Model management method, communication node and storage medium

Abstract

The invention provides a model management method, a communication node and a storage medium. The method comprises the steps of receiving model management parameters or obtaining the model management parameters in a predefined mode; measuring the power overhead of the model according to the model management parameters; and reporting control information, wherein the control information comprises at least one of the following: the power overhead, the power overhead level and a request model operation instruction.

Description

Technical Field

[0001] This application relates to the field of mobile communication technologies, and for example, relates to a model management method, a communication node, and a storage medium. Background Art

[0002] In recent years, artificial intelligence (AI) / machine learning (ML) models have been introduced into mobile communication technologies. Training, inferring, and model-related functions for the models will inevitably bring new energy consumption. The hardware on which AI / ML models are based, such as a central processing unit (CPU), a graphics processing unit (GPU), and a neural processing unit (NPU), etc., requires an astonishing amount of energy for processing a large number of operations. For example, the thermal design power (TDP) of an NVIDIA H100 GPU is approximately 700 watts. When dealing with large models, a large number of GPUs are required, which will bring a huge amount of energy consumption. When an AI model is in a battery-powered device, since the battery capacity of the device is limited, model management and control must be performed on the operation of the AI model to maintain the battery life as long as possible.

[0003] Generally speaking, AI models will improve the performance and efficiency of mobile communication systems, but correspondingly, they also need to pay the price of the power consumption of the AI models themselves. Currently, there is no unified architecture that can manage all models and functions, and it is impossible to reduce the energy consumption of model operation in communication systems. It is even more difficult to balance the two conflicting goals of the power consumption cost and the improvement of mobile communication performance. Summary of the Invention

[0004] This application provides a model management method, a communication node, and a storage medium.

[0005] An embodiment of this application provides a model management method, which is applied to a first communication node and includes:

[0006] Receiving model management parameters or obtaining model management parameters in a predefined manner;

[0007] Measuring the power consumption of the model according to the model management parameters;

[0008] Reporting control information, where the control information includes at least one of the following: the power consumption, the power consumption level, and a request for a model operation instruction.

[0009] An embodiment of the present application provides a model management method, which is applied to a second communication node and includes:

[0010] Sending model management parameters, where the model management parameters are used to instruct a first communication node to measure the power overhead of a model;

[0011] Receiving control information reported by the first communication node;

[0012] The control information includes at least one of the following: the power overhead, the power overhead level, and a request model operation instruction.

[0013] An embodiment of the present application further provides a communication node, including: a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned model management method is implemented.

[0014] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the above-mentioned model management method is implemented. Description of the Drawings

[0015] Figure 1 A schematic diagram of an AI / ML framework provided for an embodiment;

[0016] Figure 2 A flowchart of a model management method provided for an embodiment;

[0017] Figure 3 A flowchart of another model management method provided for an embodiment;

[0018] Figure 4 A schematic diagram of managing the power overhead of a model provided for an embodiment;

[0019] Figure 5 A schematic diagram of another method for managing the power overhead of a model provided for an embodiment;

[0020] Figure 6 A schematic diagram of managing the discontinuous operation period of a model provided for an embodiment;

[0021] Figure 7 A schematic diagram of configuring the relationship between the power overhead level and model influencing factors provided for an embodiment;

[0022] Figure 8 A schematic diagram of a model determined by a terminal provided for an embodiment;

[0023] Figure 9 A schematic diagram of a model determined by a network device provided for an embodiment;

[0024] Figure 10 Schematic diagram for a network device to determine activation or deactivation of a model provided for an embodiment;

[0025] Figure 11 Schematic structural diagram of a model management device provided for an embodiment;

[0026] Figure 12 Schematic structural diagram of another model management device provided for an embodiment;

[0027] Figure 13 Schematic hardware structure diagram of a communication node provided for an embodiment. Detailed implementation manners

[0028] The present application will be described below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, rather than limiting the present application. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined arbitrarily with each other. Additionally, it should be noted that, for the sake of description, only the parts related to the present application rather than all the structures are shown in the drawings.

[0029] Artificial Intelligence (AI) includes devices, components, software, algorithms, or modules with self-learning capabilities such as Machine Learning (ML), deep learning, reinforcement learning, transfer learning, deep reinforcement learning, and meta-learning. In some cases, AI is implemented through an artificial intelligence network (or neural network). A neural network consists of multiple layers, with each layer including at least one node. Typically, a neural network includes an input layer, an output layer, and at least one hidden layer. Each layer of the neural network includes, but is not limited to, at least one of a fully connected layer, a dense layer, a convolutional layer, a transposed convolutional layer, a direct connection layer, an activation function, a normalization layer, and a pooling layer. In other cases, each layer of the neural network can include a sub-neural network, such as a Residual Network block (or Resnetblock), a Densenet Block, or a Recurrent Neural Network (RNN). An artificial intelligence network includes a neural network model and / or the neural network parameters corresponding to the neural network model. Here, the neural network model can be abbreviated as the network model, and the neural network parameters can be abbreviated as network parameters. A network model defines the architecture of the neural network, including the number of layers, the size of each layer, the activation function, the connection situation, the convolution kernel and its size, the convolution stride, and the convolution type (such as 1D convolution, 2D convolution, 3D convolution, dilated convolution, transposed convolution, separable convolution, grouped convolution, or extended convolution, etc.). The network parameters are the weights and / or biases of each layer in the network model and their values. A network model can correspond to multiple different sets of neural network parameter values to adapt to different scenarios. The values of the network parameters can be obtained through offline training and / or online training. For example, by inputting at least one sample and label, training the neural network model to obtain the neural network parameters. A neural network model can correspond to multiple different neural network parameter values.

[0030] AI / ML (Artificial Intelligence / Machine Learning) is a promising enhancement direction for mobile communication systems. Introducing AI / ML technologies into mobile communication systems, such as 5G (Fifth Generation), 5G-A (5G-Advanced), and 6G (Sixth Generation), can improve the system's operating efficiency. For example, by reducing the overhead of reference signals through AI / ML inference and prediction, reducing the overhead of channel state information feedback, or improving the accuracy of terminal positioning, etc.

[0031] For a communication system employing AI / ML technology, in all parts of this application, "model" is a general term used to describe the capabilities of a device in a mobile system to perform processing methods, functions, features, or groups of features. A "model" can be understood as a function, a functional module, a processing method, an information processing method, an implementation method, a group of functions, a configuration, a set of configurations, a data set (e.g., for model training), or a data-driven algorithm, etc.

[0032] Figure 1 Schematic diagram of an AI / ML framework provided for an embodiment. As Figure 1 shown, the data collection functional module can provide input data for model training, management, and inference functions. The input data required for the model training function is training data (also known as training data). The input data required for managing the model or function is monitoring data. The input data required for the AI / ML inference function is inference data. The collected monitoring data may provide a calibration-like function for the label data and is fed into the model management functional module for comparison with the inference output to output corresponding rulings or management commands.

[0033] The model training functional module has the functions of performing model training, verification, and testing.

[0034] The model management functional module can perform the following functions: (1) Indicate related operations of the model, such as selection / activation / deactivation / switching / rollback, etc. for the model. (2) Conduct model performance monitoring. (3) Make decisions or indications to ensure correct inference operations based on the data received from the data collection function and the inference function. (4) This functional module is also responsible for requesting a model from the model storage function for model transfer / delivery requests. (5) Input performance feedback / retraining requests to the model training functional module for model (re)training or updating.

[0035] The model inference functional module takes the data provided by the data collection function as input and provides the output from the applied model or function. The inference output is the final output of the entire AI / ML model or function and is used by other units in the mobile system. The output data can also be input into the management functional module for internally monitoring the performance of the AI / ML model or the AI / ML function.

[0036] Model storage is a function responsible for storing trained / updated models that can be used to perform inference functions.

[0037] All the functional modules related to the models mentioned in the above framework belong to the processing process, and the processing processes mentioned in the embodiments of this application can all be implemented in the form of models and the above framework.

[0038] The model can improve the performance and efficiency of the mobile communication system, but correspondingly, it also needs to pay the cost of the model's own power overhead. Therefore, it is necessary to balance between the cost of power overhead and the improvement of mobile communication performance. In the embodiments of the present application, a unified architecture is defined to manage all models and function implementations to better control its power overhead. This architecture can be implemented through the model management function module to uniformly control the power overhead of the model and functions.

[0039] Figure 2 FIG. 4 is a flowchart of a model management method provided for an embodiment. This method can be applied to the first communication node, and the first communication node can be the node where the model is located, the managed node, or the node whose power overhead is controlled. As Figure 2 shown, the method provided in this embodiment includes the following steps:

[0040] Step 110, receiving model management parameters or obtaining model management parameters in a predefined manner.

[0041] Step 120, measuring the power overhead of the model according to the model management parameters.

[0042] Step 130, reporting control information, where the control information includes at least one of the following: the power overhead, the power overhead level, and the request model operation instruction.

[0043] In this embodiment, the model mainly refers to the AI model. The model management parameters can be understood as the parameters used to manage the model. For example, the parameters indicating the related operations of the model, including the selection, activation, deactivation, switching, and / or fallback of the model, etc.; for example, the parameters for monitoring the model performance, including the power overhead, the model status, and / or the model parameters, etc.; for example, the decisions or instructions made for the model, such as whether to measure or report control information, whether to train and / or whether to update the model, etc. The model management parameters can be obtained by a defined method (for example, they are predefined or default in the system), or can be indicated or configured by the second communication node. The first communication node reports control information to the second communication node based on the measurement of the power overhead of the model, and the first communication node uniformly controls the power overhead of the model according to the control information. Among them, the power overhead can also be referred to as power consumption or power expenditure, etc. In one embodiment, the model management parameters include at least one of the following: the measurement period, the measurement start point, the measurement duration, and the reporting format.

[0044] In one embodiment, the reporting format includes at least one of the following:

[0045] The reported measured value is at least power or work;

[0046] The reported measurement values are quantized at equal or unequal intervals between the upper and lower bounds of the measurement, and the upper and lower bounds of the measurement are obtained by predefined means.

[0047] In one embodiment, reporting control information includes one of the following:

[0048] Reporting control information periodically;

[0049] Reporting control information upon receiving a reporting indication.

[0050] In one embodiment, the method further includes:

[0051] Reporting auxiliary information for model operation, where the auxiliary information includes at least one of the following: time information and level information of model operation;

[0052] Wherein, the time information includes at least one of the following: period, duration, start point of activation time.

[0053] In one embodiment, the method further includes:

[0054] Receiving an activation indication message or a deactivation indication message;

[0055] Upon receiving the activation indication message, running the model;

[0056] Upon receiving the deactivation indication message, stopping the operation of the model.

[0057] In one embodiment, the requested model operation instruction includes at least one of the following instructions: request for model activation, request for model deactivation, request for model configuration instruction, request for model transmission, request for configuring power overhead level.

[0058] In one embodiment, the requested model operation instruction includes at least one of the following: index of the model, related parameters associated with the model index.

[0059] In one embodiment, the method further includes:

[0060] Controlling the operation of the model to alternate between an active state and a dormant state according to the indication information of the discontinuous operation period.

[0061] In one embodiment, the method further includes:

[0062] Receiving the indication information of the discontinuous operation period through at least one of radio resource control, Medium Access Control (MAC) layer signaling, and Downlink Control Information (DCI);

[0063] The indication information of the discontinuous operation period is configured or indicated semi-statically or dynamically by the second communication node based on the reported power overhead.

[0064] In one embodiment, the indication information of the discontinuous operation period includes at least one of the following:

[0065] The length of the discontinuous operation period, the starting point of the discontinuous operation period, the length of the active time, the starting point of the active time, the length of the dormant time, the starting point of the dormant time.

[0066] In one embodiment, the indication information of the discontinuous operation period includes the indication information corresponding to one or more discontinuous operation periods.

[0067] In one embodiment, the indication information of the discontinuous operation period is bound to the operating state of the first communication node.

[0068] In one embodiment, the indication information of the discontinuous operation period is bound to the energy-saving mode of the first communication node.

[0069] In one embodiment, the discontinuous operation period is consistent with or matches the activation period of the energy-saving mode; or,

[0070] The duration of the activation period within the discontinuous operation period is consistent with or matches the activation period of the energy-saving mode;

[0071] Wherein, matching the activation period of the energy-saving mode includes:

[0072] Being less than or equal to the activation period of the energy-saving mode.

[0073] In one embodiment, the method further includes:

[0074] Obtaining the relationship between the control information and the influencing factors of the model;

[0075] The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, number of model parameters, splitting point of the bilateral model.

[0076] In one embodiment, the relationship between the control information and the influencing factors of the model is determined by the second communication node or obtained according to predefined information.

[0077] In one embodiment, the method further includes:

[0078] Receiving the updated model, where the updated model is determined by the second communication node according to the control information.

[0079] In one embodiment, the method further includes:

[0080] Determine the model according to the power overhead level configured by the second communication node.

[0081] In one embodiment, the relationship between the control information and the influencing factors of the model is adaptively adjusted by the second communication node according to the reported control information and / or the adjustment information of the model.

[0082] In one embodiment, the method further includes:

[0083] Receive random access resource configuration information, where the partition of the random access resource is associated with the control information of the model;

[0084] Select a random access resource and transmit a random access signal according to the random access resource configuration information.

[0085] Figure 3 A flowchart of a model management method provided for an embodiment. This method can be applied to the second communication node, and the second communication node can be a node for managing the model or a node for controlling power overhead. As Figure 3 shown, the method provided in this embodiment includes step 210 and step 220.

[0086] In step 210, send model management parameters, where the model management parameters are used to instruct the first communication node to measure the power overhead of the model.

[0087] In step 220, receive the control information reported by the first communication node. The control information includes at least one of the following: the power overhead, the power overhead level, and a request for a model operation instruction.

[0088] In one embodiment, the model management parameters include at least one of the following: measurement period, measurement start point, measurement duration, and reporting format.

[0089] In one embodiment, the reporting format includes at least one of the following:

[0090] The reported measurement value is at least power or work;

[0091] The reported measurement value is quantized at equal intervals or unequal intervals between the upper bound value and the lower bound value of the measurement, and the upper bound value and the lower bound value of the measurement are obtained by predefined.

[0092] In one embodiment, receiving the control information includes one of the following:

[0093] Receive the control information periodically;

[0094] Receive the control information when a reporting indication is sent.

[0095] In one embodiment, the method further includes:

[0096] Receive auxiliary information for the operation of the model, where the auxiliary information includes at least one of the following: time information of model operation, level information;

[0097] Among them, the time information includes at least one of the following: period, duration, activation time start point

[0098] In one embodiment, the method further includes:

[0099] Send an activation indication message or a deactivation indication message;

[0100] The activation indication message is used to instruct the first communication node to run the artificial intelligence model;

[0101] The deactivation indication message is used to instruct the first communication node to stop running the artificial intelligence model.

[0102] In one embodiment, the request model operation instruction includes at least one of the following instructions: request model activation, request model deactivation, request model configuration instruction, request model transmission, request configuration of power overhead level.

[0103] In one embodiment, the method further includes:

[0104] Determine the indication information of the discontinuous operation period, where the indication information of the discontinuous operation period is used to indicate the time when the operation of the artificial intelligence model is in the active state and the dormant state.

[0105] In one embodiment, the method further includes:

[0106] Configure or indicate the indication information of the discontinuous operation period according to the power overhead semi-statically or dynamically;

[0107] The indication information of the discontinuous operation period is sent through at least one of radio resource control, MAC layer signaling, and DCI signaling.

[0108] In one embodiment, the method further includes:

[0109] Determine the relationship between the control information and the influencing factors of the model;

[0110] The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, number of model parameters, segmentation point of the bilateral model.

[0111] In one embodiment, the relationship between the control information and the influencing factors of the model is determined by the second communication node or obtained according to predefined information.

[0112] In one embodiment, the method further includes:

[0113] Update the model according to the control information;

[0114] Send the updated model to the first communication node.

[0115] In one embodiment, the method further includes:

[0116] Configure and send a power overhead level;

[0117] The power overhead level is used to assist the first communication node in determining a model according to the power overhead level.

[0118] In one embodiment, the method further includes:

[0119] Adaptively adjust the relationship between the control information and the influencing factors of the model according to the reported control information and / or the adjustment information of the model.

[0120] The following uses some embodiments to exemplarily illustrate the model management method of the present application.

[0121] Embodiment 1

[0122] This embodiment mainly describes the process of power overhead measurement and reporting of the model. If power overhead control of the model is required, first, it is necessary to understand the real-time power overhead during the operation of the model in the device, and the device needs to perform power overhead measurement and reporting.

[0123] Assume that the control entity of the power overhead is located in the model management module, the model management module is located in the first communication node (such as a network device), and the model is located in the second communication node (such as a terminal). In this embodiment, the signaling for power overhead control and management between the network device and the terminal is defined.

[0124] For example, the model management module is located in a network device (such as a core network, a radio access network, a base station, or a network management device, etc.), and the model is located in the terminal. In this embodiment, the relevant signaling between the network device and the terminal is defined.

[0125] For example, the model management module is located in a network device (such as a core network, a radio access network, or a network management device, etc.), and the model is located in the base station. This embodiment can be extended to support this scenario.

[0126] Figure 4 A schematic diagram for managing the power overhead of a model provided for an embodiment. As Figure 4As shown, the terminal can periodically measure the power overhead during model operation and report it periodically. Before measurement, model management parameters such as the configured power overhead measurement period, measurement start point, and / or reporting format can be obtained from the network device. The periodicity can be configured separately according to different states of the terminal, such as being in the Radio Resource Control (RRC) connected state, inactive state, or idle state, or according to other measurement periods of the terminal. Multiple measurements can be concentrated in the same time period as much as possible to avoid unnecessary power overhead caused by scattered measurements.

[0127] Figure 5 FIG. is a schematic diagram of another method for managing the power overhead of a model provided in an embodiment. As Figure 5 shown, the network device can also request the measurement results of the power overhead from the terminal device aperiodically. Before the request or along with the request signaling, model management parameters such as the measurement duration, measurement start point, or reporting format of the power overhead can be configured. The above model management parameters can also be auxiliary information for model management. After the terminal obtains these model management parameters, the network device can request or instruct the terminal to report the power overhead. After measuring the power overhead, the terminal reports the power overhead aperiodically.

[0128] In some embodiments, periodic measurement reporting and aperiodic measurement reporting can be used in combination. For example, when the period is set relatively long, aperiodic measurement reporting can be initiated within a certain period to make up for the problem of excessive measurement delay in periodic reporting.

[0129] After obtaining the power overhead, the network device can decide on the next action based on the power overhead measurement results.

[0130] In this embodiment, the model management parameters include information related to assisting power overhead measurement. For example, the measurement period, measurement duration, measurement start point, or reporting format can be defined as follows and the device where the power overhead measurement is located can be indicated through RRC signaling or MAC layer signaling.

[0131] Measurement period: Periodic measurement reporting can perform periodic measurements following the configuration of the measurement period. The measurement period can be set in units of superframe, frame, subframe, time slot, and / or symbol.

[0132] Measurement duration: The measurement duration configures the actual measurement time length of the power overhead measurement. In the case of periodic measurement reporting, the measurement duration should not exceed the measurement period. When aperiodic measurement reporting is performed, the measurement duration should not exceed the next measurement start point. The measurement duration can be set in units of superframe, frame, subframe, time slot, and / or symbol. The longer the measurement duration, the smoother and more accurate the results for long-term measurement, but the greater the power consumption for measurement. Therefore, a balance needs to be considered between the measurement duration and measurement accuracy.

[0133] Measurement start point: It configures the specific start position of the measurement, which is required to be configured in both periodic measurement reports and aperiodic measurement reports. When the measurement report is periodic, the measurement start point can be defined as the relative time offset with respect to the superframe header, frame header, subframe header, time slot header, and / or the first symbol. When the measurement report is aperiodic, the measurement start point is configured with reference to the signaling indicating the power overhead report. For example, if the signaling indicating the power overhead report is DCI, the measurement start point can be delayed by a time length of ΔT with respect to this DCI signaling. ΔT can be the length of one or more time slots, or the length of one or more symbols, or a time deviation obtained by superimposing the lengths of one or more time slots and one or more symbols.

[0134] Reporting format: The basic measurement quantity for the power overhead report can be the energy consumption per unit time, that is, the power value, with the unit of watt (W). Or, when the measurement duration is determined, the energy consumption, that is, the work, can also be reported, with the unit of joule or kilowatt-hour or other equivalent units. Due to the limitation of the measurement report overhead, the power overhead report should be quantized before reporting. The reported measurement value is quantized at equal intervals or unequal intervals between the upper bound value and the lower bound value of the measurement. The upper bound value and the lower bound value are obtained through predefined means. For example, the power report can be quantized into 256 levels and represented by 8 bits in binary. 00000000 indicates that the power is lower than X watts, 11111111 indicates that the power is higher than Y watts, and the power difference between X and Y watts is quantized at equal intervals or unequal intervals by the binary indices in between. This quantization method can be expressed in the form of a table and stored in the network model and the terminal through predefined means. When the power overhead levels are agreed upon through signaling interaction between the network device and the terminal, the measurement report can also directly report the power overhead levels.

[0135] The above-mentioned model management parameters including the measurement period, measurement duration, measurement start point, and / or reporting format can also be determined or obtained through predefined means.

[0136] When the power overhead control unit is located in the core network and the model is located in the base station, a similar signaling process can also be used between the core network and the base station to report the power overhead.

[0137] Whether it is periodic power overhead measurement reporting or aperiodic power overhead measurement reporting, the effective time of the measurement does not necessarily exactly match the activation or running time of the model. Therefore, in order to better adapt the measurement time period of the power overhead to the activation / running time period of the model, the device where the model is located can report auxiliary information on model activation or running, and the reported content includes the time information and / or level information of model running; among them, the time information includes but is not limited to the period, duration, activation time start point, and / or activation running level, etc. of model activation or running; among them, the level information, such as the activation running level, its related different levels define the possible high and low power overheads during model running. For example, the power overhead during model training is the highest, the power overhead during model inference is the second highest, and the power overhead when the model only processes basic management functions is the lowest. The device can report the time information, activation running level, etc. during model training alone, or can also report the time information, activation running level, etc. during model inference alone. The auxiliary information reported by the device where the model is located on model activation or running can be managed by the model management device, such as the network device configuring periodic reporting, or can also be indicated by the network device for the device where the model is located (such as the terminal) to report aperiodically.

[0138] Embodiment 2

[0139] This embodiment mainly describes the activation or deactivation process of the model. Assume that the control entity for model activation or deactivation is located in the model management module, the model management module is located in the first communication node (such as a network device), and the model is located in the second communication node (such as a terminal).

[0140] Based on the power overhead measurement reporting results reported by the device where the model is located, the network device can comprehensively judge the balance between the performance gain obtained by the model and its own power consumption, and decide whether to activate the model, deactivate the model, or make the model enter the non-activated state.

[0141] For the case where there is a certain periodic time pattern for the activation time or deactivation time of the model, the network device can also set a discontinuous processing period for the device where the model is located. During this period, the model's operation alternates between an active state and a dormant state. During the on-duration, the model's operation can be enabled, and during the off-duration, the model's operation can be turned off to save energy until the next active time window.

[0142] Figure 6 It is a schematic diagram for managing the discontinuous operation period of a model provided in an embodiment. As Figure 6As shown, the specific duration or timing of the DP period can be semi-statically or dynamically configured by the network device based on the power consumption reporting result of the device where the model is located, and notified to the terminal through RRC signaling, MAC layer signaling, or DCI signaling. These specific parameters may involve the length of the DP period, the definition of the period start point, the length and / or start point definition of the active time and the sleep time, etc.

[0143] The DP period can be set in units of superframe, frame, subframe, time slot, and / or symbol.

[0144] Measurement duration: The measurement duration indicates the actual measurement time length of the power consumption measurement. When reporting periodic measurements, the measurement duration should not exceed the measurement period; when reporting aperiodic measurements, the measurement duration should not exceed the next measurement start point. The measurement duration can be set in units of superframe, frame, subframe, time slot, and / or symbol. The longer the measurement duration, the smoother and more accurate the results for long-term measurements, but the greater the power consumption for measurement. Therefore, a balance needs to be considered between the measurement duration and the measurement accuracy.

[0145] The start point of the DP period can be defined as the relative time offset with respect to the superframe header, frame header, subframe header, time slot header, and / or the first symbol.

[0146] The active time or the sleep time can be defined simultaneously, or only one of them can be defined, and the other can be inferred in combination with the period definition. When defining the active time, its length unit and start point can be set in units of superframe, frame, subframe, time slot, and / or symbol.

[0147] The network device can configure multiple sets of DP periods, including at least one long DP period and one short DP period, to adapt to the requirements of the service.

[0148] The setting and configuration of the DP period can also be aligned or bound with the operating state of the device. For example, different DP period parameters can be configured for the RRC connection state, inactive state, and / or idle state of the terminal device respectively. For example, the DP period is shorter in the connected state, and the DP period gradually becomes longer in the inactive state and the idle state.

[0149] The setting and configuration of the DP period can also be aligned or bound with other energy-saving modes of the device. For example, when there is a strong binding relationship between the activation period of the model operation and the energy-saving mode of the terminal, the DP period can also be made consistent or matched with the activation period of the energy-saving mode, and the duration of the activation period within the DP period can also be made consistent or matched with the activation period of the energy-saving mode. The matching method can be that the DP period is less than or equal to the transmission period, or the duration of the activation period is less than or equal to the transmission time.

[0150] The network device decides whether to activate or deactivate the model, in addition to the power consumption measurement report results reported by the device where the model is located, the device where the model is located can also directly send a request model operation instruction to the network device to request activation or deactivation of the model or function. For example, as the device where the model is located, after the terminal performs power consumption measurement, when it finds that its own model power consumption exceeds the threshold or expectation, it can directly request the network device to configure a deactivation instruction; when it finds that the model needs to improve performance, it can also directly request the network device to configure an activation instruction.

[0151] Table 1 is an organizational structure table of activation or deactivation request model operation instructions. As shown in Table 1, the terminal, as the device where the model is located, may maintain multiple types of models at the same time, such as a model for beam management, a model for channel prediction compression, a model for positioning functions, etc. The activation or deactivation request sent by the terminal may be for a specific model, a group of models, or all models; even models with the same function, such as beam management models, may be in different states, such as training state, inference state, or low power maintenance state, etc. The activation or deactivation request sent by the terminal can be a targeted request for different states of the model, for example, the training state can be deactivated, but the inference state can be activated.

[0152] Table 1 An organizational structure table of activation or deactivation request model operation instructions

[0153]

[0154] Example 3

[0155] This embodiment mainly describes the power consumption level of the management model. Assume that the control subject of model activation or deactivation is located in the model management module, the model management module is located in the first communication node (such as a network device), and the model is located in the second communication node (such as a terminal).

[0156] Referring to Example 2, in order to reduce the energy or power consumption of the model, the deactivation method can be adopted, but after deactivation, it is difficult to obtain the performance gain that can be obtained by running the model. Therefore, between the two extreme states of activation or deactivation, some appropriate compromises and balances can be made, by setting a certain level for the power consumption, and selecting a suitable model for the device according to the different configurations of the model corresponding to different power consumption levels. The power consumption mentioned in this facility example is equivalent to the power consumption.

[0157] In this embodiment, the energy consumption influencing factors of the model may depend on one or more of the following aspects:

[0158] 1. Model size and complexity, including:

[0159] Number of parameters: The more parameters a model has, the higher its complexity and the more computational resources are required to process it. Large models such as Transformer and BERT consume more power due to their huge number of parameters and deep network structures.

[0160] Depth: The depth of the network directly affects the learning ability and computational requirements of the model. The more layers there are, the more forward and backward propagation calculations are needed, and correspondingly, the higher the energy consumption.

[0161] 2. Training algorithms, including:

[0162] Optimization algorithms: The optimization algorithms used (such as Adam or Stochastic Gradient Descent (SGD)) affect the convergence speed and stability, indirectly affecting the energy consumption during the training process. Some algorithms may require more iterations to converge, increasing the overall energy consumption.

[0163] Batch size: The batch size during training affects GPU utilization and memory requirements. Larger batch sizes can improve the efficiency of parallel processing but may also increase memory pressure and energy consumption.

[0164] 3. Number of training times and iterations, including:

[0165] Number of epochs: The more epochs the model is trained for, the greater the overall computational amount required to complete the training, and correspondingly, the higher the energy consumption.

[0166] Early stopping technique: Using the early stopping technique can stop the training when the validation loss no longer improves, thus reducing unnecessary calculations and energy consumption.

[0167] 4. Hardware efficiency, including:

[0168] Processor type: The type of hardware used (such as CPU or GPU, etc.) greatly affects energy efficiency. GPUs are generally more energy-efficient than CPUs when performing large-scale matrix calculations in parallel.

[0169] Hardware utilization: Insufficient utilization of hardware resources will lead to energy waste. Optimizing hardware utilization can reduce energy consumption during idle times.

[0170] 5. Model deployment and inference, including:

[0171] Quantization: By quantizing the model, converting floating-point parameters to a low-precision (such as int8) format, the model size and runtime energy consumption can be reduced.

[0172] Pruning: Removing redundant connections in the network (weight pruning) can reduce model complexity and energy consumption during the inference stage.

[0173] Distillation: By training a smaller model (student model) to mimic the behavior of a larger model (teacher model), it is possible to reduce the energy consumption during the inference process without significantly reducing the accuracy.

[0174] 6. Software and tool optimization, including:

[0175] Software framework: The software framework (such as TensorFlow, PyTorch) used for model implementation also affects the running efficiency and energy consumption. Different frameworks may vary in memory management and computational optimization.

[0176] Compilation optimization: Through advanced compiler optimization techniques, such as graph optimization and operation fusion, it is possible to improve the running efficiency and reduce the energy consumption.

[0177] Specific to the model used in a mobile communication system or terminal, under the same conditions of hardware, software algorithms, implementation tools, etc., the main influencing factors controllable by the communication system for the model power overhead come from the model type, accuracy (quantization bits), model depth and width, model parameter quantity, split point of the bilateral model, etc. If there are differences in the hardware conditions for implementing the model, the model power consumption is related to the hardware structure, type, channel size of data transmission, etc. In the model power management mechanism, it is necessary to flexibly handle the above factors related to the model itself, software and hardware, and / or tools, so as to be able to change, adjust, and optimize the model configuration under the premise of power overhead constraints.

[0178] Figure 7 FIG. is a schematic diagram showing the relationship between the configured power overhead level and model influencing factors provided for an embodiment. As Figure 7 shown, the network device can establish the association relationship or mapping relationship between the power overhead level and various influencing factors of the model. These levels and the relationship between each level and various influencing factors of the model can be predefined and stored in the network device and the terminal. Or when the network device sets the level and relationship, it can be configured to the terminal through signaling. The corresponding level and relationship can be terminal-specific or shared by multiple terminals.

[0179] It should be noted that other control information, such as power overhead and / or request model operation instructions, etc., can also have an association relationship with the model influencing factors.

[0180] Table 2 is a table showing the association relationship between the power overhead level and model influencing factors in an embodiment. Taking the case where the power overhead increases from level 1 to level 3 as an example.

[0181] Table 2 Relationship between power overhead level and model influencing factors

[0182]

[0183] Among them, MLP refers to Multi-Layer Perceptron; CNN refers to Convolutional Neural Network.

[0184] Figure 8 A schematic diagram of determining a model by a terminal provided for an embodiment. As Figure 8 shown, when the power overhead control and management center determines a certain power overhead level applicable to the model in real time, the network device can configure or indicate the power overhead level to the device where the model is located, such as the terminal. The terminal determines the set structures or parameters for executing the model function according to the mapping relationship between each level and various influencing factors of the model that has been pre-configured. Further, the terminal can first request to configure the power overhead level, and then the network device configures or indicates the power overhead level to the terminal, and the subsequent operations of the terminal are similar to the foregoing.

[0185] Figure 9 A schematic diagram of determining a model by a network device provided for an embodiment. As Figure 9 shown, when the power overhead control and management center determines a certain power overhead level applicable to the model in real time, the network device can also select a suitable model according to the determined power overhead level and transmit the model to the device where the model is located. The transmission forms of the specific model are diverse, such as transmitting the index of the model, transmitting the data set or data set index for model training, transmitting the structure and parameters of the model, etc.

[0186] The device where the model is located can determine the power overhead level according to the measurement result of its own power overhead and the mapping relationship between the power overhead and the power overhead level, and report it to the device where the model management is located. The mapping relationship between the power overhead and the power overhead level is optionally configured by the device managing the model to the device where the model is located. The device where the model management is located further adjusts the expected power overhead level according to the reported power overhead level, and determines the model according to the expected power overhead level. After the model is determined, the model is transmitted to the device where the model is located.

[0187] Figure 10 A schematic diagram of determining to activate or deactivate a model by a network device provided for an embodiment. As Figure 10 shown, the device managing the model can also directly decide whether to activate or deactivate the model.

[0188] The device where the model is located can also, according to the measurement result of its own power overhead, directly request the network device to configure a certain type or level of model, and the requested model can reduce or enhance the power overhead to achieve the purpose of adapting to energy saving or improving performance. The said request can include the index of the model for which configuration is desired or the relevant parameters or descriptions associated with the model index.

[0189] The establishment of the power overhead level and the mapping relationship between the power overhead level and the model can be initially configured by the network administrator. However, the initial configuration only considers the current status of the model at the time of configuration and does not take into account the further evolution of the model and the long-term optimization of the system terminal in terms of power overhead. Therefore, during the long-term operation of the model, it is still necessary for the network device to adaptively adjust the power overhead level or the mapping relationship between the power overhead level and the influencing factors of the model according to the reported power overhead measurement or the adjustment information of the model structure / parameters; this adaptive adjustment process can be intervened by humans or completed by artificial intelligence management.

[0190] The power overhead level of the model is basically determined after the model is determined. Devices where the model is located, such as terminals, can map the power overhead level to random access resources, that is, partition the random access resources and associate them with the power overhead level. The terminal selects the corresponding random access resources according to the power overhead level of the model during the random access process, and can implicitly notify the network device where the model management function is located of the power overhead level of the configured model, so as to determine subsequent access parameters such as bandwidth, modulation method, waveform, Multiple-In Multiple-Out (MIMO) layer number, Bandwidth Part (BWP), carrier, and reference signal configuration to adapt to the model.

[0191] The embodiment of the present application also provides a model management device. Figure 11 A structural schematic diagram of a model management device provided for an embodiment is as follows. Figure 11 As shown, the model management device includes:

[0192] A parameter acquisition module 310, configured to receive model management parameters or obtain model management parameters in a predefined manner;

[0193] A measurement module 320, configured to measure the power overhead of the model according to the model management parameters;

[0194] An information reporting module 330, configured to report control information, where the control information includes at least one of the following: the power overhead, the power overhead level, and a request for a model operation instruction.

[0195] In one embodiment, the model management parameters include at least one of the following:

[0196] The measurement period, the measurement start point, the measurement duration, and the reporting format.

[0197] In one embodiment, the reporting format includes at least one of the following:

[0198] The reported measured value is at least power or work;

[0199] The reported measurement values are quantized at equal or unequal intervals between the upper and lower bounds of the measurement, and the upper and lower bounds of the measurement are obtained by predefined means.

[0200] In one embodiment, the reported control information includes one of the following:

[0201] Periodic reporting of control information;

[0202] Reporting control information upon receipt of a reporting indication.

[0203] In one embodiment, the reporting module is further configured to: report auxiliary information on the operation of the model, where the auxiliary information includes at least one of the following: time information and level information on the operation of the model;

[0204] Wherein, the time information includes at least one of the following: period, duration, start point of activation time.

[0205] In one embodiment, the device further includes:

[0206] A receiving module configured to receive activation indication information or deactivation indication information;

[0207] An operating module configured to operate the model upon receipt of activation indication information;

[0208] A stopping module configured to stop the operation of the model upon receipt of deactivation indication information.

[0209] In one embodiment, the requested model operation instruction includes at least one of the following instructions: request for model activation, request for model deactivation, request for model configuration instruction, request for model transmission, request for power overhead level configuration.

[0210] In one embodiment, the requested model operation instruction includes at least one of the following: index of the model, relevant parameters associated with the index of the associated model.

[0211] In one embodiment, the device further includes:

[0212] A control module configured to control the operation of the model to alternate between an active state and a dormant state according to the indication information of the discontinuous operation period.

[0213] In one embodiment, the indication information of the discontinuous operation period is bound to the energy-saving mode of the first communication node.

[0214] In one embodiment, the discontinuous operation period is consistent with or matches the activation period of the energy-saving mode; or,

[0215] The duration of the activation period within the discontinuous operation period is the same as or matches the activation period of the energy-saving mode;

[0216] Among them, matching the activation period of the energy-saving mode includes:

[0217] It is less than or equal to the activation period of the energy-saving mode.

[0218] In one embodiment, the device further includes:

[0219] A relationship acquisition module, configured to acquire the relationship between control information and influencing factors of the model;

[0220] The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, number of model parameters, splitting point of the bilateral model.

[0221] In one embodiment, the relationship between the control information and the influencing factors of the model is determined by a second communication node or obtained according to predefined information.

[0222] In one embodiment, the device further includes:

[0223] A model receiving module, configured to receive the updated model, and the updated model is determined by the second communication node according to the control information.

[0224] In one embodiment, the device further includes:

[0225] A model determination module, configured to determine a model according to the power overhead level configured by the second communication node.

[0226] In one embodiment, the relationship between the control information and the influencing factors of the model is adaptively adjusted by the second communication node according to the reported control information and / or adjustment information of the model.

[0227] In one embodiment, the device further includes:

[0228] A random access module, configured to receive random access resource configuration information, where the partition of the random access resource is associated with the control information of the model; according to the random access resource configuration information, select a random access resource and transmit a random access signal.

[0229] The model management device proposed in this embodiment and the model management method proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to any of the above embodiments, and this embodiment has the same beneficial effects as the execution of the model management method.

[0230] An embodiment of the present application further provides a model management device. Figure 12Schematic diagram of the structure of a model management device provided for an embodiment. As Figure 12 shown, the model management device includes:

[0231] A parameter sending module 410, configured to send model management parameters, where the model management parameters are used to instruct a first communication node to measure the power overhead of a model;

[0232] An information receiving module 420, configured to receive control information reported by the first communication node;

[0233] The control information includes at least one of the following: the power overhead, the power overhead level, and a request model operation instruction.

[0234] In one embodiment, the model management parameters include at least one of the following:

[0235] Measurement period, measurement starting point, measurement duration, reporting format.

[0236] In one embodiment, the reporting format includes at least one of the following:

[0237] The reported measurement value is at least power or work;

[0238] The reported measurement value is quantized at equal intervals or unequal intervals between the upper bound value and the lower bound value of the measurement, and the upper bound value and the lower bound value of the measurement are obtained by predefined.

[0239] In one embodiment, receiving the control information includes one of the following:

[0240] Receiving the control information periodically;

[0241] Receiving the control information in the case of sending a reporting indication.

[0242] In one embodiment, the information receiving module 420 is further configured to:

[0243] Receive auxiliary information of model operation, where the auxiliary information includes at least one of the following: time information of model operation, level information; wherein, the time information includes at least one of the following: period, duration, activation time starting point.

[0244] In one embodiment, the device further includes:

[0245] An activation module, configured to send activation indication information or deactivation indication information;

[0246] The activation indication information is used to instruct the first communication node to run the artificial intelligence model;

[0247] The deactivation indication information is used to instruct the first communication node to stop running the artificial intelligence model.

[0248] In one embodiment, the request model operation instruction includes at least one of the following instructions: request model activation, request model deactivation, request model configuration instruction, request model transmission, and request power overhead level configuration.

[0249] In one embodiment, the device further includes:

[0250] A period determination module, configured to determine indication information of a discontinuous operation period, where the indication information of the discontinuous operation period is used to indicate the time when the operation of the artificial intelligence model is in an active state and a dormant state.

[0251] In one embodiment, the device further includes:

[0252] A period indication module, configured to configure or indicate the indication information of the discontinuous operation period semi-statically or dynamically according to the power overhead;

[0253] The indication information of the discontinuous operation period is sent through at least one of radio resource control, MAC layer signaling, and DCI signaling.

[0254] In one embodiment, the device further includes:

[0255] A relationship determination module, configured to determine the relationship between the control information and the influencing factors of the model;

[0256] The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, number of model parameters, and splitting point of the bilateral model.

[0257] In one embodiment, the relationship between the control information and the influencing factors of the model is determined by a second communication node or obtained according to predefined information.

[0258] In one embodiment, the device further includes:

[0259] An update module, configured to update the model according to the control information; and send the updated model to the first communication node.

[0260] In one embodiment, the device further includes:

[0261] A level configuration module, configured to configure and send a power overhead level; the power overhead level is used to assist the first communication node in determining the model according to the power overhead level.

[0262] In one embodiment, the device further includes:

[0263] An adjustment module, configured to adaptively adjust the relationship between the control information and the influencing factors of the model according to the reported control information and / or adjustment information of the model.

[0264] In one embodiment, the apparatus further includes:

[0265] A determination module, configured to determine random access resource configuration information, wherein the partition of the random access resource is associated with the control information of the model.

[0266] The model management apparatus proposed in this embodiment and the model management method proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to in any of the above embodiments, and this embodiment has the same beneficial effects as the execution of the model management method.

[0267] An embodiment of the present application further provides a communication node. Figure 13 As shown in the schematic diagram of the hardware structure of a communication node provided for an embodiment, Figure 13 the communication node provided by the present application includes a processor 510 and a memory 520; the processor 510 in the communication node can be one or more, Figure 13 taking one processor 510 as an example; the memory 520 is configured to store one or more programs; the one or more programs are executed by the one or more processors 510, so that the one or more processors 510 implement the model management method as described in the embodiments of the present application.

[0268] The communication node further includes: a communication device 530, an input device 540, and an output device 550.

[0269] The processor 510, memory 520, communication device 530, input device 540, and output device 550 in the communication node can be connected by a bus or other means, Figure 13 taking connection by bus as an example.

[0270] The input device 540 can be used to receive input digital or character information, and generate key signal inputs related to the user settings and function control of the communication node. The output device 550 may include a display device such as a display screen.

[0271] The communication device 530 may include a receiver and a transmitter. The communication device 530 is configured to perform information transceiver communication according to the control of the processor 510.

[0272] The memory 520, being a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the model management method described in the embodiments of the present application (for example, the parameter acquisition module 310, the measurement module 320, and the information reporting module 330 in the model management device). The memory 520 can include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the communication node, etc. In addition, the memory 520 can include high-speed random access memory and can also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some instances, the memory 520 can further include a memory remotely disposed relative to the processor 510, and these remote memories can be connected to the communication node through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0273] The embodiments of the present application further provide a storage medium. The storage medium stores a computer program, and when the computer program is executed by a processor, it implements any one of the model management methods in the embodiments of the present application. The method includes: receiving model management parameters or obtaining model management parameters in a predefined manner; measuring the power overhead of the model according to the model management parameters; reporting control information, where the control information includes at least one of the following: the power overhead, the power overhead level, and a request for a model operation instruction. Alternatively, the method includes: sending model management parameters, where the model management parameters are used to instruct a first communication node to measure the power overhead of the model; receiving the control information reported by the first communication node. The control information includes at least one of the following: the power overhead, the power overhead level, and a request for a model operation instruction.

[0274] The embodiments of the present application further provide a computer program product, including a computer program / instructions, and when the computer program / instructions are executed by a processor, they implement any one of the model management methods in the embodiments of the present application.

[0275] The computer storage medium of the embodiments of the present application may adopt any combination of one or more computer-readable media. The computer-readable media may be computer-readable signal media or computer-readable storage media. The computer-readable storage media may be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable CD-ROM, an optical storage device, a magnetic storage device, or any suitable combination of the above. The computer-readable storage media may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

[0276] The computer-readable signal media may include data signals propagated in a baseband or as part of a carrier wave, which carry computer-readable program codes. Such propagated data signals may take various forms, including but not limited to: electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal media may also be any computer-readable media other than the computer-readable storage media, and the computer-readable media may send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0277] The program codes contained on the computer-readable media may be transmitted by any appropriate media, including but not limited to: wireless, wire, optical cable, radio frequency (RF), etc., or any suitable combination of the above.

[0278] Computer program code for performing the operations of this application can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0279] The embodiments of this application also provide a computer program product, including a computer program / instructions, which when executed by a processor implement the video encoding method as described in any of the above embodiments.

[0280] As described above, the above are only exemplary embodiments of this application and are not intended to limit the protection scope of this application.

[0281] Those skilled in the art should understand that the term user terminal covers any suitable type of wireless user equipment, such as a mobile phone, a portable data processing device, a portable network browser, or a vehicle-mounted mobile station.

[0282] Generally speaking, various embodiments of this application can be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software that can be executed by a controller, a microprocessor, or other computing devices, although this application is not limited thereto.

[0283] The embodiments of this application can be implemented by a data processor of a mobile device executing computer program instructions, for example, in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0284] Any block diagram of a logical process in the accompanying drawings of this application may represent program steps, or may represent interconnected logical circuits, modules, and functions, or may represent a combination of program steps and logical circuits, modules, and functions. A computer program may be stored in a memory. The memory may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as but not limited to Read-Only Memory (ROM), Random Access Memory (RAM), optical memory devices and systems (such as Digital Video Disc (DVD) or Compact Disk (CD), etc.). The computer-readable medium may include a non-transitory storage medium. The data processor may be of any type suitable for the local technical environment, such as but not limited to a general-purpose computer, a special-purpose computer, a microprocessor, a Digital Signal Processing (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FGPA), and a processor based on a multi-core processor architecture.

[0285] By way of illustrative and non-limiting examples, a detailed description of exemplary embodiments of this application has been provided above. However, upon consideration of the accompanying drawings and the claims, various modifications and adaptations of the above embodiments will be apparent to those skilled in the art without departing from the scope of this application. Accordingly, the proper scope of this application will be determined in accordance with the claims.

Claims

1. A model management method, applied to a first communication node, characterized in that: include: Receiving model management parameters or obtaining model management parameters by a predefined method; measuring a power consumption of a model according to the model management parameters; Reporting control information, the control information comprising at least one of the following: the power consumption, the power consumption level, and the request model operation instruction.

2. The method according to claim 1, characterized in that The model management parameters include at least one of the following: Measurement cycle, measurement starting point, measurement duration, and reporting format.

3. The method according to claim 2, characterized in that The reporting format includes at least one of the following: The reported measurement value is at least power or work; The reported measurement values ​​are quantized in an equidistant or unequally spaced manner between an upper limit value and a lower limit value of the measurement, and the upper limit value and the lower limit value of the measurement are obtained by predefinition.

4. The method according to claim 1, characterized in that Report control information, including one of the following: Periodically report control information; Report control information when a reporting instruction is received.

5. The method according to claim 1, characterized in that Also includes: Report auxiliary information of the model operation, the auxiliary information includes at least one of the following: Model running time information and level information; The time information includes at least one of the following: a period, a duration, and an activation time starting point.

6. The method according to claim 1, characterized in that Also includes: Receiving activation instruction information or deactivation instruction information; When receiving the activation instruction information, running the model; When the deactivation instruction information is received, the model is stopped from running.

7. The method according to claim 1, characterized in that The request model operation instruction includes at least one of the following instructions: request model activation, request model deactivation, request model configuration instruction, request model transmission, and request configuration of power overhead level.

8. The method according to claim 7, characterized in that The model operation request instruction includes at least one of the following: an index of a model, and related parameters associated with the model index.

9. The method according to claim 1, characterized in that: Also includes: According to the indication information of the non-continuous operation cycle, the operation of the model is controlled to alternately be in an active state and a dormant state.

10. The method according to claim 9, characterized in that The indication information of the discontinuous operation cycle is bound to the energy saving mode of the first communication node.

11. The method according to claim 10, characterized in that The discontinuous operation period is consistent with or matches the activation period of the energy-saving mode; or, The duration of the activation period in the discontinuous operation cycle is consistent with or matches the activation period of the energy-saving mode; Among them, matching with the activation period of the energy-saving mode includes: being less than or equal to the activation period of the energy-saving mode.

12. The method according to claim 1, characterized in that Also includes: Obtain the relationship between control information and the influencing factors of the model; The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, model parameter quantity, and a split point of a bilateral model.

13. The method according to claim 12, characterized in that The relationship between the control information and the influencing factors of the model is determined by the second communication node, or obtained according to predefined information.

14. The method according to claim 1, characterized in that Also includes: An updated model is received, the updated model being determined by the second communication node based on the control information.

15. The method according to claim 1, characterized in that Also includes: The model is determined according to the power overhead level configured by the second communication node.

16. The method according to claim 12, characterized in that The relationship between the control information and the influencing factors of the model is adaptively adjusted by the second communication node according to the reported control information and / or the adjustment information of the model.

17. The method according to claim 1, characterized in that Also includes: receiving random access resource configuration information, wherein the partitions of the random access resources are associated with control information of the model; According to the random access resource configuration information, a random access resource is selected and a random access signal is transmitted.

18. A model management method, applied to a second communication node, characterized in that: include: Sending a model management parameter, where the model management parameter is used to indicate a power overhead of a measurement model of the first communication node; Receiving control information reported by the first communication node; The control information includes at least one of the following: the power consumption, the power consumption level, and a request model operation instruction.

19. The method according to claim 18, characterized in that The model management parameters include at least one of the following: Measurement cycle, measurement starting point, measurement duration, and reporting format.

20. The method according to claim 19, characterized in that The reporting format includes at least one of the following: The reported measurement value is at least power or work; The reported measurement values ​​are quantized in an equidistant or unequally spaced manner between an upper limit value and a lower limit value of the measurement, and the upper limit value and the lower limit value of the measurement are obtained by predefinition.

21. The method according to claim 18, characterized in that Receiving the control information includes one of the following: Periodically receiving the control information; The control information is received in case of sending a reporting indication.

22. The method according to claim 18, characterized in that Also includes: Receiving auxiliary information of the model running, the auxiliary information comprising at least one of the following: time information and level information of the model running; The time information includes at least one of the following: a period, a duration, and an activation time starting point.

23. The method according to claim 18, characterized in that Also includes: Sending activation instruction information or deactivation instruction information; The activation instruction information is used to instruct the first communication node to run the model; The deactivation indication information is used to instruct the first communication node to stop running the model.

24. The method according to claim 18, characterized in that The request model operation instruction includes at least one of the following instructions: request model activation, request model deactivation, request model configuration instruction, request model transmission, and request configuration of power overhead level.

25. The method according to claim 18, characterized in that Also includes: Determine the indication information of the discontinuous operation cycle, wherein the indication information of the discontinuous operation cycle is used to indicate the time when the operation of the model is in an active state and a dormant state.

26. The method according to claim 18, characterized in that Also includes: Determine the relationship between control information and the influencing factors of the model; The influencing factors include at least one of the following: model type, model accuracy, model depth, model width, model parameter quantity, and a split point of a bilateral model.

27. The method according to claim 26, characterized in that The relationship between the control information and the influencing factors of the model is determined by the second communication node, or obtained according to predefined information.

28. The method according to claim 18, characterized in that Also includes: updating the model according to the control information; The updated model is sent to the first communication node.

29. The method according to claim 18, characterized in that Also includes: Configure and send power overhead levels; The power overhead level is used to assist the first communication node in determining a model according to the power overhead level.

30. The method according to claim 26, characterized in that Also includes: According to the reported control information and / or the adjustment information of the model, the relationship between the control information and the influencing factors of the model is adaptively adjusted.

31. The method according to claim 18, characterized in that Also includes: Random access resource configuration information is determined, wherein the partitions of the random access resources are associated with control information of the model.

32. A communication node, characterized in that: include: memory, and one or more processors; The memory is configured to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the model management method according to any one of claims 1 to 31.

33. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the model management method as described in any one of claims 1 to 31 is implemented.

Images