WO2024125525A1 - Ai算力上报方法、终端及网络侧设备 - Google Patents

Ai算力上报方法、终端及网络侧设备 Download PDF

Info

Publication number
WO2024125525A1
WO2024125525A1 PCT/CN2023/138240 CN2023138240W WO2024125525A1 WO 2024125525 A1 WO2024125525 A1 WO 2024125525A1 CN 2023138240 W CN2023138240 W CN 2023138240W WO 2024125525 A1 WO2024125525 A1 WO 2024125525A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
computing power
terminal
computing
power information
Prior art date
Application number
PCT/CN2023/138240
Other languages
English (en)
French (fr)
Inventor
杨昂
孙鹏
Original Assignee
维沃移动通信有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 维沃移动通信有限公司 filed Critical 维沃移动通信有限公司
Publication of WO2024125525A1 publication Critical patent/WO2024125525A1/zh

Links

Definitions

  • the present application belongs to the field of communication technology, and specifically relates to an AI computing power reporting method, terminal and network side equipment.
  • AI artificial intelligence
  • the network side can instruct the user equipment (UE) to use a specific AI model.
  • UE user equipment
  • the network side cannot accurately estimate the remaining AI computing power of the UE, resulting in low utilization of the UE AI computing power, affecting the performance of the communication system.
  • the embodiments of the present application provide an AI computing power reporting method, a terminal, and a network-side device, which can improve the performance of a communication system.
  • a method for reporting AI computing power comprising:
  • the terminal obtains first AI computing power information
  • the terminal sends the first AI computing power information to the network side device
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • a method for reporting AI computing power comprising:
  • the network side device receives the first AI computing power information sent by the terminal
  • the network-side device acquires second AI computing power information corresponding to the terminal based on the first AI computing power information; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network-side device;
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • an AI computing power reporting device comprising:
  • a first acquisition module used to acquire first AI computing power information
  • a sending module configured to send the first AI computing power information to a network side device
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • an AI computing power reporting device comprising:
  • a receiving module configured to receive first AI computing power information sent by a terminal
  • a second acquisition module configured to acquire second AI computing power information corresponding to the terminal based on the first AI computing power information; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device;
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • a terminal comprising a processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.
  • a terminal comprising a processor and a communication interface; wherein the processor is used to: obtain first AI computing power information, and the communication interface is used to: send the first AI computing power information to a network side device; wherein the first AI computing power information is used to indicate at least one of the following: the AI model computing resources currently remaining in the terminal; the AI model computing resources currently available to the terminal; all the AI model computing resources of the terminal; all the AI model computing resources of the terminal that can be used for wireless communication.
  • a network side device which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the second aspect are implemented.
  • a network side device comprising a processor and a communication interface; wherein the communication interface is used to: receive first AI computing power information sent by a terminal, and the processor is used to: obtain second AI computing power information corresponding to the terminal based on the first AI computing power information; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device; wherein the first AI computing power information is used to indicate at least one of the following: the AI model computing resources currently remaining by the terminal; the AI model computing resources currently available to the terminal; all the AI model computing resources of the terminal; all the AI model computing resources of the terminal that can be used for wireless communication.
  • an AI computing power reporting system including: a terminal and a network side device, wherein the terminal can be used to execute the steps of the method described in the first aspect, and the network side device can be used to execute the steps of the method described in the second aspect.
  • a readable storage medium on which a program or instruction is stored.
  • the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.
  • a chip comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method described in the first aspect, or to implement the method described in the second aspect.
  • a computer program/program product is provided, wherein the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement the steps of the method described in the first aspect, or to implement the steps of the method described in the second aspect.
  • the terminal obtains the remaining AI models that can be used for AI model related operations.
  • the computing resources that is, the first AI computing power information, are then reported to the network side device, so that the network side device obtains the accurate remaining computing power of the terminal, so that the network side device can perform AI configuration or instruction based on the accurate remaining computing power of the terminal, which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • FIG1 is a schematic diagram of a wireless communication system applicable to an embodiment of the present application.
  • FIG2 is a schematic diagram of the structure of a neural network provided in an embodiment of the present application.
  • FIG3 is a schematic diagram of the computational logic of a neuron provided in an embodiment of the present application.
  • FIG4 is a flowchart of a method for reporting AI computing power provided in an embodiment of the present application.
  • FIG5 is a second flow chart of the AI computing power reporting method provided in an embodiment of the present application.
  • FIG6 is a schematic diagram of signaling interaction of the AI computing power reporting method provided in an embodiment of the present application.
  • FIG7 is a schematic diagram of a structure of an AI computing power reporting device provided in an embodiment of the present application.
  • FIG8 is a second structural diagram of the AI computing power reporting device provided in an embodiment of the present application.
  • FIG9 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application.
  • FIG10 is a schematic diagram of the structure of a terminal provided in an embodiment of the present application.
  • FIG. 11 is a schematic diagram of the structure of a network side device provided in an embodiment of the present application.
  • first, second, etc. in the specification and claims of the present application are used to distinguish similar objects, but not to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by “first” and “second” are generally of the same type, and the number of objects is not limited.
  • the first object can be one or more.
  • “and/or” in the specification and claims represents at least one of the connected objects, and the character “/” generally indicates that the objects associated with each other are in an "or” relationship.
  • the term “indication” in the specification and claims of the present application can be either an explicit indication or an implicit indication.
  • an explicit indication can be understood as the sender explicitly notifying the receiver of the operation or request result to be performed in the indication sent;
  • an implicit indication can be understood as the receiver making a judgment based on the indication sent by the sender and determining the operation or request result to be performed based on the judgment result.
  • LTE Long Term Evolution
  • LTE-A Long Term Evolution
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency Division Multiple Access
  • NR new radio
  • FIG1 is a schematic diagram of a wireless communication system applicable to an embodiment of the present application.
  • the wireless communication system shown in FIG1 includes a terminal 11 and a network side device 12.
  • the terminal 11 may be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer) or a personal digital assistant, or a personal digital assistant.
  • PDA Personal Digital Assistant
  • PDA Personal Digital Assistant
  • PDA netbook
  • ultra-mobile personal computer ultra-mobile personal computer
  • UMPC mobile Internet device
  • Mobile Internet Device MID
  • augmented reality augmented reality, AR
  • virtual reality virtual reality
  • VR virtual reality
  • robot wearable device
  • VUE vehicle-mounted equipment
  • PUE pedestrian terminal
  • smart home home equipment with wireless communication function, such as refrigerator, TV, washing machine or furniture, etc.
  • game console personal computer
  • personal computer personal computer, PC
  • wearable devices include: smart watch, smart bracelet, smart headset, smart glasses, smart jewelry (smart bracelet, smart bracelet, smart ring, smart necklace, smart anklet, smart anklet, etc.), smart wristband, smart clothing, etc.
  • terminal 11 can also be a chip in the terminal, such as a modem chip, a system-on-chip (System on Chip, SoC). It should be noted that the specific type of terminal 11 is not limited in the embodiment of the present application.
  • the network side device 12 may include an access network device or a core network device, wherein the access network device may also be referred to as a radio access network device, a radio access network (RAN), a radio access network function or a radio access network unit.
  • the access network device may include a base station, a WLAN access point or a WiFi node, etc.
  • the base station may be referred to as a node B, an evolved node B (eNB), an access point, a base transceiver station (BTS), a radio base station, a radio transceiver, a basic service set (BSS), an extended service set (ESS), a home B node, a home evolved B node, a transmitting and receiving point (TRP) or some other appropriate term in the field, as long as the same technical effect is achieved, the base station is not limited to a specific technical vocabulary, it should be noted that in the embodiment of the present application, only the base station in the NR system is used as an example for introduction, and the specific type of the base station is not limited.
  • the core network equipment may include but is not limited to at least one of the following: core network nodes, core network functions, mobility management entity (Mobility Management Entity, MME), access mobility management function (Access and Mobility Management Function, AMF), session management function (Session Management Function, SMF), user plane function (User Plane Function, UPF), policy control function (Policy Control Function, PCF), policy and charging rules function unit (Policy and Charging Rules Function, PCRF), edge application service discovery function (Edge Application Server Discovery Function, EASDF), unified data management (Unified Data Management, UDM), unified data repository (Unified Data Repos itory, UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (L-NEF), Binding Support Function (BSF), Application Function (AF), Location Management Function (LMF), Enhanced Serving Mobile Location Centre (E-SMLC), Network data analytics function (NWDAF), etc.
  • MME mobility management entity
  • AI Artificial intelligence
  • neural networks decision trees, support vector machines, Bayesian classifiers, etc. This application takes neural networks as an example for illustration, but does not limit the specific type of AI modules.
  • FIG2 is a schematic diagram of the structure of a neural network provided in an embodiment of the present application.
  • a neural network includes an input layer, a hidden layer and an output layer; wherein X 1 , X 2 , and X n are inputs of the neural network, and Y is the output layer of the neural network. The output of the network.
  • FIG3 is a schematic diagram of the calculation logic of neurons provided in an embodiment of the present application.
  • a 1 , a k , a K are inputs
  • w 1 , w k , w K are weights (multiplicative coefficients)
  • b is a bias (additive coefficient)
  • ⁇ (z) is an activation function.
  • Common activation functions include Sigmoid, tanh, linear rectification function (also known as Rectified Linear Unit (ReLU)), etc.
  • the parameters of the neural network are optimized using a gradient optimization algorithm, which is a class of algorithms that minimize or maximize an objective function (sometimes called a loss function), which is often a mathematical combination of the model parameters and the data.
  • a gradient optimization algorithm which is a class of algorithms that minimize or maximize an objective function (sometimes called a loss function), which is often a mathematical combination of the model parameters and the data.
  • f(.) For example, given data X and its corresponding label Y, we build a neural network model f(.). With the model, we can get the predicted output f(x) based on the input X, and calculate the difference between the predicted value and the true value (f(x)-Y), which is the loss function. Our goal is to find the appropriate w and b to minimize the value of the above loss function. The smaller the loss value, the closer our model is to the actual situation.
  • the common optimization algorithms are basically based on the error back propagation (BP) algorithm.
  • BP error back propagation
  • the basic idea of the BP algorithm is that the learning process consists of two processes: the forward propagation of the signal and the back propagation of the error.
  • the input sample is transmitted from the input layer, processed by each hidden layer layer by layer, and then transmitted to the output layer. If the actual output of the output layer does not match the expected output, it will enter the error back propagation stage.
  • Error back propagation is to propagate the output error layer by layer through the hidden layer to the input layer in some form, and distribute the error to all units in each layer, so as to obtain the error signal of each layer unit, and this error signal is used as the basis for correcting the weights of each unit.
  • This process of adjusting the weights of each layer of the signal forward propagation and error back propagation is repeated.
  • the process of continuous adjustment of weights is the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level, or until the pre-set number of learning times is reached.
  • the network side can instruct the UE to use a specific AI model.
  • the network side estimates that the remaining UE computing power is greater than the actual one, the network side instructs the UE to use an overly complex AI model, causing the UE to be unable to operate the AI model normally; if the network side estimates that the remaining UE computing power is lower than the actual one, the network side instructs the UE to use an overly simple AI model, resulting in a waste of UE AI computing power.
  • the network side cannot accurately estimate the remaining AI computing power of the UE, resulting in low utilization of the UE AI computing power, affecting the performance of the communication system.
  • the embodiments of the present application provide an AI computing power reporting method, terminal and network side equipment, which can improve the performance of the communication system.
  • FIG. 4 is a flowchart of a method for reporting AI computing power provided in an embodiment of the present application. As shown in FIG. 4 , the method includes steps 401-402; wherein:
  • Step 401 The terminal obtains first AI computing power information; wherein the first AI computing power information is used to indicate at least one of the following: the AI model computing resources currently remaining in the terminal; the AI model computing resources currently available to the terminal; all the AI model computing resources of the terminal; all the AI model computing resources of the terminal that can be used for wireless communication.
  • the terminals include but are not limited to the types of terminals 11 listed above; the network side devices include but are not limited to the types of network side devices 12 listed above; this application does not limit this.
  • the terminal Since the network side equipment cannot accurately estimate the remaining AI computing power of the terminal, the utilization rate of the terminal AI computing power is low, which affects the performance of the communication system; therefore, in order to improve the utilization rate of the terminal AI computing power and improve the performance of the communication system, in this embodiment, the terminal first needs to obtain the first AI computing power information.
  • the first AI computing power information includes M AI unit computing power units, where M is an integer or a decimal; each AI unit is used to indicate N1 computing resource units, where N1 is a positive integer or a decimal.
  • AI unit computing power unit refers to the unit for measuring AI model computing resources, where AI model computing resources include, for example, the number of AI model operations.
  • the computing resource unit includes at least one of the following:
  • FLOPs floating point operations
  • FLOPS floating point operations per second
  • the definition of the AI unit satisfies at least one of the following:
  • the definition of the AI unit includes: the value of N1, and/or the type of computing resource unit.
  • Step 402 The terminal sends the first AI computing power information to a network side device.
  • the terminal needs to send the acquired first AI computing power information to the network side device. Accordingly, after receiving the first AI computing power information, the network side device needs to obtain the second AI computing power information corresponding to the terminal based on the first AI computing power information; wherein the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device.
  • the network side device can estimate the remaining AI model computing resources of the terminal in real time, so that it can issue a suitable first AI model to the terminal based on the remaining AI model computing resources of the terminal, or indicate the appropriate first AI model to the terminal.
  • the terminal obtains the currently remaining AI model computing resources that can be used for AI model-related operations, that is, the first AI computing power information, and then reports the first AI computing power information of the terminal to the network side device, so that the network side device obtains accurate terminal remaining computing power, so that the network side device can perform AI configuration or indication based on the accurate terminal remaining computing power, which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • the terminal obtains the first AI computing power information by any of the following methods:
  • Method 1 The terminal determines the first AI computing power information based on terminal configuration information.
  • the first AI computing power information of the terminal is pre-configured in the terminal configuration information, and the terminal can directly obtain the first AI computing power information from the terminal configuration information.
  • Method 2 The terminal determines the first AI computing power information based on terminal configuration information and occupied AI computing power information.
  • the total AI computing power information pre-configured in the terminal configuration information can be subtracted from the occupied AI computing power information to obtain the first AI computing power information.
  • the model configuration information or associated information of the AI model includes the number of AI units occupied by the AI model; the number of AI units occupied by the AI model is converted into the computational complexity of the AI model. It can be calculated.
  • the model configuration information or related information of the AI model includes the number of AI units occupied by the AI model; wherein the number of AI units occupied by the AI model is calculated according to the computational complexity of the AI model.
  • one AI unit is used to indicate the number of operations of 5 TOP.
  • the computational complexity of an AI model is the number of operations of 15 TOP.
  • the AI unit occupied by the AI model is the number of operations of 15 TOP divided by the number of operations of 5 TOP, that is, the AI model occupies 3 AI units.
  • the computational complexity of the AI model is N2 computing resource units, where N2 is a positive integer or a decimal;
  • the computational complexity of an AI model is measured by the number of AI units occupied by the AI model, and each AI unit is used to indicate N1 computing resource units; therefore, the computational complexity of an AI model can be expressed in N2 computing resource units.
  • the number of AI units occupied by the AI model is obtained by any of the following methods:
  • Method 1 When M is a decimal, calculate N2 divided by N1 to obtain the number of AI units occupied by the AI model.
  • one AI unit is used to indicate the number of operations of 4 TOP.
  • the computational complexity of an AI model is the number of operations of 15 TOP.
  • the AI unit occupied by the AI model is the number of operations of 15 TOP divided by the number of operations of 4 TOP, that is, the AI model occupies 3.75 AI units.
  • Method 2 When M is an integer, calculate N2 divided by N1, round up or approximately round the quotient to obtain the number of AI units occupied by the AI model.
  • one AI unit is used to indicate 10 TOP operations.
  • the computational complexity of an AI model is 23 TOP operations.
  • the AI unit occupied by the AI model is the number of operations of 23 TOP divided by the number of operations of 10 TOP.
  • the quotient is then rounded up, which means the AI model occupies 3 AI units.
  • the quotient is approximately rounded up, which means the AI model occupies 2 AI units.
  • the computational complexity of the AI model is represented by the number of AI units occupied by the AI model, so that the terminal can accurately send the currently remaining AI model computing resources that can be used for AI model-related operations to the network side device based on the number of AI units, so that the network side device can obtain accurate remaining computing power of the terminal.
  • the terminal sends the first AI computing power information to the network side device, which can be specifically implemented by the following steps:
  • the terminal sends the first AI computing power information to the network side device during the process of reporting the AI capability information of the terminal to the network side device.
  • the terminal when the terminal reports its AI capabilities to the network-side device, it also reports/carries its total AI unit computing power unit.
  • the AI model computing resource is used for at least one of the following AI model related operations:
  • signal processing based on AI models includes signal detection, filtering, equalization, etc.; among them, signals include demodulation reference signal (Demodulation Reference Signal, DMRS), sounding reference signal (Sounding Reference Signal, SRS), synchronization signal block (Synchronization Signal Block, SSB), tracking reference signal (Tracking Reference Signal, TRS), phase tracking reference signal (Phase-Tracking Reference Signals, PTRS), channel state information reference signal (Channel State Information-Reference Signal, CSI-RS), etc.
  • signals include demodulation reference signal (Demodulation Reference Signal, DMRS), sounding reference signal (Sounding Reference Signal, SRS), synchronization signal block (Synchronization Signal Block, SSB), tracking reference signal (Tracking Reference Signal, TRS), phase tracking reference signal (Phase-Tracking Reference Signals, PTRS), channel state information reference signal (Channel State Information-Reference Signal, CSI-RS), etc.
  • signals include demodulation reference signal (Demodulation Reference Signal, DMRS), sound
  • the channels included in the signal transmission/reception/demodulation/transmission may be, for example, a physical downlink control channel (PDCCH), a physical downlink shared channel (PDCCH), downlink shared channel (PDSCH), physical uplink control channel (Physical Uplink Control Channel, PUCCH), physical uplink shared channel (Physical Uplink Shared Channel, PUSCH), physical random access channel (Physical Random Access Channel, PRACH), physical broadcast channel (Physical broadcast channel, PBCH), etc.
  • a physical downlink control channel (PDCCH), a physical downlink shared channel (PDCCH), downlink shared channel (PDSCH), physical uplink control channel (Physical Uplink Control Channel, PUCCH), physical uplink shared channel (Physical Uplink Shared Channel, PUSCH), physical random access channel (Physical Random Access Channel, PRACH), physical broadcast channel (Physical broadcast channel, PBCH), etc.
  • PUCCH Physical Uplink Control Channel
  • PUSCH Physical Uplink shared channel
  • PRACH Physical Random Access Channel
  • Channel state information feedback including channel-related information, channel matrix-related information, channel characteristic information, channel matrix characteristic information, precoding matrix indicator (PMI), rank indicator (RI), CSI-RS resource indicator (CSI-RS Resource Indicator, CRI), channel quality indicator (CQI), layer indicator (LI), etc.
  • PMI precoding matrix indicator
  • RI rank indicator
  • RI CSI-RS resource indicator
  • CQI channel quality indicator
  • LI layer indicator
  • the base station obtains the angle and delay information according to the uplink channel, and can notify the UE of the angle and delay information through CSI-RS precoding or direct indication.
  • the UE reports according to the indication of the base station or selects and reports within the indication range of the base station, thereby reducing the UE's calculation amount and the CSI reporting overhead.
  • Beam management based on AI model including: beam measurement, beam reporting, beam prediction, beam failure detection, beam failure recovery, and new beam indication in beam failure recovery.
  • Interference suppression based on AI model including intra-cell interference, inter-cell interference, out-of-band interference, intermodulation interference, etc.
  • the specific position (including horizontal position and/or vertical position) or possible future trajectory of the UE is estimated through a reference signal (such as SRS), or information to assist position estimation or trajectory estimation.
  • a reference signal such as SRS
  • Control signaling analysis based on AI models; for example, related signaling for power control and related signaling for beam management.
  • FIG. 5 is a second flow chart of the AI computing power reporting method provided in an embodiment of the present application. As shown in FIG. 5 , the method includes steps 501-502; wherein:
  • Step 501 The network side device receives first AI computing power information sent by the terminal; wherein the first AI computing power information is used to indicate at least one of the following: the AI model computing resources currently remaining in the terminal; the AI model computing resources currently available to the terminal; all the AI model computing resources of the terminal; all the AI model computing resources of the terminal that can be used for wireless communication.
  • the terminals include but are not limited to the types of terminals 11 listed above; the network side devices include but are not limited to the types of network side devices 12 listed above; this application does not limit this.
  • the network side device cannot accurately estimate the remaining AI computing power of the terminal, the utilization rate of the terminal AI computing power is low, which affects the performance of the communication system; therefore, in order to improve the utilization rate of the terminal AI computing power and improve the performance of the communication system, in this embodiment, the network side device needs to receive the first AI computing power information sent by the terminal.
  • the first AI computing power information includes M AI unit computing power units, where M is an integer or a decimal; each AI unit is used to indicate N1 computing resource units, where N1 is a positive integer or a decimal.
  • AI unit computing power unit refers to the unit for measuring AI model computing resources, where AI model computing resources include, for example, the number of AI model operations.
  • the computing resource unit includes at least one of the following:
  • the definition of the AI unit satisfies at least one of the following:
  • the definition of the AI unit includes: the value of N1, and/or the type of computing resource unit.
  • Step 502 The network side device obtains second AI computing power information corresponding to the terminal based on the first AI computing power information; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device.
  • the network side device can estimate the remaining AI model computing resources of the terminal in real time based on the first AI computing power information sent by the terminal, so that it can issue a suitable first AI model to the terminal based on the remaining AI model computing resources of the terminal, or indicate the appropriate first AI model to the terminal.
  • the network side device obtains the accurate remaining computing power of the terminal by receiving the first AI computing power information sent by the terminal, so that the network side device can perform AI configuration or indication based on the accurate remaining computing power of the terminal, thereby improving the utilization rate of the terminal AI computing power and improving the performance of the communication system.
  • the model configuration information or associated information of the AI model includes the number of AI units occupied by the AI model; the number of AI units occupied by the AI model is calculated by the computational complexity of the AI model.
  • the model configuration information or related information of the AI model includes the number of AI units occupied by the AI model; wherein the number of AI units occupied by the AI model is calculated according to the computational complexity of the AI model.
  • the computational complexity of the AI model is N2 computing resource units, where N2 is a positive integer or a decimal;
  • the number of AI units occupied by the AI model is obtained by any of the following methods:
  • Method 1 When M is a decimal, calculate N2 divided by N1 to obtain the number of AI units occupied by the AI model.
  • Method 2 When M is an integer, calculate N2 divided by N1, round up or approximately round the quotient to obtain the number of AI units occupied by the AI model.
  • the network side device obtains the second AI computing power information corresponding to the terminal, it is also necessary to configure or indicate the first AI model to the terminal based on the remaining AI model computing resources of the terminal to improve the utilization rate of the terminal's AI computing power. This can be achieved in any of the following ways:
  • Method 1 When the number of AI units occupied by the first AI model is less than or not greater than the second AI computing power information, the network side device sends the first AI model to the terminal.
  • the complexity of the first AI model sent by the network-side device to the terminal cannot be greater than, or cannot be greater than or equal to, the terminal's current idle AI unit (i.e., the second AI computing power information).
  • the network-side device estimates that the remaining AI model computing resources of the terminal are 5 AI units; then the number of AI units occupied by the first AI model sent by the network-side device to the terminal should be less than 5.
  • Method 2 When the number of AI units occupied by the first AI model is less than or not greater than the second AI computing power information, the network side device instructs the terminal to activate the first AI model.
  • the network side device can send an indication message to the terminal so that the terminal activates the first AI model; it can be understood that the complexity of the first AI model indicated by the network side device to the terminal cannot be greater than, or cannot be greater than or equal to, the current idle AI unit of the terminal (i.e., the second AI computing power information).
  • the network-side device estimates that the remaining AI model computing resources of the terminal are 5 AI units; then the number of AI units occupied by the first AI model indicated by the network-side device to the terminal should be less than 5.
  • Method 3 When a first difference between the number of AI units occupied by the first AI model and the number of AI units occupied by the second AI model is less than or not greater than the second AI computing power information, the network side device instructs the terminal to deactivate the second AI model and activate the first AI model.
  • the network side device instructs the terminal to switch from the second AI model to the first AI model (i.e., the network side device instructs the terminal to deactivate the currently used first AI model and activate the second AI model)
  • the complexity of the first AI model exceeding that of the second AI model cannot be greater than, or cannot be greater than or equal to, the terminal's current idle AI unit (i.e., the second AI computing power information).
  • the network side device can directly instruct the terminal to deactivate the second AI model and activate the first AI model.
  • the network side device can accurately configure or instruct the AI model based on the remaining AI model computing resources of the terminal (i.e., the second AI computing power information), which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • the second AI computing power information needs to be updated, which can be specifically implemented by the following steps:
  • the network side device subtracts the number of AI units occupied by the first AI model from the second AI computing power information to obtain updated second AI computing power information.
  • the network-side device estimates that the remaining AI model computing resources of the terminal are 5 AI units, and the first AI model occupies 2 AI units; then the updated second AI computing power information is 3 AI units.
  • the second AI computing power information needs to be updated, which can be specifically implemented by the following steps:
  • the network side device subtracts the number of AI units occupied by the first AI model from the second AI computing power information to obtain updated second AI computing power information.
  • the second AI computing power information needs to be updated, which can be specifically implemented by the following steps:
  • the network-side device calculates a first difference between the number of AI units occupied by the first AI model and the number of AI units occupied by the second AI model; subtracts the first difference from the second AI computing power information to obtain updated second AI computing power information.
  • the first difference can be a negative number.
  • the network-side device estimates that the remaining AI model computing resources of the terminal are 5 AI units, the first AI model occupies 2 AI units, and the second AI model occupies 3 AI units.
  • the first difference is -1 AI unit, and the updated second AI computing power information is 6 AI units.
  • the network side device when the network side device instructs the terminal to deactivate the third AI model, the network side device adds the number of AI units occupied by the third AI model to the second AI computing power information to obtain updated second AI computing power information.
  • the network-side device estimates that the remaining AI model computing resources of the terminal are 5 AI units, and the third AI model occupies 2 AI units; when the network-side device instructs the terminal to activate the third AI model, the updated second AI computing power information is 7 AI units.
  • the network side device can realize real-time update of the second AI computing power information, so that the network side device can further perform AI configuration or instruction based on the accurate terminal remaining computing power, which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • the AI model computing resource is used for at least one of the following AI model related operations:
  • FIG6 is a schematic diagram of signaling interaction of the AI computing power reporting method provided in an embodiment of the present application. As shown in FIG6 , it specifically includes steps 1 to 7:
  • Step 1 The terminal obtains first AI computing power information.
  • the first AI computing power information includes M AI unit computing power units, where M is an integer or a decimal; each AI unit is used to indicate N1 computing resource units, where N1 is a positive integer or a decimal.
  • the computing resource unit includes at least one of the following: a) the number of operations of a single operation; b) the number of operations of a trillion operations; c) the number of operations of floating-point operations; d) the memory access cost; e) the number of operations of multiplication and addition operations.
  • AI unit satisfies at least one of the following: a) agreed upon by the protocol; b) defined by the terminal; c) configured by the network side device.
  • Step 2 The terminal sends the AI capability information of the terminal to the network side device, wherein the AI capability information of the terminal includes the first AI computing power information.
  • Step 3 The network side device obtains the second AI computing power information corresponding to the terminal based on the first AI computing power information.
  • the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device.
  • step 3 is completed, at least one of steps 4 to 6 is started.
  • Step 4 The network-side device sends the first AI model to the terminal.
  • the network side device sends the first AI model to the terminal.
  • Step 5 The network side device sends first indication information to the terminal; the first indication information is used to indicate the first AI model.
  • the network side device instructs the terminal to activate the first AI model.
  • Step 6 The network side device sends second indication information to the terminal; the second indication information is used to instruct the terminal to deactivate the second AI model and activate the first AI model.
  • the network side device instructs the terminal to deactivate the second AI model and activate the first AI model.
  • Step 7 The network side device updates the second AI computing power information to obtain the updated second AI computing power information.
  • the network side device needs to subtract the number of AI units occupied by the first AI model from the second AI computing power information to obtain the updated second AI computing power information.
  • the network-side device After executing step 5, the network-side device needs to subtract the number of AI units occupied by the first AI model from the second AI computing power information to obtain the updated second AI computing power information.
  • the network-side device After executing step 6, the network-side device needs to calculate the first difference between the number of AI units occupied by the first AI model and the number of AI units occupied by the second AI model; then subtract the first difference from the second AI computing power information to obtain the updated second AI computing power information.
  • the network-side device When the network-side device instructs the terminal to deactivate the third AI model, the network-side device needs to add the number of AI units occupied by the third AI model to the second AI computing power information to obtain the updated second AI computing power information.
  • the AI computing power reporting method provided in the embodiment of the present application can be executed by an AI computing power reporting device.
  • the AI computing power reporting device executing the AI computing power reporting method as an example to illustrate the AI computing power reporting device provided in the embodiment of the present application.
  • FIG. 7 is a schematic diagram of a structure of an AI computing power reporting device provided in an embodiment of the present application. As shown in FIG. 7 , the AI computing power reporting device 700 is applied to a terminal, and includes:
  • a first acquisition module 701 is used to acquire first AI computing power information
  • a sending module 702 is used to send the first AI computing power information to a network side device
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • the network side device by obtaining the currently remaining AI model computing resources that can be used for AI model-related operations, that is, the first AI computing power information, and then reporting the first AI computing power information of the terminal to the network side device, the network side device obtains accurate terminal remaining computing power, so that the network side device can perform AI configuration or indication based on the accurate terminal remaining computing power, which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • the first AI computing power information includes M AI unit computing power units, where M is an integer or a decimal; each AI unit is used to indicate N1 computing resource units, where N1 is a positive integer or a decimal.
  • the computing resource unit includes at least one of the following:
  • the definition of the AI unit satisfies at least one of the following: agreed upon by the protocol; defined by the terminal; configured by the network side device.
  • the model configuration information or associated information of the AI model includes the number of AI units occupied by the AI model; the number of AI units occupied by the AI model is calculated by the computational complexity of the AI model.
  • the computational complexity of the AI model is N2 computing resource units, where N2 is a positive integer or a decimal;
  • the number of AI units occupied by the AI model is obtained by any of the following methods:
  • N2 is divided by N1 to obtain the number of AI units occupied by the AI model
  • N2 When M is an integer, calculate N2 divided by N1, round up or approximately round the quotient to obtain the number of AI units occupied by the AI model.
  • the first acquisition module 701 is further used for any of the following:
  • the first AI computing power information is determined.
  • the sending module 702 is further configured to:
  • the first AI computing power information is sent to the network side device.
  • the AI model computing resource is used for at least one of the following AI model related operations:
  • FIG8 is a second structural diagram of an AI computing power reporting device provided in an embodiment of the present application. As shown in FIG8 , the AI computing power reporting device 800 is applied to a network side device, including:
  • a receiving module 801 is configured to receive first AI computing power information sent by a terminal
  • a second acquisition module 802 is used to acquire second AI computing power information corresponding to the terminal based on the first AI computing power information; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device;
  • the first AI computing power information is used to indicate at least one of the following:
  • the AI model computing resources currently remaining on the terminal are the AI model computing resources currently remaining on the terminal.
  • the AI model computing resources currently available to the terminal are the AI model computing resources currently available to the terminal.
  • the network side device by receiving the first AI computing power information sent by the terminal, the network side device obtains the accurate terminal remaining computing power, so that the network side device can perform AI configuration or indication based on the accurate terminal remaining computing power, which can improve the utilization rate of the terminal AI computing power and improve the performance of the communication system.
  • the first AI computing power information includes M AI unit computing power units, where M is an integer or a decimal; each AI unit is used to indicate N1 computing resource units, where N1 is a positive integer or a decimal.
  • the computing resource unit includes at least one of the following:
  • the definition of the AI unit satisfies at least one of the following: agreed upon by the protocol; defined by the terminal; configured by the network side device.
  • the model configuration information or associated information of the AI model includes the number of AI units occupied by the AI model; the number of AI units occupied by the AI model is calculated by the computational complexity of the AI model.
  • the computational complexity of the AI model is N2 computing resource units, where N2 is a positive integer or a decimal;
  • the number of AI units occupied by the AI model is obtained by any of the following methods:
  • N2 is divided by N1 to obtain the number of AI units occupied by the AI model
  • N2 When M is an integer, calculate N2 divided by N1, round up or approximately round the quotient to obtain the number of AI units occupied by the AI model.
  • the device further comprises at least one of the following:
  • a sending module configured to send the first AI model to the terminal when the number of AI units occupied by the first AI model is less than or not greater than the second AI computing power information
  • a first indication module configured to instruct the terminal to activate the first AI model when the number of AI units occupied by the first AI model is less than or not greater than the second AI computing power information
  • the second indication module is used to instruct the terminal to deactivate the second AI model and activate the first AI model when a first difference between the number of AI units occupied by the first AI model and the number of AI units occupied by the second AI model is less than or not greater than the second AI computing power information.
  • the device after sending the first AI model to the terminal, the device further includes:
  • the first update module is used to subtract the number of AI units occupied by the first AI model from the second AI computing power information to obtain updated second AI computing power information.
  • the device after instructing the terminal to activate the first AI model, the device further includes:
  • the second update module is used to subtract the number of AI units occupied by the first AI model from the second AI computing power information to obtain updated second AI computing power information.
  • the device after instructing the terminal to deactivate the second AI model and activating the first AI model, the device further includes:
  • the third update module is used to calculate a first difference between the number of AI units occupied by the first AI model and the number of AI units occupied by the second AI model; and subtract the first difference from the second AI computing power information to obtain updated second AI computing power information.
  • the device further comprises:
  • a third instruction module used to instruct the terminal to deactivate a third AI model
  • the fourth updating module is used to add the number of AI units occupied by the third AI model to the second AI computing power information to obtain updated second AI computing power information.
  • the AI model computing resource is used for at least one of the following AI model related operations:
  • the AI computing power reporting device in the embodiment of the present application can be an electronic device, such as an electronic device with an operating system, or a component in an electronic device, such as an integrated circuit or a chip.
  • the electronic device can be a terminal, or it can be other devices other than a terminal.
  • the terminal can include but is not limited to the types of terminal 11 listed above, and other devices can be servers, network attached storage (NAS), etc., which are not specifically limited in the embodiment of the present application.
  • the AI computing power reporting device provided in the embodiment of the present application can implement the various processes implemented by the method embodiments of Figures 4 to 5 and achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • FIG9 is a schematic diagram of the structure of a communication device provided in an embodiment of the present application.
  • the communication device 900 includes a processor 901 and a memory 902.
  • the memory 902 stores programs or instructions that can be run on the processor 901.
  • the program or instruction is executed by the processor 901 to implement the various steps of the above-mentioned AI computing power reporting method embodiment, and can achieve the same technical effect.
  • the communication device 900 is a network-side device, the program or instruction is executed by the processor 901 to implement the various steps of the above-mentioned AI computing power reporting method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the embodiment of the present application also provides a terminal, including a processor and a communication interface, wherein the processor is used to: obtain first AI computing power information, and the communication interface is used to: send the first AI computing power information to a network side device;
  • the first AI computing power information is used to indicate at least one of the following: the remaining AI model computing resources of the terminal; the AI model computing resources currently available to the terminal; the AI model computing resources owned by the terminal; the AI model computing resources owned by the terminal that can be used for wireless communication.
  • This terminal embodiment corresponds to the above-mentioned terminal side method embodiment, and each implementation process and implementation method of the above-mentioned method embodiment can be applied to this terminal embodiment and can achieve the same technical effect.
  • Figure 10 is a structural diagram of the terminal provided in an embodiment of the present application.
  • the terminal 1000 includes but is not limited to: a radio frequency unit 1001, a network module 1002, an audio output unit 1003, an input unit 1004, a sensor 1005, a display unit 1006, a user input unit 1007, an interface unit 1008, a memory 1009 and at least some of the components in the processor 1010.
  • the terminal 1000 can also include a power supply (such as a battery) for supplying power to each component, and the power supply can be logically connected to the processor 1010 through a power management system, so as to implement functions such as charging, discharging, and power consumption management through the power management system.
  • a power supply such as a battery
  • the terminal structure shown in FIG10 does not constitute a limitation on the terminal, and the terminal can include more or fewer components than shown in the figure, or combine certain components, or arrange components differently, which will not be described in detail here.
  • the input unit 1004 may include a graphics processing unit (GPU) 10041 and a microphone 10042, and the graphics processor 10041 processes the image data of the static picture or video obtained by the image capture device (such as a camera) in the video capture mode or the image capture mode.
  • the display unit 1006 may include a display panel 10061, and the display panel 10061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc.
  • the user input unit 1007 includes a touch panel 10071 and at least one of other input devices 10072.
  • the touch panel 10071 is also called a touch screen.
  • the touch panel 10071 may include two parts: a touch detection device and a touch controller.
  • Other input devices 10072 may include, but are not limited to, a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.
  • the RF unit 1001 can transmit the data to the processor 1010 for processing; in addition, the RF unit 1001 can send uplink data to the network side device.
  • the RF unit 1001 includes but is not limited to an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc.
  • the memory 1009 can be used to store software programs or instructions and various data.
  • the memory 1009 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instruction required for at least one function (such as a sound playback function, an image playback function, etc.), etc.
  • the memory 1009 may include a volatile memory or a non-volatile memory, or the memory 1009 may include a transient and non-transient memory.
  • the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.
  • the volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM).
  • the memory x09 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.
  • Processor 1010 may include one or more processing units; optionally, processor x10 integrates an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application program.
  • the modem processor mainly processes wireless communication signals, such as a baseband processor. It can be understood that the modem processor may not be integrated into the processor 1010.
  • the embodiment of the present application also provides a network side device, including a processor and a communication interface, the communication interface is used to: receive first AI computing power information sent by a terminal, the processor is used to: based on the first AI computing power information, obtain second AI computing power information corresponding to the terminal; the second AI computing power information is used to indicate the remaining AI model computing resources of the terminal estimated by the network side device; wherein the first AI computing power information is used to indicate at least one of the following: the AI model computing resources currently remaining in the terminal; the AI model computing resources currently available to the terminal; all the AI model computing resources of the terminal; all the AI model computing resources of the terminal that can be used for wireless communication.
  • This network side device embodiment corresponds to the above-mentioned network side device method embodiment, and each implementation process and implementation method of the above-mentioned method embodiment can be applied to this network side device embodiment, and can achieve the same technical effect.
  • FIG11 is a schematic diagram of the structure of a network side device provided in an embodiment of the present application.
  • the network side device 1100 includes: an antenna 1101, a radio frequency device 1102, a baseband device 1103, a processor 1104, and a memory 1105.
  • the antenna 1101 is connected to the radio frequency device 1102.
  • the radio frequency device 1102 receives information through the antenna 1101 and sends the received information to the baseband device 1103 for processing.
  • the baseband device 1103 processes the information to be sent and sends it to the radio frequency device 1102.
  • the radio frequency device 1102 processes the received information and sends it out through the antenna 1101.
  • the method executed by the network-side device in the above embodiment may be implemented in the baseband device 1103, which includes a baseband processor.
  • the baseband device 1103 may include, for example, at least one baseband board, on which multiple chips are arranged, as shown in Figure 11, one of which is, for example, a baseband processor, which is connected to the memory 1105 through a bus interface to call the program in the memory 1105 and execute the network device operations shown in the above method embodiment.
  • the network side device may also include a network interface 1106, which is, for example, a common public radio interface (CPRI).
  • a network interface 1106, which is, for example, a common public radio interface (CPRI).
  • CPRI common public radio interface
  • the network side device 1100 of the embodiment of the present invention also includes: instructions or programs stored in the memory 1105 and executable on the processor 1104.
  • the processor 1104 calls the instructions or programs in the memory 1105 to execute the AI computing power reporting method as described above and achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • An embodiment of the present application also provides an AI computing power reporting system, including: a terminal and a network side device, wherein the terminal can be used to execute the steps of the AI computing power reporting method shown in FIG. 4 above, and the network side device can be used to execute the steps of the AI computing power reporting method shown in FIG. 5 above.
  • the embodiment of the present application also provides a readable storage medium, which can be volatile or non-volatile.
  • the readable storage medium stores a program or instruction.
  • the program or instruction is executed by the processor, the various processes of the above-mentioned AI computing power reporting method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.
  • the processor is the processor in the terminal described in the above embodiment.
  • the readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.
  • An embodiment of the present application further provides a chip, which includes 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 various processes of the above-mentioned AI computing power reporting method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.
  • the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.
  • the present application embodiment further provides a computer program/program product, wherein the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement the above-mentioned AI
  • the various processes of the computing power reporting method embodiment can achieve the same technical effect, and to avoid repetition, they will not be repeated here.
  • the technical solution of the present application can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM/RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.
  • a storage medium such as ROM/RAM, a magnetic disk, or an optical disk
  • a terminal which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)

Abstract

本申请公开了一种AI算力上报方法、终端及网络侧设备，属于通信技术领域，本申请实施例的AI算力上报方法包括：终端获取第一AI算力信息；所述终端向所述网络侧设备发送所述第一AI算力信息；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。

Description

AI算力上报方法、终端及网络侧设备
相关申请的交叉引用
本申请主张在2022年12月15日在中国提交的申请号为202211616288.6的中国专利的优先权，其全部内容通过引用包含于此。
技术领域
本申请属于通信技术领域，具体涉及一种AI算力上报方法、终端及网络侧设备。
背景技术
目前，人工智能(Artificial Intelligence，AI)技术在多个领域获得了广泛的应用，将人工智能融入无线通信网络，显著提升吞吐量、时延以及用户容量等技术指标是未来的无线通信网络的重要任务。
相关技术中，网络侧可以指示用户设备(User Equipment，UE)使用特定的AI模型。
但是，网络侧无法准确估计UE剩余AI算力，导致UE AI算力的利用率较低，影响通信***的性能。
发明内容
本申请实施例提供一种AI算力上报方法、终端及网络侧设备，能够提高通信***的性能。
第一方面，提供了一种AI算力上报方法，该方法包括：
终端获取第一AI算力信息；
所述终端向网络侧设备发送所述第一AI算力信息；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
第二方面，提供了一种AI算力上报方法，该方法包括：
网络侧设备接收终端发送的第一AI算力信息；
所述网络侧设备基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示所述网络侧设备估计的所述终端剩余的AI模型计算资源；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
第三方面，提供了一种AI算力上报装置，该装置包括：
第一获取模块，用于获取第一AI算力信息；
发送模块，用于向网络侧设备发送所述第一AI算力信息；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
第四方面，提供了一种AI算力上报装置，该装置包括：
接收模块，用于接收终端发送的第一AI算力信息；
第二获取模块，用于基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示网络侧设备估计的所述终端剩余的AI模型计算资源；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
第五方面，提供了一种终端，该终端包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤。
第六方面，提供了一种终端，包括处理器及通信接口；其中，所述处理器用于：获取第一AI算力信息，所述通信接口用于：向网络侧设备发送所述第一AI算力信息；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。
第七方面，提供了一种网络侧设备，该网络侧设备包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第二方面所述的方法的步骤。
第八方面，提供了一种网络侧设备，包括处理器及通信接口；其中，所述通信接口用于：接收终端发送的第一AI算力信息，所述处理器用于：基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示网络侧设备估计的所述终端剩余的AI模型计算资源；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。
第九方面，提供了一种AI算力上报***，包括：终端及网络侧设备，所述终端可用于执行如第一方面所述的方法的步骤，所述网络侧设备可用于执行如第二方面所述的方法的步骤。
第十方面，提供了一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤，或者实现如第二方面所述的方法的步骤。
第十一方面，提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如第一方面所述的方法，或实现如第二方面所述的方法。
第十二方面，提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的方法的步骤，或者实现如第二方面所述的方法的步骤。
在本申请实施例中，终端通过获取当前剩余的可用于AI模型相关操作的AI模型 计算资源，即第一AI算力信息，然后向网络侧设备上报终端的第一AI算力信息，使得网络侧设备获得准确的终端剩余算力，从而网络侧设备可以基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
附图说明
图1是本申请实施例可应用的无线通信***的示意图；
图2是本申请实施例提供的神经网络的结构示意图；
图3是本申请实施例提供的神经元的计算逻辑示意图；
图4是本申请实施例提供的AI算力上报方法的流程示意图之一；
图5是本申请实施例提供的AI算力上报方法的流程示意图之二；
图6是本申请实施例提供的AI算力上报方法的信令交互示意图；
图7是本申请实施例提供的AI算力上报装置的结构示意图之一；
图8是本申请实施例提供的AI算力上报装置的结构示意图之二；
图9是本申请实施例提供的通信设备的结构示意图；
图10是本申请实施例提供的终端的结构示意图；
图11是本申请实施例提供的网络侧设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员所获得的所有其他实施例，都属于本申请保护的范围。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象，而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施，且“第一”、“第二”所区别的对象通常为一类，并不限定对象的个数，例如第一对象可以是一个，也可以是多个。此外，说明书以及权利要求中“和/或”表示所连接对象的至少其中之一，字符“/”一般表示前后关联对象是一种“或”的关系。本申请的说明书和权利要求书中的术语“指示”既可以是一个明确的指示，也可以是一个隐含的指示。其中，明确的指示可以理解为，发送方在发送的指示中明确告知了接收方需要执行的操作或请求结果；隐含的指示可以理解为，接收方根据发送方发送的指示进行判断，根据判断结果确定需要执行的操作或请求结果。
值得指出的是，本申请实施例所描述的技术不限于长期演进型(Long Term Evolution，LTE)/LTE的演进(LTE-Advanced，LTE-A)***，还可用于其他无线通信***，诸如码分多址(Code Division Multiple Access，CDMA)、时分多址(Time Division Multiple Access，TDMA)、频分多址(Frequency Division Multiple Access，FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access，OFDMA)、单载波频分多址(Single-carrier Frequency Division Multiple Access，SC-FDMA)和其他***。本申请实施例中的术语“***”和“网络”常被可互换地使用，所描述的技术既可用于以上提及的***和无线电技术，也可用于其他***和无线电技术。以下描述出于示例目的描述了新空口(New Radio，NR)***，并且在以下大部分描述中使用NR术语，但是这些技术也可应用于NR***应用以外的通信***，如第6代(6th Generation，6G)通信***。
图1是本申请实施例可应用的无线通信***的示意图，图1示出的无线通信***包括终端11和网络侧设备12。其中，终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字 助理(Personal Digital Assistant，PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer，UMPC)、移动上网装置(Mobile Internet Device，MID)、增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(VUE)、行人终端(PUE)、智能家居(具有无线通信功能的家居设备，如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(personal computer，PC)、柜员机或者自助机等终端侧设备，可穿戴式设备包括：智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。除了上述终端设备，终端11也可以是终端内的芯片，例如调制解调器(Modem)芯片，***级芯片(System on Chip，SoC)。需要说明的是，在本申请实施例并不限定终端11的具体类型。
网络侧设备12可以包括接入网设备或核心网设备，其中，接入网设备也可以称为无线接入网设备、无线接入网(Radio Access Network，RAN)、无线接入网功能或无线接入网单元。接入网设备可以包括基站、WLAN接入点或WiFi节点等，基站可被称为节点B、演进节点B(eNB)、接入点、基收发机站(Base Transceiver Station，BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set，BSS)、扩展服务集(Extended Service Set，ESS)、家用B节点、家用演进型B节点、发送接收点(Transmitting Receiving Point，TRP)或所述领域中其他某个合适的术语，只要达到相同的技术效果，所述基站不限于特定技术词汇，需要说明的是，在本申请实施例中仅以NR***中的基站为例进行介绍，并不限定基站的具体类型。核心网设备可以包含但不限于如下至少一项：核心网节点、核心网功能、移动管理实体(Mobility Management Entity，MME)、接入移动管理功能(Access and Mobility Management Function，AMF)、会话管理功能(Session Management Function，SMF)、用户平面功能(User Plane Function，UPF)、策略控制功能(Policy Control Function，PCF)、策略与计费规则功能单元(Policy and Charging Rules Function，PCRF)、边缘应用服务发现功能(Edge Application Server Discovery Function，EASDF)、统一数据管理(Unified Data Management，UDM)，统一数据仓储(Unified Data Repository，UDR)、归属用户服务器(Home Subscriber Server，HSS)、集中式网络配置(Centralized network configuration，CNC)、网络存储功能(Network Repository Function，NRF)，网络开放功能(Network Exposure Function，NEF)、本地NEF(Local NEF，或L-NEF)、绑定支持功能(Binding Support Function，BSF)、应用功能(Application Function，AF)、位置管理功能(location manage function，LMF)、增强服务移动定位中心(Enhanced Serving Mobile Location Centre，E-SMLC)、网络数据分析功能(network data analytics function，NWDAF)等。需要说明的是，在本申请实施例中仅以NR***中的核心网设备为例进行介绍，并不限定核心网设备的具体类型。
为了便于更加清晰地理解本申请各实施例提供的技术方案，首先对一些相关的背景知识进行如下介绍。
人工智能(AI)
人工智能(AI)目前在各个领域获得了广泛的应用，将人工智能融入无线通信网络，显著提升吞吐量、时延以及用户容量等技术指标是未来的无线通信网络的重要任务。AI模块有多种实现方式，例如神经网络、决策树、支持向量机、贝叶斯分类器等。本申请以神经网络为例进行说明，但是并不限定AI模块的具体类型。
图2是本申请实施例提供的神经网络的结构示意图，如图2所示，一个神经网络包括输入层、隐层及输出层；其中，X1、X2、Xn为神经网络的输入，Y为神经网 络的输出。
其中，神经网络由神经元组成，图3是本申请实施例提供的神经元的计算逻辑示意图，如图3所示，a1、ak、aK为输入，w1、wk、wK为权值(乘性系数)，b为偏置(加性系数)，σ(z)为激活函数。常见的激活函数包括Sigmoid、tanh、线性整流函数(又称修正线性单元(Rectified Linear Unit，ReLU))等等；z可以通过以下公式(1)表示：
z＝a1w1+...+akwk+...+aKwK+b         (1)
神经网络的参数通过梯度优化算法进行优化。梯度优化算法是一类最小化或者最大化目标函数(有时候也叫损失函数)的算法，而目标函数往往是模型参数和数据的数学组合。
例如给定数据X和其对应的标签Y，我们构建一个神经网络模型f(.)，有了模型后，根据输入X就可以得到预测输出f(x)，并且可以计算出预测值和真实值之间的差距(f(x)-Y)，这个就是损失函数。我们的目的是找到合适的w、b使上述的损失函数的值达到最小，损失值越小，则说明我们的模型越接近于真实情况。
目前常见的优化算法，基本都是基于误差反向传播(error Back Propagation，BP)算法。BP算法的基本思想是，学习过程由信号的正向传播与误差的反向传播两个过程组成。正向传播时，输入样本从输入层传入，经各隐层逐层处理后，传向输出层。若输出层的实际输出与期望的输出不符，则转入误差的反向传播阶段。误差反传是将输出误差以某种形式通过隐层向输入层逐层反传，并将误差分摊给各层的所有单元，从而获得各层单元的误差信号，此误差信号即作为修正各单元权值的依据。这种信号正向传播与误差反向传播的各层权值调整过程，是周而复始地进行的。权值不断调整的过程，也就是网络的学习训练过程。此过程一直进行到网络输出的误差减少到可接受的程度，或进行到预先设定的学习次数为止。
常见的优化算法有梯度下降(Gradient Descent)、随机梯度下降(Stochastic Gradient Descent，SGD)、小批量梯度下降(mini-batch gradient descent)、动量法(Momentum)、带动量的随机梯度下降(Nesterov)、自适应梯度下降(ADAptive GRADient descent，Adagrad)、Adadelta、均方根误差降速(root mean square prop，RMSprop)、自适应动量估计(Adaptive Moment Estimation，Adam)等。
这些优化算法在误差反向传播时，都是根据损失函数得到的误差/损失，对当前神经元求导数/偏导，加上学习速率、之前的梯度/导数/偏导等影响，得到梯度，将梯度传给上一层。
相关技术中，网络侧可以指示UE使用特定的AI模型。但是，目前尚无UE上报自己剩余AI算力的方法，导致网络侧无法准确估计UE剩余AI算力。如果网络侧估计的UE剩余算力多于实际，则网络侧指示UE使用过于复杂的AI模型，造成UE无法正常运行AI模型；如果网络侧估计的UE剩余算力低于实际，则网络侧指示UE使用过于简单的AI模型，造成UE AI算力的浪费。也就是说，网络侧无法准确估计UE剩余AI算力，导致UE AI算力的利用率较低，影响通信***的性能。
综上所述，针对上述存在的问题，本申请实施例提供了一种AI算力上报方法、终端及网络侧设备，能够提高通信***的性能。
图4是本申请实施例提供的AI算力上报方法的流程示意图之一，如图4所示，该方法包括步骤401-402；其中：
步骤401、终端获取第一AI算力信息；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。
需要说明的是，本申请实施例可应用于基于AI模型进行通信的场景中。所述终端包括但不限于上述所列举的终端11的类型；网络侧设备包括但不限于上述所列举的网络侧设备12的类型；本申请对此并不限定。
由于网络侧设备无法准确估计终端剩余AI算力，导致终端AI算力的利用率较低，影响通信***的性能；因此，为了提升终端AI算力的利用率，提升通信***的性能，在本实施例中，首先终端需要获取第一AI算力信息。
可选地，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
AI unit算力单元是指衡量AI模型计算资源的单元，其中，AI模型计算资源例如是AI模型的运算次数。
可选地，所述计算资源单位，包括以下至少一项：
a)单次操作(operation，OP)的运算次数；
b)万亿操作(Trillion operation，TOP)的运算次数；
c)浮点操作(FLoating point Operation，FLOP)的运算次数；或者每秒浮点操作(FLoating point Operations per second，FLOPS)的运算次数；
d)内存访问成本(Memory Access Cost，MAC)；
e)乘加数操作(multiply-accumulate operation，MACC)的运算次数。
可选地，所述AI unit的定义满足以下至少一项：
a)由协议约定；
b)由所述终端定义；
c)由所述网络侧设备配置。
其中，所述AI unit的定义包括：N1的取值，和/或，计算资源单位的类型。
步骤402、所述终端向网络侧设备发送所述第一AI算力信息。
在本实施例中，终端需要将获取到的第一AI算力信息发送至网络侧设备。相应地，网络侧设备接收到第一AI算力信息之后，需要基于第一AI算力信息，获取终端对应的第二AI算力信息；其中，第二AI算力信息用于指示网络侧设备估计的终端剩余的AI模型计算资源。
也就是说，网络侧设备基于终端发送的第一AI算力信息，可以实时估计出终端剩余的AI模型计算资源，从而可以基于终端剩余的AI模型计算资源为终端下发，或者向终端指示合适的第一AI模型。
在本申请实施例提供的AI算力上报方法中，终端通过获取当前剩余的可用于AI模型相关操作的AI模型计算资源，即第一AI算力信息，然后向网络侧设备上报终端的第一AI算力信息，使得网络侧设备获得准确的终端剩余算力，从而网络侧设备可以基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，所述终端获取第一AI算力信息，可以通过以下任一种方式实现：
方式1、所述终端基于终端配置信息，确定所述第一AI算力信息。
在实际应用中，在终端配置信息中预配置有终端的第一AI算力信息，终端可以直接从终端配置信息中获取第一AI算力信息。
方式2、所述终端基于终端配置信息及已占用AI算力信息，确定所述第一AI算力信息。
在实际应用中，在终端当前已占用AI算力信息的情况下，可以用终端配置信息中预配置的总AI算力信息减去已占用AI算力信息，进而得到第一AI算力信息。
可选地，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换 算得到。
在本实施例中，在AI模型注册、AI模型配置、AI模型传输及AI模型传递时，AI模型的模型配置信息或关联信息中，包括AI模型占用的AI unit的数目；其中AI模型占用的AI unit的数目，由AI模型的计算复杂度换算得到。
例如，一个AI unit用于指示5 TOP的运算次数，一个AI模型的计算复杂度为15 TOP的运算次数，则AI模型占用的AI unit为15 TOP的运算次数除以5 TOP的运算次数，即AI模型占用3个AI unit。
可选地，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
AI模型的计算复杂度通过AI模型占用的AI unit的数目来衡量，而每个AI unit用于指示N1个计算资源单位；因此，AI模型的计算复杂度可以用N2个计算资源单位来表示。
所述AI模型占用的AI unit的数目，由以下任一种方式得到：
方式1、在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目。
例如，一个AI unit用于指示4 TOP的运算次数，一个AI模型的计算复杂度为15 TOP的运算次数，则AI模型占用的AI unit为15 TOP的运算次数除以4 TOP的运算次数，即AI模型占用3.75个AI unit。
方式2、在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
例如，一个AI unit用于指示10 TOP的运算次数，一个AI模型的计算复杂度为23 TOP的运算次数，则AI模型占用的AI unit为23 TOP的运算次数除以10 TOP的运算次数，然后将计算得到的商进行向上取整，即AI模型占用3个AI unit；或者将计算得到的商进行近似取整，即AI模型占用的2个AI unit。
在上述实施方式中，通过AI模型占用的AI unit的数目来表示AI模型的计算复杂度，可以使终端可以基于AI unit的数目，准确的将当前剩余的可用于AI模型相关操作的AI模型计算资源发送至网络侧设备，使得网络侧设备获得准确的终端剩余算力。
可选地，所述终端向网络侧设备发送所述第一AI算力信息，具体可以通过以下步骤实现：
所述终端在向所述网络侧设备上报所述终端的AI能力信息的过程中，向所述网络侧设备发送所述第一AI算力信息。
也就是说，终端在向网络侧设备上报自己的AI能力时，同时上报/携带自己总的AI unit算力单元。
可选地，所述AI模型计算资源用于以下至少一项AI模型相关操作：
a)基于AI模型的信号处理；
具体地，基于AI模型的信号处理包括信号检测、滤波、均衡等；其中，信号包括解调参考信号(Demodulation Reference Signal，DMRS)、探测参考信号(Sounding Reference Signal，SRS)、同步信号块(Synchronization Signal Block，SSB)、跟踪参考信号(Tracking Reference Signal，TRS)、相位跟踪参考信号(Phase-Tracking Reference Signals，PTRS)、信道状态信息参考信号(Channel State Information-Reference Signal，CSI-RS)等。
b)基于AI模型的信号传输/接收/解调/发送；
具体地，信号传输/接收/解调/发送所包括的信道例如可以是：物理下行控制信道(Physical downlink control channel，PDCCH)、物理下行共享信道(Physical  downlink shared channel，PDSCH)、物理上行控制信道(Physical Uplink Control Channel，PUCCH)、物理上行共享信道(Physical Uplink Shared Channel，PUSCH)、物理随机接入信道(Physical Random Access Channel，PRACH)、物理广播信道(Physical broadcast channel，PBCH)等。
c)基于AI模型的信道状态信息获取；包括：
信道状态信息反馈，包括信道相关信息、信道矩阵相关信息、信道特征信息、信道矩阵特征信息、预编码矩阵指示(Precoding matrix indicator，PMI)、秩指示(Rank indicator，RI)、CSI-RS资源指示(CSI-RS Resource Indicator，CRI)、信道质量指示(Channel quality indicator，CQI)、层指示(Layer Indicator，LI)等。
FDD上下行部分互易性。对于频分双工(Frequency Division Duplexing，FDD)***，根据部分互异性，基站根据上行信道获取角度和时延信息，可以通过CSI-RS预编码或者直接指示的方法，将角度信息和时延信息通知UE，UE根据基站的指示上报或者在基站的指示范围内选择并上报，从而减少UE的计算量和CSI上报的开销。
d)基于AI模型的波束管理；包括：波束测量、波束上报、波束预测、波束失败检测、波束失败恢复、波束失败恢复中的新波束指示。
e)基于AI模型的信道预测；包括信道状态信息的预测、波束预测。
f)基于AI模型的干扰抑制；包括小区内干扰、小区间干扰、带外干扰、交调干扰等。
g)基于AI模型的定位；
例如，通过参考信号(例如SRS)，估计出的UE的具***置(包括水平位置和或垂直位置)或未来可能的轨迹，或辅助位置估计或轨迹估计的信息。
h)基于AI模型的高层业务和参数的预测和管理；包括吞吐量、所需数据包大小、业务需求、移动速度、噪声信息等。
i)基于AI模型的控制信令解析；例如功率控制的相关信令，波束管理的相关信令。
图5是本申请实施例提供的AI算力上报方法的流程示意图之二，如图5所示，该方法包括步骤501-502；其中：
步骤501、网络侧设备接收终端发送的第一AI算力信息；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。
需要说明的是，本申请实施例可应用于基于AI模型进行通信的场景中。所述终端包括但不限于上述所列举的终端11的类型；网络侧设备包括但不限于上述所列举的网络侧设备12的类型；本申请对此并不限定。
由于网络侧设备无法准确估计终端剩余AI算力，导致终端AI算力的利用率较低，影响通信***的性能；因此，为了提升终端AI算力的利用率，提升通信***的性能，在本实施例中，网络侧设备需要接收终端发送的第一AI算力信息。
可选地，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
AI unit算力单元是指衡量AI模型计算资源的单元，其中，AI模型计算资源例如是AI模型的运算次数。
可选地，所述计算资源单位，包括以下至少一项：
a)单次操作的运算次数；
b)万亿操作的运算次数；
c)浮点操作的运算次数；
d)内存访问成本；
e)乘加数操作的运算次数。
可选地，所述AI unit的定义满足以下至少一项：
a)由协议约定；
b)由所述终端定义；
c)由所述网络侧设备配置。
其中，所述AI unit的定义包括：N1的取值，和/或，计算资源单位的类型。
步骤502、所述网络侧设备基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示所述网络侧设备估计的所述终端剩余的AI模型计算资源。
在本实施例中，网络侧设备基于终端发送的第一AI算力信息，可以实时估计出终端剩余的AI模型计算资源，从而可以基于终端剩余的AI模型计算资源为终端下发，或者向终端指示合适的第一AI模型。
在本申请实施例提供的AI算力上报方法中，网络侧设备通过接收终端发送的第一AI算力信息，使得网络侧设备获得准确的终端剩余算力，从而网络侧设备可以基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换算得到。
在本实施例中，在AI模型注册、AI模型配置、AI模型传输及AI模型传递时，AI模型的模型配置信息或关联信息中，包括AI模型占用的AI unit的数目；其中AI模型占用的AI unit的数目，由AI模型的计算复杂度换算得到。
可选地，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
所述AI模型占用的AI unit的数目，由以下任一种方式得到：
方式1、在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目。
方式2、在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
可选地，在网络侧设备获取到终端对应的第二AI算力信息之后，还需要基于终端剩余的AI模型计算资源，向终端配置或指示第一AI模型，以提高终端AI算力的利用率。具体可以通过以下任一种方式实现：
方式1、在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，所述网络侧设备将所述第一AI模型下发给所述终端。
具体地，网络侧设备向终端下发的第一AI模型的复杂度不能大于，或者不能大于等于终端当前空闲AI unit(即第二AI算力信息)。
例如，网络侧设备估计的终端剩余的AI模型计算资源为5个AI unit；则网络侧设备向终端下发的第一AI模型占用的AI unit的数目应该小于5。
方式2、在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，所述网络侧设备指示所述终端激活所述第一AI模型。
具体地，网络侧设备可以向终端发送指示信息，以使终端激活第一AI模型；可以理解的是，网络侧设备向终端指示的第一AI模型的复杂度不能大于，或者不能大于等于终端当前空闲AI unit(即第二AI算力信息)。
例如，网络侧设备估计的终端剩余的AI模型计算资源为5个AI unit；则网络侧设备向终端指示的第一AI模型占用的AI unit的数目应该小于5。
方式3、在第一AI模型占用的AI unit的数目与第二AI模型占用的AI unit的数目的第一差值小于或不大于所述第二AI算力信息的情况下，所述网络侧设备指示所述终端去激活所述第二AI模型，及激活所述第一AI模型。
具体地，在网络侧设备指示终端从第二AI模型切换至第一AI模型(即，网络侧设备指示终端去激活当前使用的第一AI模型，激活第二AI模型)的情况下，则第一AI模型超过第二AI模型的复杂度不能大于，或者不能大于等于终端当前空闲AI unit(即第二AI算力信息)。
可以理解的是，若第一AI模型的复杂度低于当前使用的第二AI模型，则网络侧设备可以直接指示终端去激活所述第二AI模型，及激活所述第一AI模型。
在上述实施方式中，网络侧设备可以准确的基于终端剩余的AI模型计算资源(即第二AI算力信息)进行AI模型的配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，在网络侧设备将第一AI模型下发给所述终端之后，还需要对第二AI算力信息进行更新，具体可以通过以下步骤实现：
所述网络侧设备从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
例如，网络侧设备估计的终端剩余的AI模型计算资源为5个AI unit，第一AI模型占用了2个AI unit；则更新后的第二AI算力信息即3个AI unit。
可选地，在网络侧设备指示所述终端激活所述第一AI模型之后，还需要对第二AI算力信息进行更新，具体可以通过以下步骤实现：
所述网络侧设备从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
可选地，在网络侧设备指示所述终端去激活所述第二AI模型，及激活所述第一AI模型之后，还需要对第二AI算力信息进行更新，具体可以通过以下步骤实现：
所述网络侧设备计算所述第一AI模型占用的AI unit的数目与所述第二AI模型占用的AI unit的数目的第一差值；从所述第二AI算力信息中减去所述第一差值，得到更新后的第二AI算力信息。
需要说明的是，第一差值可以为负数。例如，网络侧设备估计的终端剩余的AI模型计算资源为5个AI unit，第一AI模型占用了2个AI unit，第二AI模型占用了3个AI unit，第一差值为-1个AI unit，则更新后的第二AI算力信息即6个AI unit。
可选地，在网络侧设备指示所述终端去激活第三AI模型的情况下，所述网络侧设备从所述第二AI算力信息中加上所述第三AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
例如，网络侧设备估计的终端剩余的AI模型计算资源为5个AI unit，第三AI模型占用了2个AI unit；网络侧设备指示终端去激活第三AI模型的情况下，更新后的第二AI算力信息即7个AI unit。
在上述实施方式中，网络侧设备可以实现对第二AI算力信息的实时更新，从而使网络侧设备可以进一步地基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，所述AI模型计算资源用于以下至少一项AI模型相关操作：
a)基于AI模型的信号处理；
b)基于AI模型的信号传输/接收/解调/发送；
c)基于AI模型的信道状态信息获取；
d)基于AI模型的波束管理；
e)基于AI模型的信道预测；
f)基于AI模型的干扰抑制；
g)基于AI模型的定位；
h)基于AI模型的高层业务和参数的预测和管理；
i)基于AI模型的控制信令解析。
图6是本申请实施例提供的AI算力上报方法的信令交互示意图。如图6所示，具体包括步骤1-步骤7：
步骤1、终端获取第一AI算力信息。
具体地，第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
计算资源单位，包括以下至少一项：a)单次操作的运算次数；b)万亿操作的运算次数；c)浮点操作的运算次数；d)内存访问成本；e)乘加数操作的运算次数。
AI unit的定义满足以下至少一项：a)由协议约定；b)由所述终端定义；c)由所述网络侧设备配置。
步骤2、终端向网络侧设备发送终端的AI能力信息，其中，终端的AI能力信息中包括第一AI算力信息。
步骤3、网络侧设备基于第一AI算力信息，获取终端对应的第二AI算力信息。
具体地，第二AI算力信息用于指示网络侧设备估计的终端剩余的AI模型计算资源。
需要说明的是，在执行完毕步骤3之后，开始执行步骤4至步骤6中至少一项。
步骤4、网络侧设备向终端下发第一AI模型。
具体地，在第一AI模型占用的AI unit的数目小于或不大于第二AI算力信息的情况下，网络侧设备将第一AI模型下发给终端。
步骤5、网络侧设备向终端发送第一指示信息；第一指示信息用于指示第一AI模型。
具体地，在第一AI模型占用的AI unit的数目小于或不大于第二AI算力信息的情况下，网络侧设备指示终端激活第一AI模型。
步骤6、网络侧设备向终端发送第二指示信息；第二指示信息用于指示终端去激活第二AI模型，及激活第一AI模型。
具体地，在第一AI模型占用的AI unit的数目与第二AI模型占用的AI unit的数目的第一差值小于或不大于第二AI算力信息的情况下，网络侧设备指示终端去激活第二AI模型，及激活第一AI模型。
步骤7、网络侧设备更新第二AI算力信息，得到更新后的第二AI算力信息。
具体地，网络侧设备在执行完毕步骤4的情况下，需要从第二AI算力信息中减去第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
网络侧设备在执行完毕步骤5的情况下，需要从第二AI算力信息中减去第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
网络侧设备在执行完毕步骤6的情况下，需要计算第一AI模型占用的AI unit的数目与第二AI模型占用的AI unit的数目的第一差值；然后从第二AI算力信息中减去第一差值，得到更新后的第二AI算力信息。
在网络侧设备指示终端去激活第三AI模型的情况下，网络侧设备需要从第二AI算力信息中加上第三AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
本申请实施例提供的AI算力上报方法，执行主体可以为AI算力上报装置。本申 请实施例中以AI算力上报装置执行AI算力上报方法为例，说明本申请实施例提供的AI算力上报装置。
图7是本申请实施例提供的AI算力上报装置的结构示意图之一，如图7所示，该AI算力上报装置700，应用于终端，包括：
第一获取模块701，用于获取第一AI算力信息；
发送模块702，用于向网络侧设备发送所述第一AI算力信息；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
本申请实施例提供的AI算力上报装置中，通过获取当前剩余的可用于AI模型相关操作的AI模型计算资源，即第一AI算力信息，然后向网络侧设备上报终端的第一AI算力信息，使得网络侧设备获得准确的终端剩余算力，从而网络侧设备可以基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
可选地，所述计算资源单位，包括以下至少一项：
单次操作的运算次数；
万亿操作的运算次数；
浮点操作的运算次数；
内存访问成本；
乘加数操作的运算次数。
可选地，所述AI unit的定义满足以下至少一项：由协议约定；由所述终端定义；由所述网络侧设备配置。
可选地，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换算得到。
可选地，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
所述AI模型占用的AI unit的数目，由以下任一种方式得到：
在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目；
在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
可选地，所述第一获取模块701，进一步用于以下任一项：
基于终端配置信息，确定所述第一AI算力信息；
基于终端配置信息及已占用AI算力信息，确定所述第一AI算力信息。
可选地，发送模块702，进一步用于：
在向所述网络侧设备上报所述终端的AI能力信息的过程中，向所述网络侧设备发送所述第一AI算力信息。
可选地，所述AI模型计算资源用于以下至少一项AI模型相关操作：
基于AI模型的信号处理；
基于AI模型的信号传输/接收/解调/发送；
基于AI模型的信道状态信息获取；
基于AI模型的波束管理；
基于AI模型的信道预测；
基于AI模型的干扰抑制；
基于AI模型的定位；
基于AI模型的高层业务和参数的预测和管理；
基于AI模型的控制信令解析。
图8是本申请实施例提供的AI算力上报装置的结构示意图之二，如图8所示，该AI算力上报装置800，应用于网络侧设备，包括：
接收模块801，用于接收终端发送的第一AI算力信息；
第二获取模块802，用于基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示网络侧设备估计的所述终端剩余的AI模型计算资源；
其中，所述第一AI算力信息用于指示以下至少一项：
所述终端当前剩余的AI模型计算资源；
所述终端当前可用的AI模型计算资源；
所述终端所有的AI模型计算资源；
所述终端所有的可用于无线通信的AI模型计算资源。
本申请实施例提供的AI算力上报装置中，通过接收终端发送的第一AI算力信息，使得网络侧设备获得准确的终端剩余算力，从而网络侧设备可以基于准确的终端剩余算力进行AI配置或指示，能够提升终端AI算力的利用率，提升通信***的性能。
可选地，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
可选地，所述计算资源单位，包括以下至少一项：
单次操作的运算次数；
万亿操作的运算次数；
浮点操作的运算次数；
内存访问成本；
乘加数操作的运算次数。
可选地，所述AI unit的定义满足以下至少一项：由协议约定；由所述终端定义；由所述网络侧设备配置。
可选地，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换算得到。
可选地，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
所述AI模型占用的AI unit的数目，由以下任一种方式得到：
在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目；
在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
可选地，所述装置还包括以下至少一项：
下发模块，用于在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，将所述第一AI模型下发给所述终端；
第一指示模块，用于在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，指示所述终端激活所述第一AI模型；
第二指示模块，用于在第一AI模型占用的AI unit的数目与第二AI模型占用的AI unit的数目的第一差值小于或不大于所述第二AI算力信息的情况下，指示所述终端去激活所述第二AI模型，及激活所述第一AI模型。
可选地，在将所述第一AI模型下发给所述终端之后，所述装置还包括：
第一更新模块，用于从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
可选地，在指示所述终端激活所述第一AI模型之后，所述装置还包括：
第二更新模块，用于从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
可选地，在指示所述终端去激活所述第二AI模型，及激活所述第一AI模型之后，所述装置还包括：
第三更新模块，用于计算所述第一AI模型占用的AI unit的数目与所述第二AI模型占用的AI unit的数目的第一差值；从所述第二AI算力信息中减去所述第一差值，得到更新后的第二AI算力信息。
可选地，所述装置还包括：
第三指示模块，用于指示所述终端去激活第三AI模型；
第四更新模块，用于从所述第二AI算力信息中加上所述第三AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
可选地，所述AI模型计算资源用于以下至少一项AI模型相关操作：
基于AI模型的信号处理；
基于AI模型的信号传输/接收/解调/发送；
基于AI模型的信道状态信息获取；
基于AI模型的波束管理；
基于AI模型的信道预测；
基于AI模型的干扰抑制；
基于AI模型的定位；
基于AI模型的高层业务和参数的预测和管理；
基于AI模型的控制信令解析。
本申请实施例中的AI算力上报装置可以是电子设备，例如具有操作***的电子设备，也可以是电子设备中的部件，例如集成电路或芯片。该电子设备可以是终端，也可以为除终端之外的其他设备。示例性的，终端可以包括但不限于上述所列举的终端11的类型，其他设备可以为服务器、网络附属存储器(Network Attached Storage，NAS)等，本申请实施例不作具体限定。
本申请实施例提供的AI算力上报装置能够实现图4至图5的方法实施例实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
图9是本申请实施例提供的通信设备的结构示意图，如图9所示，该通信设备900，包括处理器901和存储器902，存储器902上存储有可在所述处理器901上运行的程序或指令，例如，该通信设备900为终端时，该程序或指令被处理器901执行时实现上述AI算力上报方法实施例的各个步骤，且能达到相同的技术效果。该通信设备900为网络侧设备时，该程序或指令被处理器901执行时实现上述AI算力上报方法实施例的各个步骤，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种终端，包括处理器和通信接口，所述处理器用于：获取第一AI算力信息，所述通信接口用于：向网络侧设备发送所述第一AI算力信息；其 中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。该终端实施例与上述终端侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该终端实施例中，且能达到相同的技术效果。
图10是本申请实施例提供的终端的结构示意图，如图10所示，该终端1000包括但不限于：射频单元1001、网络模块1002、音频输出单元1003、输入单元1004、传感器1005、显示单元1006、用户输入单元1007、接口单元1008、存储器1009以及处理器1010等中的至少部分部件。
本领域技术人员可以理解，终端1000还可以包括给各个部件供电的电源(比如电池)，电源可以通过电源管理***与处理器1010逻辑相连，从而通过电源管理***实现管理充电、放电、以及功耗管理等功能。图10中示出的终端结构并不构成对终端的限定，终端可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置，在此不再赘述。
应理解的是，本申请实施例中，输入单元1004可以包括图形处理单元(Graphics Processing Unit，GPU)10041和麦克风10042，图形处理器10041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元1006可包括显示面板10061，可以采用液晶显示器、有机发光二极管等形式来配置显示面板10061。用户输入单元1007包括触控面板10071以及其他输入设备10072中的至少一种。触控面板10071，也称为触摸屏。触控面板10071可包括触摸检测装置和触摸控制器两个部分。其他输入设备10072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆，在此不再赘述。
本申请实施例中，射频单元1001接收来自网络侧设备的下行数据后，可以传输给处理器1010进行处理；另外，射频单元1001可以向网络侧设备发送上行数据。通常，射频单元1001包括但不限于天线、放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器1009可用于存储软件程序或指令以及各种数据。存储器1009可主要包括存储程序或指令的第一存储区和存储数据的第二存储区，其中，第一存储区可存储操作***、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播放功能等)等。此外，存储器1009可以包括易失性存储器或非易失性存储器，或者，存储器1009可以包括瞬态和非瞬态存储器。其中，非易失性存储器可以是只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable ROM，PROM)、可擦除可编程只读存储器(Erasable PROM，EPROM)、电可擦除可编程只读存储器(Electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory，RAM)，静态随机存取存储器(Static RAM，SRAM)、动态随机存取存储器(Dynamic RAM，DRAM)、同步动态随机存取存储器(Synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM，DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(Synch link DRAM，SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM，DRRAM)。本申请实施例中的存储器x09包括但不限于这些和任意其它适合类型的存储器。
处理器1010可包括一个或多个处理单元；可选的，处理器x10集成应用处理器和调制解调处理器，其中，应用处理器主要处理涉及操作***、用户界面和应用程序 等的操作，调制解调处理器主要处理无线通信信号，如基带处理器。可以理解的是，上述调制解调处理器也可以不集成到处理器1010中。
本申请实施例还提供一种网络侧设备，包括处理器和通信接口，所述通信接口用于：接收终端发送的第一AI算力信息，所述处理器用于：基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示网络侧设备估计的所述终端剩余的AI模型计算资源；其中，所述第一AI算力信息用于指示以下至少一项：所述终端当前剩余的AI模型计算资源；所述终端当前可用的AI模型计算资源；所述终端所有的AI模型计算资源；所述终端所有的可用于无线通信的AI模型计算资源。该网络侧设备实施例与上述网络侧设备方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中，且能达到相同的技术效果。
图11是本申请实施例提供的网络侧设备的结构示意图，如图11所示，该网络侧设备1100包括：天线1101、射频装置1102、基带装置1103、处理器1104和存储器1105。天线1101与射频装置1102连接。在上行方向上，射频装置1102通过天线1101接收信息，将接收的信息发送给基带装置1103进行处理。在下行方向上，基带装置1103对要发送的信息进行处理，并发送给射频装置1102，射频装置1102对收到的信息进行处理后经过天线1101发送出去。
以上实施例中网络侧设备执行的方法可以在基带装置1103中实现，该基带装置1103包括基带处理器。
基带装置1103例如可以包括至少一个基带板，该基带板上设置有多个芯片，如图11所示，其中一个芯片例如为基带处理器，通过总线接口与存储器1105连接，以调用存储器1105中的程序，执行以上方法实施例中所示的网络设备操作。
该网络侧设备还可以包括网络接口1106，该接口例如为通用公共无线接口(common public radio interface，CPRI)。
具体地，本发明实施例的网络侧设备1100还包括：存储在存储器1105上并可在处理器1104上运行的指令或程序，处理器1104调用存储器1105中的指令或程序执行如上所述的AI算力上报方法，并达到相同的技术效果，为避免重复，故不在此赘述。
本申请实施例还提供了一种AI算力上报***，包括：终端及网络侧设备，所述终端可用于执行如上所述图4所示的AI算力上报方法的步骤，所述网络侧设备可用于执行如上所述图5所示的AI算力上报方法的步骤。
本申请实施例还提供一种可读存储介质，所述可读存储介质可以是以易失性的，也可以是非易失性的，所述可读存储介质上存储有程序或指令，该程序或指令被处理器执行时实现上述AI算力上报方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
其中，所述处理器为上述实施例中所述的终端中的处理器。所述可读存储介质，包括计算机可读存储介质，如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。
本申请实施例另提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现上述AI算力上报方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
应理解，本申请实施例提到的芯片还可以称为***级芯片，***芯片，芯片***或片上***芯片等。
本申请实施例另提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现上述AI 算力上报方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外，需要指出的是，本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能，还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能，例如，可以按不同于所描述的次序来执行所描述的方法，并且还可以添加、省去、或组合各种步骤。另外，参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以计算机软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述，但是本申请并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本申请的启示下，在不脱离本申请宗旨和权利要求所保护的范围情况下，还可做出很多形式，均属于本申请的保护之内。

Claims (26)

  1. 一种人工智能AI算力上报方法，包括：
    终端获取第一AI算力信息；
    所述终端向网络侧设备发送所述第一AI算力信息；
    其中，所述第一AI算力信息用于指示以下至少一项：
    所述终端当前剩余的AI模型计算资源；
    所述终端当前可用的AI模型计算资源；
    所述终端所有的AI模型计算资源；
    所述终端所有的可用于无线通信的AI模型计算资源。
  2. 根据权利要求1所述的AI算力上报方法，其中，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
  3. 根据权利要求2所述的AI算力上报方法，其中，所述计算资源单位，包括以下至少一项：
    单次操作的运算次数；
    万亿操作的运算次数；
    浮点操作的运算次数；
    内存访问成本；
    乘加数操作的运算次数。
  4. 根据权利要求2或3所述的AI算力上报方法，其中，所述AI unit的定义满足以下至少一项：由协议约定；由所述终端定义；由所述网络侧设备配置。
  5. 根据权利要求2至4任一项所述的AI算力上报方法，其中，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换算得到。
  6. 根据权利要求5所述的AI算力上报方法，其中，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
    所述AI模型占用的AI unit的数目，由以下任一种方式得到：
    在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目；
    在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
  7. 根据权利要求1至6任一项所述的AI算力上报方法，其中，所述终端获取第一AI算力信息，包括以下任一项：
    所述终端基于终端配置信息，确定所述第一AI算力信息；
    所述终端基于终端配置信息及已占用AI算力信息，确定所述第一AI算力信息。
  8. 根据权利要求1至7任一项所述的AI算力上报方法，其中，所述终端向网络侧设备发送所述第一AI算力信息，包括：
    所述终端在向所述网络侧设备上报所述终端的AI能力信息的过程中，向所述网络侧设备发送所述第一AI算力信息。
  9. 根据权利要求1至8任一项所述的AI算力上报方法，其中，所述AI模型计算资源用于以下至少一项AI模型相关操作：
    基于AI模型的信号处理；
    基于AI模型的信号传输/接收/解调/发送；
    基于AI模型的信道状态信息获取；
    基于AI模型的波束管理；
    基于AI模型的信道预测；
    基于AI模型的干扰抑制；
    基于AI模型的定位；
    基于AI模型的高层业务和参数的预测和管理；
    基于AI模型的控制信令解析。
  10. 一种人工智能AI算力上报方法，包括：
    网络侧设备接收终端发送的第一AI算力信息；
    所述网络侧设备基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示所述网络侧设备估计的所述终端剩余的AI模型计算资源；
    其中，所述第一AI算力信息用于指示以下至少一项：
    所述终端当前剩余的AI模型计算资源；
    所述终端当前可用的AI模型计算资源；
    所述终端所有的AI模型计算资源；
    所述终端所有的可用于无线通信的AI模型计算资源。
  11. 根据权利要求10所述的AI算力上报方法，其中，所述第一AI算力信息包括M个AI unit算力单元，M为整数或小数；每个AI unit用于指示N1个计算资源单位，N1为正整数或小数。
  12. 根据权利要求11所述的AI算力上报方法，其中，所述计算资源单位，包括以下至少一项：
    单次操作的运算次数；
    万亿操作的运算次数；
    浮点操作的运算次数；
    内存访问成本；
    乘加数操作的运算次数。
  13. 根据权利要求11或12所述的AI算力上报方法，其中，所述AI unit的定义满足以下至少一项：由协议约定；由所述终端定义；由所述网络侧设备配置。
  14. 根据权利要求11至13任一项所述的AI算力上报方法，其中，在所述AI模型的模型配置信息或关联信息中，包括所述AI模型占用的AI unit的数目；所述AI模型占用的AI unit的数目，由所述AI模型的计算复杂度换算得到。
  15. 根据权利要求14所述的AI算力上报方法，其中，所述AI模型的计算复杂度为N2个计算资源单位，N2为正整数或小数；
    所述AI模型占用的AI unit的数目，由以下任一种方式得到：
    在M为小数的情况下，计算N2除以N1，得到所述AI模型占用的AI unit的数目；
    在M为整数的情况下，计算N2除以N1，将计算得到的商进行向上取整或近似取整，得到所述AI模型占用的AI unit的数目。
  16. 根据权利要求10所述的AI算力上报方法，其中，所述方法还包括以下至少一项：
    在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，所述网络侧设备将所述第一AI模型下发给所述终端；
    在第一AI模型占用的AI unit的数目小于或不大于所述第二AI算力信息的情况下，所述网络侧设备指示所述终端激活所述第一AI模型；
    在第一AI模型占用的AI unit的数目与第二AI模型占用的AI unit的数目的第一差值小于或不大于所述第二AI算力信息的情况下，所述网络侧设备指示所述终端去 激活所述第二AI模型，及激活所述第一AI模型。
  17. 根据权利要求16所述的AI算力上报方法，其中，在所述网络侧设备将所述第一AI模型下发给所述终端之后，所述方法还包括：
    所述网络侧设备从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
  18. 根据权利要求16所述的AI算力上报方法，其中，在所述网络侧设备指示所述终端激活所述第一AI模型之后，所述方法还包括：
    所述网络侧设备从所述第二AI算力信息中减去所述第一AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
  19. 根据权利要求16所述的AI算力上报方法，其中，在所述网络侧设备指示所述终端去激活所述第二AI模型，及激活所述第一AI模型之后，所述方法还包括：
    所述网络侧设备计算所述第一AI模型占用的AI unit的数目与所述第二AI模型占用的AI unit的数目的第一差值；从所述第二AI算力信息中减去所述第一差值，得到更新后的第二AI算力信息。
  20. 根据权利要求10所述的AI算力上报方法，其中，所述方法还包括：
    所述网络侧设备指示所述终端去激活第三AI模型；
    所述网络侧设备从所述第二AI算力信息中加上所述第三AI模型占用的AI unit的数目，得到更新后的第二AI算力信息。
  21. 根据权利要求10至20任一项所述的AI算力上报方法，其中，所述AI模型计算资源用于以下至少一项AI模型相关操作：
    基于AI模型的信号处理；
    基于AI模型的信号传输/接收/解调/发送；
    基于AI模型的信道状态信息获取；
    基于AI模型的波束管理；
    基于AI模型的信道预测；
    基于AI模型的干扰抑制；
    基于AI模型的定位；
    基于AI模型的高层业务和参数的预测和管理；
    基于AI模型的控制信令解析。
  22. 一种人工智能AI算力上报装置，包括：
    第一获取模块，用于获取第一AI算力信息；
    发送模块，用于向网络侧设备发送所述第一AI算力信息；
    其中，所述第一AI算力信息用于指示以下至少一项：
    所述终端当前剩余的AI模型计算资源；
    所述终端当前可用的AI模型计算资源；
    所述终端所有的AI模型计算资源；
    所述终端所有的可用于无线通信的AI模型计算资源。
  23. 一种人工智能AI算力上报装置，包括：
    接收模块，用于接收终端发送的第一AI算力信息；
    第二获取模块，用于基于所述第一AI算力信息，获取所述终端对应的第二AI算力信息；所述第二AI算力信息用于指示网络侧设备估计的所述终端剩余的AI模型计算资源；
    其中，所述第一AI算力信息用于指示以下至少一项：
    所述终端当前剩余的AI模型计算资源；
    所述终端当前可用的AI模型计算资源；
    所述终端所有的AI模型计算资源；
    所述终端所有的可用于无线通信的AI模型计算资源。
  24. 一种终端，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求1至9任一项所述的AI算力上报方法的步骤。
  25. 一种网络侧设备，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求10至21任一项所述的AI算力上报方法的步骤。
  26. 一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如权利要求1-9任一项所述的AI算力上报方法，或者实现如权利要求10至21任一项所述的AI算力上报方法的步骤。
PCT/CN2023/138240 2022-12-15 2023-12-12 Ai算力上报方法、终端及网络侧设备 WO2024125525A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211616288.6A CN118214750A (zh) 2022-12-15 2022-12-15 Ai算力上报方法、终端及网络侧设备
CN202211616288.6 2022-12-15

Publications (1)

Publication Number Publication Date
WO2024125525A1 true WO2024125525A1 (zh) 2024-06-20

Family

ID=91453021

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/138240 WO2024125525A1 (zh) 2022-12-15 2023-12-12 Ai算力上报方法、终端及网络侧设备

Country Status (2)

Country Link
CN (1) CN118214750A (zh)
WO (1) WO2024125525A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111819872A (zh) * 2020-06-03 2020-10-23 北京小米移动软件有限公司 信息传输方法、装置、通信设备及存储介质
CN113238867A (zh) * 2021-05-19 2021-08-10 浙江凡双科技有限公司 一种基于网络卸载的联邦学习方法
WO2022145551A1 (ko) * 2020-12-29 2022-07-07 엘지전자 주식회사 지능적 신호 송수신 방법 및 이를 위한 장치
CN114756104A (zh) * 2021-01-08 2022-07-15 展讯通信（上海）有限公司 过热处理方法及装置、存储介质
CN114930945A (zh) * 2020-01-14 2022-08-19 Oppo广东移动通信有限公司 信息上报方法、装置、设备和存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114930945A (zh) * 2020-01-14 2022-08-19 Oppo广东移动通信有限公司 信息上报方法、装置、设备和存储介质
CN111819872A (zh) * 2020-06-03 2020-10-23 北京小米移动软件有限公司 信息传输方法、装置、通信设备及存储介质
WO2022145551A1 (ko) * 2020-12-29 2022-07-07 엘지전자 주식회사 지능적 신호 송수신 방법 및 이를 위한 장치
CN114756104A (zh) * 2021-01-08 2022-07-15 展讯通信（上海）有限公司 过热处理方法及装置、存储介质
CN113238867A (zh) * 2021-05-19 2021-08-10 浙江凡双科技有限公司 一种基于网络卸载的联邦学习方法

Also Published As

Publication number Publication date
CN118214750A (zh) 2024-06-18

Similar Documents

Publication Publication Date Title
US20220247472A1 (en) Coding method, decoding method, and device
WO2022105913A1 (zh) 通信方法、装置及通信设备
WO2023040887A1 (zh) 信息上报方法、装置、终端及可读存储介质
WO2024125525A1 (zh) Ai算力上报方法、终端及网络侧设备
CN116017493A (zh) 模型请求方法、模型请求处理方法及相关设备
WO2024067281A1 (zh) Ai模型的处理方法、装置及通信设备
CN116488747A (zh) 信息交互方法、装置及通信设备
WO2024067280A1 (zh) 更新ai模型参数的方法、装置及通信设备
WO2024067434A1 (zh) Rs配置方法、装置、终端及网络侧设备
WO2024099091A1 (zh) 波束预测方法、装置、终端、网络侧设备及存储介质
WO2024099094A1 (zh) 波束测量方法、装置、终端、网络侧设备及存储介质
US20240224082A1 (en) Parameter selection method, parameter configuration method, terminal, and network side device
WO2023125951A1 (zh) 通信模型配置方法、装置和通信设备
WO2024032695A1 (zh) Csi预测处理方法、装置、通信设备及可读存储介质
WO2024120447A1 (zh) 模型监督触发方法、装置、ue、网络侧设备、可读存储介质及通信***
WO2024093713A1 (zh) 资源配置方法、装置、通信设备及可读存储介质
WO2024093771A1 (zh) 参考信号确定方法、终端及网络侧设备
WO2024140422A1 (zh) Ai单元的性能监测方法、终端及网络侧设备
WO2024067665A1 (zh) Csi预测处理方法、装置、通信设备及可读存储介质
WO2023186099A1 (zh) 信息反馈方法、装置及设备
WO2023040888A1 (zh) 数据传输方法及装置
WO2023103911A1 (zh) 测量方法、装置、设备及存储介质
WO2024008111A1 (zh) 数据采集方法及装置
WO2024083004A1 (zh) Ai模型配置方法、终端及网络侧设备
WO2024032694A1 (zh) Csi预测处理方法、装置、通信设备及可读存储介质