first AI unit; Configuration information of the second signal associated with the first AI unit; The complexity of the first AI unit; AI resources occupied by the first AI unit; The first AI unit occupies the first unit's AI resources; The input information associated with the first AI unit; The output information associated with the first AI unit; The priority of the first AI unit.

[0189] Optionally, the AI ​​resources for the first unit are determined according to any one of the following: The first AI resource is preset; wherein, the first AI resource is related to the first reference report; A pre-defined second AI resource; wherein the second AI resource is related to the first reference AI function; A pre-defined third AI resource, wherein the third AI resource is related to the first reference AI unit; The first preset number of floating-point operations (FLOPs); The second preset number of AI parameters.

[0190] Optionally, the AI ​​resources of the first unit include at least one of the following: Unit AI storage resources; Unit AI computing resources.

[0191] Optionally, the first capability information includes at least one of the following: Storage capacity information; Processing capacity information; Processing efficiency information; The storage capacity information includes that the first device can store up to M1 AI resources, and / or that the first device can store up to N1 baseline AI resources. The processing capacity information includes at least one of the following: The first device can process a maximum of M2 AI resources; The first device can process a maximum of N2 baseline AI units; The first device can process up to M3 AI resources in the first time. The first device can process a maximum of N3 baseline AI units in the first time. The first device can process up to M4 AI resources at the same time. The first device can process a maximum of N4 baseline AI units at the same time; The first device has already occupied M5 units of AI resources; The first device has already occupied N5 baseline AI units; The first device has already occupied M6 AI unit resources at the first reference time; The first device has already occupied N6 reference AI units at the first reference time; The processing efficiency information includes the first processing time for processing a third preset number of AI unit resources, and / or the second processing time for processing a fourth preset number of baseline AI units.

[0192] Optionally, the baseline AI unit is assumed to be an AI unit that requires a fifth preset number of AI units.

[0193] Optionally, the first device includes K state information corresponding to different AI capabilities, and the first capability information includes at least one of the following: The storage capacity information corresponding to the K status information; The processing capacity information corresponding to the K status information; The processing efficiency information corresponds to each of the K state information.

[0194] Optionally, the processing module 901 is also used for: The target AI unit is determined based on the first information and the status information of the first device.

[0195] Optionally, the first capability information includes the support capabilities of the AI ​​unit. If there is an AI unit supported by the first device, the first capability information may also include at least one of the following: At least one of the inputs, outputs, and network conditions of the supported AI unit; Supported AI unit association configuration information; Supported AI unit associated reference signal information.

[0196] Optionally, the first capability information may also include at least one of the following: First-hand information; First state information; The first time information is used to indicate the time information associated with the first capability information or the time information when the first capability information changes; the first status information is used to indicate the status information associated with the first capability information or the status information of the first device.

[0197] Optionally, the processing module 901 is also used for: Determine whether to run the target AI unit and / or update the target report based on the priority of the target AI unit and / or the target report; and / or, Whether to run the target AI unit and / or update the target report is determined based on the AI ​​resources used by the target AI unit.

[0198] Optionally, the processing module 901 is also used for: Determine whether to run the target AI unit and / or update the target report based on the AI ​​storage resources occupied by the target AI unit; and / or, Whether to run the target AI unit and / or update the target report is determined based on the AI ​​computing resources used by the target AI unit.

[0199] Optionally, the processing module 901 is also used for: Determine whether to run the target AI unit and / or update the target report based on the AI ​​resources already occupied and the AI ​​resources occupied by the target AI unit; Determine whether to run the target AI unit and / or update the target report based on the unused AI resources and the AI ​​resources occupied by the target AI unit; The decision on whether to run the target AI unit and / or update the target report is based on the AI ​​resources already occupied and the AI ​​resources occupied by the target AI unit as determined at the first reference time. Whether to run the target AI unit and / or update the target report is determined based on the unused AI resources identified at the first reference time and the AI ​​resources occupied by the target AI unit.

[0200] Optionally, the processing module 901 is also used for: Based on the first and third information, identify the target AI unit and / or update the target report; The third information includes any one of the following: AI resource utilization rate; Busyness rate.

[0201] Optionally, the processing module 901 is also used for: If the AI ​​resource utilization rate is greater than or equal to a first threshold, or if the busy rate is greater than or equal to a second threshold, send a first indication message regarding the AI ​​resource utilization rate; or... If the AI ​​resource occupancy rate is greater than or equal to the first threshold, or if the busy rate is greater than or equal to the fourth threshold, a second indication message indicating that the AI ​​resource is busy or there is no available AI resource is sent.

[0202] The processing time of the target AI unit is determined based on either a first range of AI resource utilization or a second range of busy rate.

[0203] Optionally, the processing module 901 is also used for: The target AI unit is determined and / or the target report is updated based on the first and second information; The second information includes at least one of the following: Resources for processing the first signal; Monitoring configuration information; Monitoring instructions; Indication information for candidate AI units; First reference information.

[0204] Optionally, the processing module 901 is also used for: Monitor the performance of at least one second AI unit; The target AI unit is determined from at least one second AI unit based on the performance of the first information and the second AI unit.

[0205] Optionally, the processing module 901 is also used for: Monitor the performance of the target AI unit; If the performance of the target AI unit is lower than a first performance threshold, an updated target AI unit is determined from at least one second AI unit.

[0206] Optionally, the processing module 901 is also used for: If the performance of the third AI unit in the second AI unit is lower than the second performance threshold, the third AI unit will be released from the first device, or the state of the third AI unit will be changed to an unavailable or inactive state; and / or, If the performance of the fourth AI unit in the second AI unit is lower than the third performance threshold, stop the operation and / or monitoring of the fourth AI unit.

[0207] Optionally, the processing module 901 is also used for at least one of the following: The performance of at least one second AI unit is monitored within the first window; Monitor the target AI unit within the second window; Run at least one second AI unit within the third window; Run the target AI unit in the fourth window.

[0208] Optionally, the processing module 901 is also used for at least one of the following: If the resources in the current first window do not meet the performance monitoring conditions, monitor the performance of the second AI unit in the next first window. If the resources in the current second window do not meet the performance monitoring conditions, monitor the performance of the target AI unit in the next second window. If the resources in the current third window do not meet the resource requirements for running at least one second AI unit, then run at least one second AI unit in the next third window. If the resources in the current fourth window do not meet the resource requirements for running the target AI unit, the target AI unit will be run in the next fourth window.

[0209] Optionally, the AI ​​resources occupied by the target AI unit may also include at least one of the following: AI resources occupied by one or more second AI units; Monitor the performance of the target AI unit; Monitor the AI ​​resources consumed by the performance of one or more second AI units.

[0210] Optionally, the processing module 901 is also used for: The decision to run the target AI unit and / or update the target report is based on the first processing resources used by the target report; or... The decision on whether to run the target AI unit and / or update the target report is based on the second processing resources used to acquire the input information of the target AI unit.

[0211] Optionally, the AI ​​resources occupied by the target AI unit are related to the number of cycles of the first signal or the number of resources of the first signal.

[0212] Optionally, the processing module 901 is also used for: If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device are released in the order of inactive to active states; or, If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device that are inactive or active are released in order of priority.

[0213] Optionally, the processing module 901 is also used for: Based on priority, stop the fifth AI unit in the first device from operating; Deactivate the sixth AI unit in the first device according to priority; Release the seventh AI unit in the first device according to priority; Release the corresponding AI unit in the first device according to the monitoring instructions.

[0214] Among them, the fifth AI unit is an AI unit in the first device that is in the running state and has a lower priority than the target AI unit; the sixth AI unit is an AI unit in the first device that is in the active state and has a lower priority than the target AI unit; and the seventh AI unit is an AI unit in the first device that is in the inactive state and has a lower priority than the target AI unit.

[0215] Optionally, the processing module 901 is also used for: Receive capability information of the second device from the second device; Receive the remaining AI resources from the second device; Receive the number of AI resources associated with one or more AI units from the second device; Send the first capability information to the second device; Send the remaining amount of AI resources to the second device; Send the number of AI resources associated with one or more AI units to the second device.

[0216] The control device provided in this application embodiment can implement the various processes implemented in the method embodiment of the first aspect and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0217] like Figure 10 As shown in the illustration, this application also provides an electronic device 1000, including a processor 1001 and a memory 1002. The memory 1002 stores a program or instructions that can run on the processor 1001. For example, when the electronic device 1000 is a terminal, the program or instructions executed by the processor 1001 implement the various steps of the above-described control method embodiments and achieve the same technical effect. When the electronic device 1000 is a network-side device, the program or instructions executed by the processor 1001 implement the various steps of the above-described control method embodiments and achieve the same technical effect. When the electronic device 1000 is a server, the program or instructions executed by the processor 1001 implement the various steps of the above-described control method embodiments and achieve the same technical effect. To avoid repetition, further details are omitted here.

[0218] This application also provides a terminal, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps in the method embodiment shown in the first aspect. This terminal embodiment corresponds to the above-described terminal-side method embodiment; all implementation processes and methods of the above-described method embodiments can be applied to this terminal embodiment and achieve the same technical effect. The terminal can be... Figure 9 The control device shown. Specifically, Figure 11 A schematic diagram of the hardware structure of a terminal to implement an embodiment of this application. The terminal 1100 includes, but is not limited to, at least some of the following components: radio frequency unit 1101, network module 1102, audio output unit 1103, input unit 1104, sensor 1105, display unit 1106, user input unit 1107, interface unit 1108, memory 1109, and processor 1110.

[0219] Those skilled in the art will understand that the terminal 1100 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor x10 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 11The terminal structure shown does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0220] It should be understood that, in this embodiment, the input unit 1104 may include a graphics processor 11041 and a microphone 11042. The graphics processor 11041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode. The display unit 1106 may include a display panel 11061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like. The user input unit 1107 includes at least one of a touch panel 11071 and other input devices 11072. The touch panel 11071 is also called a touch screen. The touch panel 11071 may include a touch detection device and a touch controller. Other input devices 11072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

[0221] In this embodiment, after receiving downlink data from the network-side device, the radio frequency unit 1101 can transmit it to the processor 1110 for processing; in addition, the radio frequency unit 1101 can send uplink data to the network-side device. Typically, the radio frequency unit 1101 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.

[0222] The memory x09 can be used to store software programs or instructions, as well as various data. The memory 1109 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 1109 may include volatile memory or non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory x09 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0223] Processor x10 may include one or more processing units; optionally, processor 1110 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may not be integrated into processor 1110.

[0224] The processor 1110 is used to determine the target AI unit and / or update the target report based on the first information.

[0225] It is understood that the implementation process of each implementation method mentioned in this embodiment can refer to the relevant description of the first aspect method embodiment and achieve the same or corresponding technical effects. To avoid repetition, it will not be described again here.

[0226] This application also provides a network-side device, including a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the method embodiment shown in the first aspect embodiment. This network-side device embodiment corresponds to the above-described network-side device method embodiment. All implementation processes and methods of the above-described method embodiments can be applied to this network-side device embodiment and achieve the same technical effects.

[0227] Specifically, embodiments of this application also provide a network-side device, which can be... Figure 9 The control device shown. (As shown) Figure 12 As shown, the network-side device 1200 includes: an antenna 1201, a radio frequency (RF) device 1202, a baseband device 1203, a processor 1204, and a memory 1205. The antenna 1201 is connected to the RF device 1202. In the uplink direction, the RF device 1202 receives information through the antenna 1201 and transmits the received information to the baseband device 1203 for processing. In the downlink direction, the baseband device 1203 processes the information to be transmitted and sends it to the RF device 1202. The RF device 1202 processes the received information and transmits it through the antenna 1201.

[0228] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 1203, which includes a baseband processor.

[0229] The baseband device 1203 may include at least one baseband board, on which multiple chips are disposed, as shown in FIG. y. One of the chips is, for example, a baseband processor, which is connected to the memory y5 via a bus interface to call the program or instructions in the memory y5 to execute the network-side device operation shown in the above method embodiment.

[0230] The network-side device may also include a network interface 1206, such as a Common Public Radio Interface (CPRI).

[0231] The processor 1204 is used to determine the target AI unit and / or update the target report based on the first information.

[0232] In addition, the network-side device 1200 of this application embodiment also includes: a program or instructions stored in memory 1205 and executable on processor 1204. The processor 1204 calls the program or instructions in memory 1205 to execute the steps of the method embodiment shown in the first aspect embodiment and achieve the same technical effect. To avoid repetition, it will not be described in detail here.

[0233] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described control method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0234] The processor mentioned above is the processor in the terminal described in the above embodiments, the processor in the network-side device, or the processor in the server. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0235] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Furthermore, it should be pointed out that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0236] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.), and the computer software product includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[0237] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

1. A control method, characterized in that, Applied to a first device, the method includes: Based on the initial information, identify the target AI unit and / or update the target report; The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The AI ​​units that can be applied to the first device; The AI ​​unit activated by the first device; The AI ​​unit that the first device will run; AI unit used to obtain the target report; Reference AI unit used to obtain the target report; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report. The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

2. The method according to claim 1, characterized in that, The configuration information related to the target report includes at least one of the following: The configuration information of the first signal, wherein if the configuration information related to the target report includes the configuration information of the first signal, the target report is obtained by the first device based on the first signal; The content of the target report; The time-domain location where the target report was sent; The time-domain type of the target report sent; The frequency of target report submissions; Prioritize target reports; The target report contains instructions and information.

3. The method according to claim 1, characterized in that, The configuration information related to the first AI unit includes at least one of the following: The indication information associated with the first AI unit; Configuration information of the second signal associated with the first AI unit; The complexity of the first AI unit; The AI ​​resources occupied by the first AI unit; The first AI unit occupies the first unit's AI resources; The input information associated with the first AI unit; The output information associated with the first AI unit; The priority of the first AI unit.

4. The method according to any one of claims 1 to 3, characterized in that, The AI ​​resources of the first unit are determined according to any one of the following: The first AI resource is preset; wherein, the first AI resource is related to the first reference report; A pre-defined second AI resource; wherein the second AI resource is related to the first reference AI function; A pre-defined third AI resource, wherein the third AI resource is related to the first reference AI unit; The first preset number of floating-point operations (FLOPs); The second preset number of AI parameters.

5. The method according to any one of claims 1 to 3, characterized in that, The first unit's AI resources include at least one of the following: Unit AI storage resources; Unit AI computing resources.

6. The method according to claim 1, characterized in that, The first capability information includes at least one of the following: Storage capacity information; Processing capacity information; Processing efficiency information; The storage capacity information includes that the first device can store up to M1 first unit AI resources, and / or that the first device can store up to N1 base AI units. The processing capability information includes at least one of the following: The first device can process a maximum of M2 first unit AI resources; The first device can process a maximum of N2 baseline AI units; The first device can process a maximum of M3 first unit AI resources in the first time period; The first device can process a maximum of N3 baseline AI units in the first time period; The first device can process a maximum of M4 first unit AI resources at the same time; The first device can process a maximum of N4 baseline AI units simultaneously; The first device has already occupied M5 first-unit AI resources; The first device has already occupied N5 baseline AI units; The first device has already occupied M6 first-unit AI resources at the first reference time; The first device has already occupied N6 reference AI units at the first reference time; The processing efficiency information includes the first processing time for processing a third preset number of first unit AI resources, and / or the second processing time for processing a fourth preset number of benchmark AI units.

7. The method according to claim 6, characterized in that, The baseline AI unit is assumed to be an AI unit that requires a fifth preset number of first unit AI resources.

8. The method according to claim 6, characterized in that, The first device includes K state information corresponding to different AI capabilities, and the first capability information includes at least one of the following: The K status information respectively correspond to the storage capacity information; The processing capability information corresponds to each of the K status information; The processing efficiency information corresponds to each of the K state information.

9. The method according to claim 8, characterized in that, The step of determining the target AI unit based on the first information includes: The target AI unit is determined based on the first information and the status information of the first device.

10. The method according to claim 1, characterized in that, The first capability information includes the support capabilities of the AI ​​unit. If an AI unit supported by the first device exists, the first capability information further includes at least one of the following: At least one of the inputs, outputs, and network conditions of the supported AI unit; Supported AI unit association configuration information; Supported AI unit associated reference signal information.

11. The method according to claim 1, characterized in that, The first capability information also includes at least one of the following: First-hand information; First state information; Wherein, the first time information is used to indicate the time information associated with the first capability information or the time information when the first capability information changes; the first status information is used to indicate the status information associated with the first capability information or the status information of the first device.

12. The method according to any one of claims 1 to 11, characterized in that, The step of determining the target AI unit and / or updating the target report based on the first information includes: Determine whether to run the target AI unit and / or update the target report based on the priority of the target AI unit and / or the target report; and / or, The decision on whether to run the target AI unit and / or update the target report is based on the AI ​​resources occupied by the target AI unit.

13. The method according to claim 12, characterized in that, The step of determining whether to run the target AI unit and / or update the target report based on the AI ​​resources occupied by the target AI unit includes: Determine whether to run the target AI unit and / or update the target report based on the AI ​​storage resources occupied by the target AI unit; and / or, The decision on whether to run the target AI unit and / or update the target report is based on the AI ​​computing resources occupied by the target AI unit.

14. The method according to claim 13, characterized in that, The step of determining whether to run the target AI unit and / or update the target report based on the AI ​​resources occupied by the target AI unit further includes at least one of the following: Determine whether to run the target AI unit and / or update the target report based on the AI ​​resources already occupied and the AI ​​resources occupied by the target AI unit; Determine whether to run the target AI unit and / or update the target report based on the unused AI resources and the AI ​​resources occupied by the target AI unit; Based on the AI ​​resources already occupied as determined at the first reference time and the AI ​​resources occupied by the target AI unit, determine whether to run the target AI unit and / or whether to update the target report; Whether to run the target AI unit and / or update the target report is determined based on the unoccupied AI resources determined at the first reference time and the AI ​​resources occupied by the target AI unit.

15. The method according to claim 1, characterized in that, The step of determining the target AI unit and / or updating the target report based on the first information includes: Based on the first and second information, identify the target AI unit and / or update the target report; The second information includes at least one of the following: Resources for processing the first signal; Monitoring configuration information; Monitoring instructions; Indication information for candidate AI units; First reference information.

16. The method according to claim 1, characterized in that, The step of determining the target AI unit and / or updating the target report based on the first information includes: Based on the first and third information, identify the target AI unit and / or update the target report; The third information includes any one of the following: AI resource utilization rate; Busyness rate.

17. The method according to claim 16, characterized in that, The process of determining the target AI unit and / or updating the target report based on the first and third information includes: If the AI ​​resource occupancy rate is greater than or equal to a first threshold, or if the busy rate is greater than or equal to a second threshold, send a first indication of the AI ​​resource occupancy rate; or... If the AI ​​resource occupancy rate is greater than or equal to a third threshold, or if the busy rate is greater than or equal to a fourth threshold, a second indication message indicating that AI resources are busy or there are no available AI resources is sent; or... The processing time of the target AI unit is determined based on a first range of AI resource occupancy or a second range of busy rate.

18. The method according to any one of claims 1 to 17, characterized in that, The step of determining the target AI unit based on the first information includes: Monitor the performance of at least one second AI unit; The target AI unit is determined from the at least one second AI unit based on the first information and the performance of the second AI unit.

19. The method according to claim 18, characterized in that, The method further includes: Monitor the performance of the target AI unit; If the performance of the target AI unit is lower than a first performance threshold, an updated target AI unit is determined from the at least one second AI unit.

20. The method according to claim 18, characterized in that, The method further includes: If the performance of the third AI unit in the second AI unit is lower than the second performance threshold, the third AI unit is released from the first device, or the state of the third AI unit is changed to an unavailable or inactive state; and / or, If the performance of the fourth AI unit in the second AI unit is lower than the third performance threshold, the operation and / or monitoring of the fourth AI unit shall be stopped.

21. The method according to claim 18, characterized in that, The method further includes at least one of the following: The performance of at least one of the second AI units is monitored within the first window; Monitor the target AI unit within the second window; At least one of the second AI units is run within the third window; Run the target AI unit in the fourth window.

22. The method according to claim 21, characterized in that, The method further includes at least one of the following: If the resources in the current first window do not meet the performance monitoring conditions, monitor the performance of the second AI unit in the next first window. If the resources in the current second window do not meet the performance monitoring conditions, monitor the performance of the target AI unit in the next second window. If the resources in the current third window do not meet the resource requirements for running at least one of the second AI units, then run at least one of the second AI units in the next third window. If the resources in the current fourth window do not meet the resource requirements for running the target AI unit, the target AI unit will be run in the next fourth window.

23. The method according to claim 18, characterized in that, The AI ​​resources occupied by the target AI unit also include at least one of the following: AI resources occupied by the second AI unit; Monitor the performance of the target AI unit; Monitor the performance of the second AI unit and the AI ​​resources it consumes.

24. The method according to claim 1, characterized in that, The method further includes: Based on the first processing resources occupied by the target report, determine whether to run the target AI unit and / or whether to update the target report; or... Based on the second processing resources used to acquire the input information of the target AI unit, it is determined whether to run the target AI unit and / or whether to update the target report.

25. The method according to any one of claims 1 to 24, characterized in that, The AI ​​resources occupied by the target AI unit are related to the number of cycles of the first signal or the number of resources of the first signal.

26. The method according to claim 1, characterized in that, The method further includes: If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device are released in the order of inactive state to active state; or, If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device that are inactive or active are released in order of priority.

27. The method according to claim 1, characterized in that, The method further includes at least one of the following: The fifth AI unit in the first device will be stopped from operating according to priority. The sixth AI unit in the first device was deactivated based on priority. Release the seventh AI unit in the first device according to priority; Release the corresponding AI unit in the first device according to the monitoring instructions; The fifth AI unit is an AI unit in the first device that is in a running state and has a lower priority than the target AI unit. The sixth AI unit is an AI unit in the first device that is in an active state and has a lower priority than the target AI unit. The seventh AI unit is an AI unit in the first device that is in an inactive state and has a lower priority than the target AI unit.

28. The method according to claim 1, characterized in that, The method further includes at least one of the following: Receive capability information of the second device from the second device; Receive the remaining AI resources of the second device from the second device; Receive the number of AI resources associated with one or more AI units from the second device; Send the first capability information to the second device; Send the remaining amount of AI resources to the second device; Send the number of AI resources associated with one or more AI units to the second device.

29. A control device, characterized in that, Applied to the first device, the device includes: The processing module is used to determine the target AI unit and / or update the target report based on the first information; The target AI unit is one or more of the first AI units, and the first AI unit includes at least one of the following: The AI ​​units that can be applied to the first device; The AI ​​unit activated by the first device; The AI ​​unit that the first device will run; AI unit used to obtain the target report; Reference AI unit used to obtain the target report; The first information includes at least one of the following: The first configuration information includes at least one of the configuration information related to the first AI unit and the configuration information related to the target report. The first capability information of the first device includes capability information related to the AI ​​resources of the first unit.

30. The apparatus according to claim 29, characterized in that, The first device includes K state information corresponding to different AI capabilities, and the first capability information includes at least one of the following: The storage capacity information corresponding to the K status information; The processing capability information corresponding to the K status information; The processing efficiency information corresponds to each of the K state information.

31. The apparatus according to claim 30, characterized in that, The processing module is also used for: The target AI unit is determined based on the first information and the status information of the first device.

32. The apparatus according to any one of claims 29 to 31, characterized in that, The processing module is also used for: Determine whether to run the target AI unit and / or update the target report based on the priority of the target AI unit and / or the target report; and / or, The decision on whether to run the target AI unit and / or update the target report is based on the AI ​​resources occupied by the target AI unit.

33. The apparatus according to claim 32, characterized in that, The processing module is also used for: Determine whether to run the target AI unit and / or update the target report based on the AI ​​storage resources occupied by the target AI unit; and / or, The decision on whether to run the target AI unit and / or update the target report is based on the AI ​​computing resources occupied by the target AI unit.

34. The apparatus according to claim 33, characterized in that, The processing module is also used for: Determine whether to run the target AI unit and / or update the target report based on the AI ​​resources already occupied and the AI ​​resources occupied by the target AI unit; Determine whether to run the target AI unit and / or update the target report based on the unused AI resources and the AI ​​resources occupied by the target AI unit; Based on the AI ​​resources already occupied as determined at the first reference time and the AI ​​resources occupied by the target AI unit, determine whether to run the target AI unit and / or whether to update the target report; Whether to run the target AI unit and / or update the target report is determined based on the unoccupied AI resources determined at the first reference time and the AI ​​resources occupied by the target AI unit.

35. The apparatus according to claim 29, characterized in that, The processing module is also used for: Based on the first and second information, identify the target AI unit and / or update the target report; The second information includes at least one of the following: Resources for processing the first signal; Monitoring configuration information; Monitoring instructions; Indication information for candidate AI units; First reference information.

36. The apparatus according to claim 29, characterized in that, The processing module is also used for: Based on the first and third information, identify the target AI unit and / or update the target report; The third information includes any one of the following: AI resource utilization rate; Busyness rate.

37. The apparatus according to claim 36, characterized in that, The processing module is also used for: If the AI ​​resource occupancy rate is greater than or equal to a first threshold, or if the busy rate is greater than or equal to a second threshold, send a first indication of the AI ​​resource occupancy rate; or... If the AI ​​resource occupancy rate is greater than or equal to the third threshold, or if the busy rate is greater than or equal to the fourth threshold, a second indication message indicating that the AI ​​resource is busy or there is no available AI resource is sent. The processing time of the target AI unit is determined based on a first range of AI resource occupancy or a second range of busy rate.

38. The apparatus according to any one of claims 29 to 37, characterized in that, The processing module is also used for: Monitor the performance of at least one second AI unit; The target AI unit is determined from the at least one second AI unit based on the first information and the performance of the second AI unit.

39. The apparatus according to claim 38, characterized in that, The processing module is also used for: Monitor the performance of the target AI unit; If the performance of the target AI unit is lower than a first performance threshold, an updated target AI unit is determined from the at least one second AI unit.

40. The apparatus according to claim 38, characterized in that, include: If the performance of the third AI unit in the second AI unit is lower than the second performance threshold, the third AI unit is released from the first device, or the state of the third AI unit is changed to an unavailable or inactive state; and / or, If the performance of the fourth AI unit in the second AI unit is lower than the third performance threshold, the operation and / or monitoring of the fourth AI unit shall be stopped.

41. The apparatus according to claim 38, characterized in that, The processing module is also used for at least one of the following: The performance of at least one of the second AI units is monitored within the first window; Monitor the target AI unit within the second window; At least one of the second AI units is run within the third window; Run the target AI unit in the fourth window.

42. The apparatus according to claim 41, characterized in that, The processing module is also used for at least one of the following: If the resources in the current first window do not meet the performance monitoring conditions, monitor the performance of the second AI unit in the next first window. If the resources in the current second window do not meet the performance monitoring conditions, monitor the performance of the target AI unit in the next second window. If the resources in the current third window do not meet the resource conditions for running at least one of the second AI units, then at least one of the second AI units shall be run in the next third window. If the resources in the current fourth window do not meet the resource requirements for running the target AI unit, the target AI unit will be run in the next fourth window.

43. The apparatus according to claim 29, characterized in that, The processing module is also used for: Based on the first processing resources occupied by the target report, determine whether to run the target AI unit and / or whether to update the target report; or... Based on the second processing resources used to acquire the input information of the target AI unit, it is determined whether to run the target AI unit and / or whether to update the target report.

44. The apparatus according to claim 29, characterized in that, The processing module is also used for: If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device are released in the order of inactive state to active state; or, If, based on storage capacity information, it is determined that the storage resource requirements of the target AI unit are not met, the AI ​​units in the first device that are inactive or active are released in order of priority.

45. The apparatus according to claim 29, characterized in that, The processing module is also used for at least one of the following: The fifth AI unit in the first device will be stopped from operating according to priority. The sixth AI unit in the first device was deactivated based on priority. Release the seventh AI unit in the first device according to priority; Release the corresponding AI unit in the first device according to the monitoring instructions; The fifth AI unit is an AI unit in the first device that is in a running state and has a lower priority than the target AI unit. The sixth AI unit is an AI unit in the first device that is in an active state and has a lower priority than the target AI unit. The seventh AI unit is an AI unit in the first device that is in an inactive state and has a lower priority than the target AI unit.

46. ​​The apparatus according to claim 29, characterized in that, The processing module is also used for at least one of the following: Receive capability information of the second device from the second device; Receive the remaining AI resources of the second device from the second device; Receive the number of AI resources associated with one or more AI units from the second device; Send the first capability information to the second device; Send the remaining amount of AI resources to the second device; Send the number of AI resources associated with one or more AI units to the second device.

47. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the control method as described in any one of claims 1 to 28.

Images