Wireless communication method and communication equipment

Abstract

The invention provides a wireless communication method and communication equipment. The method comprises the following steps: a first device sends capability information of the first device to a second device, wherein the capability information is associated with one or more of the following: model training; compiling the model; and deploying the model. In the embodiment of the application, the first device can send the capability information related to the model to the second device to inform the first device of the capability information, so that compared with the traditional scheme that the first device and the second device cannot know the capability information of the opposite side, the second device is helped to know the capability information of the first device, and the user experience is improved. In one embodiment, tasks (e.g., a model training task, a model updating task, a model parameter fine tuning task, a model deployment task, a model compiling task, etc.) matching the capability information are executed by the first device to improve the likelihood of model-based communication between the first device and the second device.

Description

Wireless communication method and communication device Technical Field

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

[0002] Currently, model-based communication solutions are being introduced in practical communication systems. This means that multiple devices can communicate based on their own models. As communication systems become increasingly widespread, they incorporate a wide range of devices, such as terminal devices, terminal-side servers, network devices, and network-side servers. Different devices may have different model-related capabilities, or in other words, different levels of model support. If model-related capabilities are lumped together across devices, model-based communication between them may fail.

[0003] Summary of the Invention

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

[0005] In a first aspect, a method for wireless communication is provided, comprising: a first device sending capability information of the first device to a second device, wherein the capability information is associated with one or more of the following: model training; model compilation; and model deployment.

[0006] In a second aspect, a method for wireless communication is provided, including: a second device receiving capability information of the first device sent by a first device, wherein the capability information is associated with one or more of the following: model training; model compilation; model deployment.

[0007] According to a third aspect, a communication device is provided, which is a first device and includes: a sending unit for sending capability information of the first device to a second device, wherein the capability information is associated with one or more of the following: model training; model compilation; model deployment.

[0008] In a fourth aspect, a communication device is provided, which is a second device and includes: a receiving unit for receiving capability information of the first device sent by a first device, wherein the capability information is associated with one or more of the following: model training; model compilation; model deployment.

[0009] In a fifth aspect, a communication device is provided, comprising a processor, a memory, and a communication interface, wherein the memory is used to store one or more computer programs, and the processor is used to call the computer program in the memory so that the communication device executes part or all of the steps in the methods of the above aspects.

[0010] In a sixth aspect, an embodiment of the present application provides a communication system, which includes the above-mentioned terminal device and / or network device. In another possible design, the system may also include other devices that interact with the first device or the second device in the solution provided in the embodiment of the present application.

[0011] In the seventh aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program enables a communication device (for example, a first device or a second device) to perform some or all of the steps in the methods of the above aspects.

[0012] In an eighth aspect, embodiments of the present application provide a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a communication device (e.g., a first device or a second device) to perform some or all of the steps of the methods described in each of the above aspects. In some implementations, the computer program product may be a software installation package.

[0013] In a ninth aspect, an embodiment of the present application provides a chip comprising a memory and a processor, wherein the processor can call and run a computer program from the memory to implement some or all of the steps described in the methods of the above aspects.

[0014] In an embodiment of the present application, the first device can send capability information related to the model to the second device to inform the capability information of the first device. Compared with the traditional solution in which the first device and the second device cannot know each other's capability information, this helps the second device to know the capability information of the first device, so as to select tasks that match the capability information (for example, model training tasks, model update tasks, fine-tuning model parameter tasks, model deployment tasks, model compilation tasks, etc.) to be executed by the first device, so as to increase the possibility of model-based communication between the first device and the second device. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] FIG1 is a wireless communication system 100 used in an embodiment of the present application.

[0016] FIG2 is a schematic diagram of channel estimation and signal recovery applicable to an embodiment of the present application.

[0017] FIG3 is a schematic diagram of a channel-state information (CSI) feedback system based on an artificial intelligence (AI) model applicable to an embodiment of the present application.

[0018] FIG4 shows a schematic diagram of an AI model-based positioning solution applicable to an embodiment of the present application.

[0019] Figure 5 shows a schematic diagram of AI model-based beam management applicable to an embodiment of the present application.

[0020] FIG6 is a schematic diagram of a neural network applicable to an embodiment of the present application.

[0021] FIG7 is a schematic diagram of a convolutional neural network (CNN) applicable to an embodiment of the present application.

[0022] FIG8 is a flowchart of a wireless communication method according to an embodiment of the present application.

[0023] FIG9 is a flowchart of a wireless communication method according to another embodiment of the present application.

[0024] FIG10 is a flowchart of a wireless communication method according to another embodiment of the present application.

[0025] FIG11 is a schematic diagram of a communication device according to an embodiment of the present application.

[0026] FIG12 is a schematic diagram of a communication device according to another embodiment of the present application.

[0027] FIG13 is a schematic structural diagram of a communication device according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] The technical solution in this application will be described below with reference to the accompanying drawings.

[0029] Figure 1 illustrates a wireless communication system 100 used in an embodiment of the present application. The wireless communication system 100 may include a network device 110 and a terminal device 120. The network device 110 may be a device that communicates with the terminal device 120. The network device 110 may provide communication coverage for a specific geographic area and may communicate with the terminal device 120 within the coverage area.

[0030] FIG1 exemplarily shows a network device and two terminals. Optionally, the wireless communication system 100 may include multiple network devices and each network device may include other numbers of terminal devices within its coverage area, which is not limited in the embodiments of the present application.

[0031] Optionally, the wireless communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiment of the present application.

[0032] It should be understood that the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: fifth generation (5G) system or new radio (NR), long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), etc. The technical solutions provided in this application can also be applied to future communication systems, such as the sixth generation mobile communication system, satellite communication system, etc.

[0033] The terminal device in the embodiments of the present application may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station (MS), mobile terminal (MT), remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device. The terminal device in the embodiments of the present application may refer to a device that provides voice and / or data connectivity to a user and can be used to connect people, objects and machines, such as a handheld device with wireless connection function, a vehicle-mounted device, etc. The terminal device in the embodiments of the present application can be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal in industrial control, a wireless terminal in self-driving, a wireless terminal in remote medical surgery, a wireless terminal in a smart grid, a wireless terminal in transportation safety, a wireless terminal in a smart city, a wireless terminal in a smart home, etc. Optionally, the UE can be used to act as a base station. For example, the UE can act as a scheduling entity that provides sidelink signals between UEs in V2X or D2D, etc. For example, a cellular phone and a car communicate with each other using sidelink signals. The cellular phone and smart home devices communicate without relaying the communication signal through the base station.

[0034] The network device in the embodiments of the present application may be a device for communicating with a terminal device, and may also be referred to as an access network device or a radio access network device. For example, the network device may be a base station. The network device in the embodiments of the present application may refer to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network. A base station can broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station MeNB, secondary station SeNB, multi-standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), positioning node, etc. A base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. A base station can also refer to a communication module, a modem or a chip used to be set in the aforementioned device or apparatus. The base station can also be a mobile switching center and a device that performs base station functions in device-to-device D2D, vehicle-to-everything (V2X), and machine-to-machine (M2M) communications, a network-side device in a 6G network, or a device that performs base station functions in future communication systems. The base station can support networks with the same or different access technologies. The embodiments of this application do not limit the specific technology and specific device form used by the network equipment.

[0035] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0036] In some deployments, the network device in the embodiments of the present application may refer to a CU or a DU, or the network device may include a CU and a DU. The gNB may also include an AAU.

[0037] The network equipment and terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; they can also be deployed in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the scenarios in which the network equipment and terminal devices are located.

[0038] It should be understood that all or part of the functions of the communication device in this application can also be implemented through software functions running on hardware, or through virtualization functions instantiated on a platform (such as a cloud platform).

[0039] With the development of artificial intelligence (AI) technology, AI models are being introduced into more and more communication processes. To facilitate understanding, the following describes the AI ​​models used in communication processes with reference to Figures 2 to 5.

[0040] Channel estimation and signal recovery based on AI model

[0041] Due to the complexity and time-varying nature of wireless channel environments, in wireless communication systems (e.g., the wireless communication systems described above), a receiver needs to recover received signals based on channel estimation results. Figure 2 is a schematic diagram of channel estimation and signal recovery applicable to embodiments of the present application.

[0042] As shown in FIG2 , in step S210 , the transmitter transmits, in addition to the data signal, a series of pilot signals known to the receiver on the time-frequency resources, such as the channel state information-reference signal (CSI-RS) and the demodulation reference signal (DMRS).

[0043] In step S211, the transmitter transmits the above data signal and pilot signal to the transmitter through the channel.

[0044] In step S212, after receiving the pilot signal, the receiver may perform channel estimation. In one possible implementation, the receiver may estimate channel information of the channel transmitting the pilot signal based on a pre-stored pilot sequence and the received pilot sequence using a channel estimation algorithm (e.g., a least squares (LS) channel estimation method).

[0045] In step S213, the receiver may recover the channel information on all time-frequency resources using an interpolation algorithm based on the channel information of the channel transmitting the pilot sequence, for subsequent channel state information (CSI) feedback or data recovery.

[0046] CSI feedback system based on AI model

[0047] In wireless communication systems, codebook-based solutions are primarily used to extract and provide feedback on channel characteristics. This means that after the receiver performs channel estimation, it selects the precoding matrix that best matches the current channel from a pre-set precoding codebook based on the estimation results and an optimization criterion. The receiver then feeds the precoding matrix index (PMI) back to the transmitter via an air interface feedback link for precoding. In some implementations, the receiver can also provide the transmitter with a measured channel quality indicator (CQI) to facilitate adaptive modulation and coding.

[0048] Figure 3 is a schematic diagram of a CSI feedback system based on an AI model applicable to an embodiment of the present application. As shown in Figure 3, the entire feedback system includes an AI encoder 311 and an AI decoder 321 part of the autoencoder, wherein the AI ​​encoder 311 is deployed at the transmitter 310 and the AI ​​decoder 321 is deployed at the receiver 320. The transmitter 310 compresses and encodes the CSI to be transmitted through the AI ​​encoder 311 to obtain compressed CSI. The compressed CSI is then fed back to the receiver 320 through the feedback link, and the receiver 320 decodes the compressed CSI through the AI ​​decoder 321 to obtain the recovered CSI. In this way, the communication overhead of feedback CSI can be saved without affecting the accuracy of CSI transmission.

[0049] Positioning based on AI models

[0050] In cellular wireless positioning, the straight-line propagation of electromagnetic waves between network devices and terminal devices is called line-of-sight (LOS) wireless propagation. In some cases, electromagnetic wave signals cannot propagate in a straight line due to obstruction by buildings or trees, which is usually called non-line-of-sight (NLOS) wireless propagation. Traditional positioning algorithms such as time difference of arrival (TDOA) and angle-of-arrival (AOA) are based on LOS channels and are no longer applicable in environments where NLOS is predominant. In most scenarios, the number of network devices with LOS channels to terminal devices is often small, resulting in the inability of traditional positioning algorithms to meet the requirements of high-precision positioning. In addition, there may be some non-ideal factors in actual systems, which can lead to reduced positioning accuracy.

[0051] Therefore, a high-precision positioning method based on AI models has been proposed for scenarios where LOS / NLOS channels coexist. Existing research results have shown that by using machine learning methods to train models based on large amounts of channel data and to explore the mapping relationship between channel responses and location coordinates, it is possible to address the limitations of traditional positioning algorithms in LOS / NLOS channel coexistence scenarios and improve positioning accuracy.

[0052] FIG4 shows a schematic diagram of an AI model-based positioning solution applicable to an embodiment of the present application. Referring to FIG4 , in a positioning solution based on an AI model 410 in a LOS / NLOS channel coexistence scenario, the channel response can be used as the input of the AI ​​model 410, and the location coordinates can be used as the output of the AI ​​model 410. The AI ​​model 410 learns the intrinsic relationship between the wireless channel and the location of the terminal device. In this way, even in a scenario where there are not enough LOS channels and / or in a scenario where there are non-ideal conditions, the positioning solution based on the AI ​​model 410 can also output the location coordinates of the terminal device with higher accuracy, which helps to meet the needs of high-precision positioning.

[0053] The above introduces several communication processes applicable to the AI ​​model. The following introduces the AI ​​model applicable to the embodiments of the present application. It should be noted that the AI ​​model applicable to the embodiments of the present application is not limited to the several AI models introduced below.

[0054] AI-based beam management

[0055] In the traditional beam selection process, it is usually necessary to traverse all combinations of receive beams and transmit beams to select the appropriate beam. However, traversing all combinations takes a long time, resulting in low beam selection efficiency.

[0056] For example, suppose the network equipment deploys 64 different downlink transmission directions in FR2 (carried by up to 64 synchronization signals and physical broadcast channel blocks (SSB)). Accordingly, the terminal device uses one or more antenna panels to simultaneously scan the receiving beams when receiving, and each antenna panel has 4 receiving beams. Then the terminal device needs to measure at least 256 beam pairs, which means that 256 resources of downlink resource overhead are required. From a time perspective, each SSB cycle is approximately 20ms, and 4 SSB cycles are required to complete the measurement of 4 receiving beams. Assuming that multiple receiving antenna panels can perform beam scanning simultaneously, it will take at least 80ms.

[0057] As the number of beams in future massive multiple-input, multiple-output (MIMO) systems increases, using beam scanning-based beam management solutions to match optimal beam pairs will only result in increased reference signal transmission overhead and beam scanning latency. Therefore, to avoid these issues, AI-based beam management was proposed in Release 18. The following describes this AI-based beam management solution, combining the training and prediction processes of the AI ​​model.

[0058] Assume that the AI ​​model is used to predict the available beams in beam set A. Accordingly, during the training phase, the beam measurement results of beam set B can be used as AI model training data. That is, the AI ​​model is trained based on the beam measurement results of beam set B so that the AI ​​model can predict the available beams from beam set A.

[0059] It should be noted that the beam measurement results of the above-mentioned beam set B may include the measurement results corresponding to the layer 1 (layer1, L1) measurement quantity, and / or the indication information of the selected beam in beam set B (for example, the transmitting beam identifier, the receiving beam identifier or the beam pair identifier, etc.).

[0060] In some implementations, the training data may also include label information of beam set A, and the label information is used to indicate one or more of the following beams in beam set A: optimal transmit beam, optimal receive beam, optimal beam pair, better multiple transmit beams, better multiple receive beams, better beam pair, etc.

[0061] As shown in Figure 5, in the prediction stage, the input of the AI ​​model 510 may include the link quality measurement results (for example, L1 measurement quantity) corresponding to the beams in the beam set A, and the prediction results output by the AI ​​model 510 may include the target beam selected from the beam set A, and the link quality corresponding to the target beam.

[0062] In some implementations, the target beam may be one or more beams. For example, if the target beam is a single beam, the target beam may be the optimal beam or a relatively optimal beam in beam set A. For example, if the target beam is multiple beams, the target beam may be multiple beams in beam set A that meet the requirements. "Meeting the requirements" may be understood as meaning that the link quality corresponding to the beam meets the requirements, for example, the link quality corresponding to the beam is greater than or equal to a threshold.

[0063] In other implementations, the target beam may refer to one or more beam pairs, each of which may include a receive beam or a transmit beam. For example, if the target beam is a single beam pair, the target beam may be the optimal beam pair or a relatively optimal beam pair in beam set A. For example, if the target beam is multiple beam pairs, the target beam may be multiple beam pairs in beam set A that meet the requirements. "Meeting the requirements" may be understood as meaning that the link quality corresponding to the beam pair meets the requirements, for example, the link quality corresponding to the beam pair is greater than or equal to a threshold.

[0064] It should be noted that the link quality in the embodiment of the present application can be determined by one or more measurement quantities described above. Of course, the link quality in the embodiment of the present application can also be determined based on other measurement quantities in future communication systems, and the embodiment of the present application is not limited to this.

[0065] In addition, the link quality is determined based on one or more measurement quantities, which can be understood as the link quality being obtained by processing one or more measurement quantities. Of course, the link quality can also be a measurement quantity, which is not limited in the present embodiment.

[0066] It should also be noted that if the prediction result only indicates one beam in the beam pair, the other beam in the beam pair can be determined by other means. For example, it can be determined by one or some of the processes P1 to P3 in the traditional beam selection process. Of course, it can also be determined by one or some of the processes U1 to U3 in the traditional beam selection process. The embodiments of the present application are not limited to this.

[0067] In some implementations, the beam set B may be a different beam set from the beam set A. In some implementations, the beam set B may be a subset of the beam set A. Accordingly, by measuring fewer beams (beams in the beam set B), predictions for more beams (beams in the beam set A) may be achieved. Compared with the above-mentioned scheme of selecting beams based on traversing all combinations, it helps to reduce the time of executing the beam selection process. Of course, in the embodiment of the present application, the beams in the beam set B and the beams in the beam set A may be completely different beams. For example, there is no intersection between the beam set B and the beam set A, but the beam direction corresponding to the beam set B may be similar to the beam direction corresponding to the beam set A.

[0068] In some other implementations, the beam set B may be exactly the same as the beam set A.

[0069] The above describes the communication process applicable to the AI ​​model in conjunction with Figures 2 to 5. The following describes the AI ​​model applicable to the embodiments of the present application in conjunction with Figures 6 to 7.

[0070] AI models

[0071] In recent years, artificial intelligence research, exemplified by neural networks, has achieved remarkable success in many fields, and will continue to play a vital role in people's lives and production for a long time to come. A neural network can be understood as a computational model consisting of multiple interconnected neuron nodes. The connections between these nodes represent the weighted values ​​from input signals to output signals, often referred to as weights. Each node performs a weighted summation of different input signals and outputs the result through a specific activation function.

[0072] Common neural networks include CNN, recurrent neural network (RNN), deep neural network (DNN), etc.

[0073] The following describes a neural network applicable to embodiments of the present application in conjunction with FIG6 . The neural network shown in FIG6 can be divided into three categories based on the location of different layers: input layer 610, hidden layer 620, and output layer 630. Generally speaking, the first layer is the input layer 610, the last layer is the output layer 630, and the intermediate layers between the first and last layers are all hidden layers 620.

[0074] The input layer 610 is used to input data, where the input data can be, for example, a received signal received by a receiver. The hidden layer 620 is used to process the input data, for example, decompress the received signal. The output layer 630 is used to output processed output data, for example, a decompressed signal.

[0075] As shown in Figure 6, a neural network consists of multiple layers, each of which contains multiple neurons. The neurons between layers can be fully connected or partially connected. For connected neurons, the output of the neurons in the previous layer can serve as the input of the neurons in the next layer.

[0076] With the continuous advancement of neural network research, deep learning algorithms have been proposed in recent years. These algorithms introduce a large number of hidden layers into neural networks, forming DNNs. More hidden layers allow DNNs to better capture complex real-world situations. Theoretically, a model with more parameters has higher complexity and a greater "capacity," meaning it can handle more complex learning tasks. These neural network models are widely used in pattern recognition, signal processing, optimization and combination, anomaly detection, and other fields.

[0077] CNN is a deep neural network with a convolutional structure, and its structure is shown in FIG7 , which may include an input layer 710 , a convolutional layer 720 , a pooling layer 730 , a fully connected layer 740 , and an output layer 750 .

[0078] Each convolution layer 720 may include a plurality of convolution operators, which are also called kernels. The convolution operator can be regarded as a filter for extracting specific information from the input signal. The convolution operator can essentially be a weight matrix, which is usually predefined.

[0079] The weight values ​​in these weight matrices need to be obtained through a lot of training in practical applications. The weight matrices formed by the weight values ​​obtained through training can extract information from the input signal, thereby helping CNN to make correct predictions.

[0080] When CNN has multiple convolutional layers, the initial convolutional layer tends to extract more general features, which can also be called low-level features. As the depth of CNN increases, the features extracted by the subsequent convolutional layers become more and more complex.

[0081] Pooling layers 730 are often used periodically after convolutional layers to reduce the number of training parameters. For example, a single convolutional layer can be followed by a pooling layer, as shown in Figure 7, or multiple convolutional layers can be followed by one or more pooling layers. In signal processing, the sole purpose of a pooling layer is to reduce the spatial size of the extracted information.

[0082] The fully connected layer 740, after being processed by the convolution layer 720 and the pooling layer 730, is not sufficient for CNN to output the required output information. Because as mentioned above, the convolution layer 720 and the pooling layer 730 only extract features and reduce the parameters brought by the input data. However, in order to generate the final output information (for example, the bit stream of the original information transmitted by the transmitter), CNN also needs to use the fully connected layer 740. Generally, the fully connected layer 740 may include multiple hidden layers, and the parameters contained in the multiple hidden layers can be pre-trained based on the relevant training data of the specific task type. For example, the task type may include decoding the data signal received by the receiver. For example, the task type may also include channel estimation based on the pilot signal received by the receiver.

[0083] Following the multiple hidden layers in the fully connected layer 740, the final layer of the CNN is the output layer 750, which is used to output the results. Typically, this output layer 750 is configured with a loss function (e.g., a loss function similar to categorical cross entropy) to calculate the prediction error, or to evaluate the degree of difference between the output of the CNN model (also known as the predicted value) and the ideal result (also known as the true value).

[0084] To minimize the loss function, the CNN model needs to be trained. In some implementations, the CNN model can be trained using the backpropagation algorithm (BP). The BP training process consists of a forward propagation process and a backward propagation process. During the forward propagation process (e.g., the propagation from 710 to 750 in Figure 7 is forward propagation), the input data is fed into the aforementioned layers of the CNN model, processed layer by layer, and transmitted to the output layer. If the output result of the output layer differs significantly from the ideal result, the aforementioned loss function is minimized as the optimization goal, and the backward propagation process is switched to (e.g., the propagation from 750 to 710 in Figure 7 is backward propagation). The partial derivatives of the optimization goal with respect to each neuron weight are calculated layer by layer, forming the gradient of the optimization goal with respect to the weight vector, which serves as the basis for modifying the model weights. The CNN training process is completed during the weight modification process. When the aforementioned error reaches the desired value, the CNN training process ends.

[0085] It should be noted that the CNN shown in Figure 7 is only an example of a convolutional neural network. In specific applications, the convolutional neural network can also exist in the form of other network models, and the embodiments of the present application are not limited to this.

[0086] RNNs are designed to process sequential data. In traditional neural network models (for example, CNN models), the layers are fully connected, from the input layer to the hidden layer to the output layer, and the nodes within each layer are disconnected. However, these ordinary neural networks are inadequate for many problems. For example, if you want to predict the next word in a sentence, you generally need to use the previous word, because the previous and next words in a sentence are not independent. RNNs are called recurrent neural networks because the current output of a sequence is also related to the previous output. Specifically, the network remembers the previous information and applies it to the calculation of the current output. That is, the nodes between hidden layers are no longer disconnected but connected, and the input of the hidden layer includes not only the output of the input layer but also the output of the hidden layer at the previous moment. In theory, RNNs can process sequence data of any length.

[0087] Training an RNN is similar to training a traditional ANN (artificial neural network). The same backpropagation error algorithm is used, but there is a slight difference. If the RNN is expanded, the parameters W, U, and V are shared, while traditional neural networks are not. Furthermore, when using the gradient descent algorithm, the output of each step depends not only on the network state at the current step, but also on the state of the network at the previous steps. For example, at t = 4, three steps need to be propagated backward, and various gradients need to be added to the three subsequent steps. This learning algorithm is called backpropagation through time (BPTT).

[0088] Given the existence of artificial neural networks and convolutional neural networks, why do we still need recurrent neural networks? The reason is simple. Both convolutional and artificial neural networks assume that elements are independent of each other, and that inputs and outputs are also independent, like cats and dogs. However, in the real world, many elements are interconnected, such as the changes in stock prices over time. For example, someone said, "I love traveling, and my favorite place is Yunnan. I must visit __ someday." Everyone knows to fill in the blank with "Yunnan." This is because we infer this information based on the context, but achieving this is quite difficult. Therefore, recurrent neural networks were developed. Their essence is that they possess memory, just like humans. Therefore, their output depends on the current input and memory.

[0089] Model training

[0090] Currently, academic research and research within the international standards organization 3GPP on models used in communication systems primarily focuses on model design and their use after training. During this phase, it's often assumed that models can be pre-trained offline. For example, a model is pre-trained offline and ready for use. However, deploying pre-trained models in different scenarios and environments to solve different problems is inherently challenging both in engineering and implementation. This means it's difficult to pre-train a single model to solve all problems. Furthermore, it's also difficult to pre-prepare model-based solutions for a wide range of wireless communication-related issues. Therefore, adopting a non-offline, on-demand model building approach becomes crucial.

[0091] Typically, the model building process involves two phases: training and inference. The purpose of the training phase is to train the device to complete a specific model based on a specific dataset. Subsequently, in the inference phase, the trained model can be used for inference. Current 3GPP discussions primarily focus on single-end and dual-end models, as well as online and offline training. These are described below.

[0092] A single-ended model refers to a model used at a specific communication device. Because the model is used solely on a communication device, it can be understood that the model is used at one end of the communication device. Therefore, this type of model can also be called a "single-ended model." The communication device can be, for example, a terminal device or a network device. For example, the AI ​​model 410 described above can be deployed at a network device to locate the terminal device, and therefore serves as an example of a single-ended model.

[0093] A dual-end model refers to a model that is deployed separately in communication devices that communicate with each other, and the models in the communication devices that communicate with each other need to work together. Since this type of model needs to be deployed separately in the communication devices at both ends of the communication and work together, this model can also be called a "dual-end model." For example, the AI ​​model encoder and AI model decoder introduced above need to be deployed separately on the sending end and the receiving end, and the AI ​​model encoder and AI model decoder deployed on the sending end and the receiving end need to be configured and used with each other. Therefore, the AI ​​model encoder and AI model decoder can be used as examples of a dual-end model.

[0094] In some scenarios, the use of the above-mentioned two-end models in conjunction with each other can be understood as the two-end models cooperating with each other to perform joint reasoning, and the joint reasoning is jointly executed on devices that communicate with each other, that is, the first part of the reasoning is first executed by one device, and then the rest of the reasoning is executed by the other device.

[0095] Offline training refers to the process of data collection and model training that does not require real-time interaction with the external environment. In offline training, data can be collected and preprocessed in advance, and the model can be trained in an environment with sufficient computing resources. Typically, after offline training, the model is used for inference after a period of time, and does not enter the inference phase immediately.

[0096] Currently, academic research and research within the international standards organization 3GPP on wireless AI / ML primarily focuses on model design and how to use trained models. Model training generally assumes that the model can be trained, for example, by completing offline training. However, offline training often fails to cover all communication scenarios, making the trained model inappropriate. Therefore, online training becomes crucial.

[0097] Online training refers to the process of adjusting (or updating) a model using continuously collected real-time data after it has already been deployed and is running. In other words, the model is trained based on real-time data and used in real time (or near real time). Online training can promptly respond to changes in new data, allowing the model to continuously adapt to new situations and changes, thereby improving its accuracy and effectiveness.

[0098] In some implementations, model training for a dual-end model can be divided into three methods based on different training methods. The following describes methods 1 to 3.

[0099] Method 1: Model training for a dual-end model is performed by a single end. In this model training method, the end performing model training can transmit the trained model of the dual-end model that needs to be deployed on the other end (called the peer model) to the other end. At this time, the other end can directly use the transmitted peer model without training the model itself.

[0100] Method 2: The dual-end model is trained separately by devices that communicate with each other. Therefore, this method of model training can also be called "dual-end training." In other words, the dual-end model is trained successively by devices that communicate with each other. For example, the dual-end model includes Model 1 deployed in Device 1 and Model 2 deployed in Device 2. Device 1 can first train Model 1 and transmit the data used to train Model 2 to Device 2 after Model 1 training is completed. Accordingly, Device 2 can train Model 2 based on the data. The device that performs model training first (for example, Device 1) can be called the "first-trained device," and the device that performs model training later (for example, Device 2) can be called the "latter-trained device."

[0101] Taking the AI ​​encoder 311 and AI decoder 321 introduced in the previous text in conjunction with Figure 3 as an example, the AI ​​encoder 311 is deployed at the transmitting end and the AI ​​decoder 321 is deployed at the receiving end. Accordingly, method 2 can be used to perform model training on the AI ​​encoder 311 and the AI ​​decoder 321 respectively.

[0102] In method three, the two-end model is jointly trained by communicating devices. Therefore, this model training method can also be called "dual-end training." Assume that the two-end model includes Model 1 deployed on Device 1 and Model 2 deployed on Device 2. The output of Model 1 on Device 1 can be used as the input of Model 2 on Device 2. Then, Model 1 and Model 2 can be jointly trained based on the output of Model 2.

[0103] Taking the AI ​​encoder 311 and AI decoder 321 described above in conjunction with FIG3 as an example, the AI ​​encoder 311 is deployed at the transmitting end, and the AI ​​decoder 321 is deployed at the receiving end. Accordingly, the AI ​​encoder 311 and the AI ​​decoder 321 can be jointly trained using method three.

[0104] In some implementations, the training results of the model training of the third method may include one or more of the following: obtaining a trained model deployed on the first device, and / or obtaining a trained model deployed on the second device.

[0105] As mentioned earlier, model-based communication solutions have been introduced in practical communication systems. This means that multiple devices can communicate based on their own models. As communication systems become increasingly widespread, they are incorporating a wide range of devices, such as terminal devices, terminal servers, network devices, and network servers. Different devices may have different model-related capabilities, or in other words, different levels of model support. If model-related capabilities are lumped together across devices, model-based communication between them may fail. To facilitate understanding, the following example uses the different capabilities associated with model training across different devices as an example.

[0106] In some scenarios, different devices support different types of model training. For example, if device 1 supports model training method 1 and device 2 supports model training method 2, device 1 cannot participate in model training for method 2. This means that device 1 cannot collaborate with device 2 to complete model training for method 2. This may result in poor inference accuracy for the model on device 2, or even prevent the model on device 2 from entering the inference phase. Consequently, device 1 and device 2 cannot communicate based on their respective deployed models.

[0107] In other scenarios, different devices support different model structures for model training. For example, if device 1 supports model training for model structure 1, and device 2 supports model training for model structure 2, the dataset transmitted by device 2 to device 1 for model training may not be suitable for model training on device 2. Consequently, devices 1 and 2 cannot communicate based on their respective deployed models.

[0108] In other scenarios, different devices may have different levels of support for the same model training process. For example, for the same model training process, different devices may have different computing resources available for model training. For another example, for the same model training process, different devices may have different power consumption available for model training. For another example, because different devices support different latency, even for the same model training process, the time available for model training may be different due to the latency limitations supported by different devices. For another example, different devices may support different software and hardware, and due to the limitations of the software and hardware in different devices, different devices may support different types of model training. Therefore, due to the different levels of support for the model training process on different devices, devices may not be able to cooperate with each other to complete model training, and may not be able to communicate based on the trained model afterwards.

[0109] The applicant found that the root cause of the above problems is that different devices have different capabilities associated with the model, and devices communicating based on the model cannot obtain each other's capability information, resulting in the ability to generalize the model-related capabilities in the device, which causes the problem of failure of model-based communication between devices.

[0110] Therefore, in response to the above problems, an embodiment of the present application provides a method for wireless communication. In this method, a first device can send capability information related to a model to a second device to inform the capability information of the first device. Compared with the traditional solution in which the first device and the second device cannot know each other's capability information, this method helps the second device to know the capability information of the first device, so as to select tasks that match the capability information (for example, model training tasks, model update tasks, fine-tuning model parameter tasks, model deployment tasks, model compilation tasks, etc.) to be performed by the first device, thereby increasing the possibility of model-based communication between the first device and the second device.

[0111] It should be noted that in some implementations, the above model may be an AI model, such as any of the AI ​​models described above in conjunction with Figures 2 to 5. Of course, the above model may also be a newly introduced model in future communication systems. In other implementations, the above model may be an ML model.

[0112] The following describes a flow chart of a wireless communication method according to an embodiment of the present application in conjunction with Figure 8. The method shown in Figure 8 includes step S810.

[0113] In step S810 , the first device sends capability information of the first device to the second device.

[0114] In some implementations, the capability information of the first device may be at the device level. That is, the capability information of the first device is for the entire device, and the capability information for different model use cases is not differentiated. In this case, the capability information for different model use cases may be indicated together, or the capability information for different model use cases may be determined together.

[0115] In other implementations, the capability information of the first device may be at the model use case level, that is, the capability information of the first device may be specific to the model use case. In this case, the capability information for different model use cases may be independently indicated, or the capability information for different model use cases may be independently determined.

[0116] For example, the first device may be deployed with model use case 1 for beam management and model use case 2 for CSI information compression. At this time, the capability information associated with model use case 1 and the capability information associated with model use case 2 in the first device may be independently determined and transmitted independently.

[0117] In some implementations, the first device may be associated with multiple model use cases, and the capability information of some or all of the multiple model use cases may be different. Of course, in the embodiment of the present application, the capability information for different model use cases in the multiple model use cases may be the same.

[0118] In some implementations, the capability information is associated with one or more of the following: model training; model compilation; and model deployment. These are described below with reference to Examples 1 to 3.

[0119] Example 1: Capability information is associated with model training.

[0120] In some implementations, the capability information may include one or more of the first information, the second information, and the third information. These are described below with reference to Examples 1-1 to 1-3, respectively.

[0121] Example 1-1: Capability information includes first information.

[0122] In some implementations, the first information is used to indicate whether the first device supports model training.

[0123] In some implementations, the first information may occupy one bit in the capability information, which helps reduce the overhead required to transmit the first information. In this case, if the value of the bit is the first value, the bit can be used to indicate that the first device supports model training. If the value of the bit is the second value, the bit can be used to indicate that the first device does not support model training. The first value may be different from the second value, for example, the first value may be 1 and the second value may be 0. For another example, the first value may be 0 and the second value may be 1. Of course, in an embodiment of the present application, the first information may be carried by multiple bits.

[0124] Generally, the model training phase has much higher requirements in terms of computing, storage, power consumption, etc. than the model inference phase. Therefore, a device that supports model inference does not necessarily support model training. In an embodiment of the present application, a first device can indicate whether the first device supports model training by sending a first message to a second device, thereby increasing the possibility of model-based communication between the first and second devices.

[0125] Example 1-2: The capability information includes the second information.

[0126] In some implementations, the second information indicates the model training types supported by the first device. These training types may include one or more of the following: online training, and dual-end model training methods 1 through 3. For details, see the above description of model training; for brevity, these details are omitted here.

[0127] In some implementations, if the training type of the model training includes online training, the second information also includes one or more of the following: information indicating whether the first device supports online training; information indicating the duration of time supported by the first device for online training; information indicating the training complexity of the online training supported by the first device; information indicating the amount of data supported by the first device for online training.

[0128] In an embodiment of the present application, refining the information content of the second information for online training helps the second device determine the online training that the first device can participate in based on the second information, helps avoid the second device instructing the first device to participate in online training that the first device does not support, and improves the rationality of the second device instructing the first device to participate in model training.

[0129] If the second information includes information for indicating whether the first device supports online training, the information can occupy one bit in the second information, which helps to reduce the overhead required to transmit the second information. At this time, if the value of the bit is the first value, the bit can be used to indicate that the first device supports online training. If the value of the bit is the second value, the bit can be used to indicate that the first device does not support online training. The first value can be different from the second value. For example, the first value can be 1 and the second value can be 0. For another example, the first value can be 0 and the second value can be 1. Of course, in an embodiment of the present application, the above information can be carried by multiple bits.

[0130] If the second information includes information indicating the duration, in some implementations, the first device may occupy all of the time for online training within the duration, or may occupy part of the time for online training within the duration.

[0131] In some scenarios, the online training supported by the first device may be limited in duration. For example, the computing power of the first device may only support model training for a certain duration. For another example, the usage of the first device may indicate that the first device only supports model training for a certain duration. Therefore, in embodiments of the present application, the duration can be indicated by the second information so that the second device can select a reasonable model training for the first device.

[0132] In some implementations, the duration may be determined based on a reference model. For example, the duration may be the time it takes the first device to perform online training on the reference model. In other implementations, the duration may be determined based on a reference training complexity. For example, in the case of a reference training complexity, the duration may be the time it takes the first device to perform online training on the reference model.

[0133] In the embodiment of the present application, the value of the duration is not specifically limited. For example, the value of the duration may include one or more of the following: 10ms, 50ms, 100ms, 1s, 2s, 5s, 10s, 1 minute, 2 minutes, 5 minutes, 10 minutes, 30 minutes.

[0134] If the second information includes information indicating the aforementioned training complexity, in some implementations, training complexity can be understood as the computing resources and time required during the online training process. Typically, training complexity can be determined by the following factors: data volume, model complexity, training algorithm, and computing resources.

[0135] In some scenarios, the online training supported by the first device may have limitations in terms of training complexity. For example, the computing power of the first device may only support model training within a certain training complexity. Therefore, in embodiments of the present application, the second information can be used to indicate the training complexity so that the second device can select a reasonable model training for the first device.

[0136] If the second information includes information indicating the amount of data available for online training, in some implementations, the data available for online training may include one or more of training data, validation data, test data, metadata, and prior knowledge data. Of course, in embodiments of the present application, the data available for online training may also include other data.

[0137] In some scenarios, the online training supported by the first device may be limited in terms of data volume. For example, the computing power of the first device may only support model training within a certain data volume. Therefore, in embodiments of the present application, the data volume can be indicated by the second information so that the second device can select a reasonable model training for the first device.

[0138] In some implementations, the amount of data available for online training may be determined based on a reference model. For example, the amount of data used by the first device for online training of the reference model may be used as the amount of data available for online training. In other implementations, the amount of data available for online training may be determined based on a reference training complexity. For example, in the case of a reference training complexity, the amount of data used by the first device for online training may be used as the amount of data available for online training.

[0139] In the embodiments of the present application, the value of the data amount is not specifically limited. In some implementations, the data amount can be indicated by N samples. The value of N can include one or more of the following: 1, 2, 5, 10, 20, 50, or 100. In other implementations, the data amount can be indicated by the data size. The value of the data size can be one or more of the following: 100 bytes, 1 kB, or 1 MB.

[0140] As described above, online training varies in duration, training complexity, and data volume. Therefore, online training constraints can be set based on these factors. For example, the first device can perform online training that satisfies the constraints indicated by the second information.

[0141] In some scenarios, online training can be divided into multiple types based on the different restrictions of online training. The restrictions associated with different online training in multiple types of online training may be different, or in other words, the second information associated with different types of online training in multiple types of online training is different. Of course, in the embodiment of the present application, restrictions can also be formulated based on other information. In some implementations, the use conditions of the model can be combined with one or more of the above-mentioned restrictions to form new restrictions to divide the types of online training. Of course, in the embodiment of the present application, the use conditions of the model can also be used alone to specify restrictions.

[0142] In some implementations, the usage conditions of the model can also be referred to as the deployment and usage conditions of the model. The deployment and usage conditions may include one or more of the following: hardware conditions; software conditions and data conditions. Among them, the hardware conditions can be understood as the hardware resource support required for the deployment and use of the model. For example, hardware resource support may include processors, memory, and storage space, etc. Software conditions can be understood as the software environment support required for the deployment and use of the model. For example, the software environment may include operating systems, development tools, and dependent libraries, etc. Data conditions can be understood as the data quality of relevant data (for example, training data, test data) required for the deployment and use of the model, wherein training data is used to train the model, and test data is used to evaluate the performance of the model. The quality of the data has an important impact on the accuracy and effect of the model.

[0143] In the embodiments of the present application, duration, training complexity, and data volume can be used independently as constraints for online training. Of course, the duration, training complexity, and data volume can also be used in combination as constraints for online training. The following describes constraints formed by combining training complexity with duration, and constraints formed by combining training complexity with data volume.

[0144] For example, if the training complexity of online training is reference complexity 1 and the online training can be completed within time t_1, that is, the duration of online training is t_1, then the online training belongs to the first type. If the complexity of online training is reference complexity 1 and the online training can be completed within time t_k, that is, the duration of online training is t_k, then the online training belongs to the second type.

[0145] For another example, if the complexity of online training is reference complexity 2 and can be completed based on a dataset of data size N_1, then the online training belongs to type 3. If the complexity of online training is reference complexity 2 and can be completed based on a dataset of data size N_p, then the online training belongs to type 4.

[0146] In some implementations, the second information may carry information indicating the type of online training to indicate the restrictions on the first device executing the online training. For example, the second information may carry an identifier of the type of online training to indicate the restrictions on the first device executing the online training. Of course, in embodiments of the present application, the second information may also directly carry the aforementioned restrictions.

[0147] For example, if the online training of the first device belongs to the first type described above, the first device may send second information to the second device. In some implementations, the second information may carry a type identifier of the first type, wherein the type identifier of the first type is used to indicate that if the training complexity of the online training is a reference complexity of 1, the duration of online training supported by the first device is t_1. In other implementations, the second information may carry a restriction condition associated with the first type, i.e., if the training complexity of the online training is a reference complexity of 1, the duration of online training supported by the first device is t_1.

[0148] For another example, if the online training of the first device belongs to the third type described above, the first device may send the second information to the second device. In some implementations, the second information may carry a third type of type identifier, wherein the third type of type identifier is used to indicate that if the complexity of the online training is a reference complexity of 2, the first device can complete the online training based on a data set of data volume N_1. In other implementations, the second information may carry a restriction condition associated with the third type, i.e., if the complexity of the online training is a reference complexity of 2, the first device can complete the online training based on a data set of data volume N_1.

[0149] As described above, the restrictions of online training can be determined based on the type of online training. That is, there is a mapping relationship between the restrictions of online training and the type of online training. Accordingly, the restrictions associated with the type of online training in the second information can be determined based on the mapping relationship and the type of online training indicated in the second information.

[0150] In the embodiments of the present application, the manner in which the above-mentioned mapping relationship is obtained is not limited. In some implementations, the mapping relationship may be determined based on predefined information or preconfigured information. In other implementations, the mapping relationship may be defined by the first device and sent to the second device. In other implementations, the mapping relationship may be defined by the second device and sent to the first device.

[0151] In some implementations, if the training type of the model training includes dual-end training, the second information includes one or more of the following: information indicating whether the first device supports dual-end training; information indicating the training device for performing dual-end training; information indicating the training method of dual-end training supported by the first device; information indicating whether the first device supports data transmission required for dual-end training; information indicating whether the first device supports data generation required for dual-end training; information indicating whether the first device supports data reception required for dual-end training; information indicating whether the first device supports data processing required for dual-end training; information indicating whether the first device supports dual-end training initiated by the first device; and information indicating whether the first device supports dual-end training initiated by the second device.

[0152] In an embodiment of the present application, refining the information content of the second information for dual-end training helps the second device determine the dual-end training that the first device can participate in based on the second information, helps avoid the second device instructing the first device to participate in dual-end training that the first device does not support, and improves the rationality of the second device instructing the first device to participate in model training.

[0153] If the second information includes information for indicating whether the first device supports dual-end training, the information can occupy one bit in the second information, which helps reduce the overhead required to transmit the second information. In this case, if the value of the bit is the first value, the bit can be used to indicate that the first device supports dual-end training. If the value of the bit is the second value, the bit can be used to indicate that the first device does not support dual-end training. The first value can be different from the second value. For example, the first value can be 1 and the second value can be 0. For another example, the first value can be 0 and the second value can be 1. Of course, in the embodiment of the present application, the above information can be carried by multiple bits.

[0154] If the second information includes information indicating a training device for dual-end training, in some implementations, the training device includes the first device, meaning the first device can participate in the dual-end training. In other implementations, the training device may include a target device associated with the first device, meaning the target device can participate in the dual-end training on behalf of the first device and transmit the trained model to the first device.

[0155] In the embodiments of the present application, the target device is not limited. For example, the target device may be an external device associated with the first device, wherein the external device may be a server or other device with a model training function.

[0156] If the second information includes information for indicating the training method of dual-end training supported by the first device, the training method of dual-end training includes one or more of the following: multiple devices complete model training for multiple models separately (that is, method two introduced above); multiple devices jointly train multiple models (that is, method three introduced above).

[0157] If the second information includes information indicating whether the first device supports the data transmission required for dual-end training, in some implementations, if dual-end training involves multiple devices simultaneously training multiple models (i.e., Method 3 described above), then the data transmission required for dual-end training includes transmitting one or more of the following data: model parameters required for dual-end training; model gradient data required for dual-end training; and model quantization data required for dual-end training. In this case, the data transmission required for dual-end training can be referred to as the data transmission required for dual-end training in Method 3.

[0158] In other implementations, the devices participating in mode 2 include device 1 and device 2, and the model participating in dual-end training in device 2 may be transmitted from device 1 to device 2. Therefore, in the embodiment of the present application, the data transmission for dual-end training mode 2 may include transmitting the model required for dual-end training. In this case, the data transmission required for dual-end training can be referred to as the data transmission required for dual-end training in mode 2. Of course, in other scenarios, the model participating in dual-end training can also be pre-configured or pre-defined. In this case, the data transmission for dual-end training mode 2 may no longer be transmitted through the second information.

[0159] In an embodiment of the present application, different data transmission capabilities can be reported for different types of dual-end training, which helps to improve the accuracy of the second device in obtaining the capability information of the first device to determine whether dual-end training can be performed with the second device.

[0160] In the embodiment of the present application, the content of the second information can be distinguished not only based on the dual-end training method, but also based on the order in which the devices perform model training in the dual-end training in method 3. The following example illustrates a scenario in which multiple devices complete model training for multiple models.

[0161] In some implementations, the second information includes one or more of the following: information indicating whether the first device supports data reception required for dual-end training; information indicating whether the first device supports data processing required for dual-end training.

[0162] Based on the aforementioned dual-end training method 2, it can be seen that the device being trained later needs to receive data for model training from the device being trained earlier and / or process the data used for model training. In other words, a device with one or more of the above capabilities can serve as a device being trained later in dual-end training method 2. Therefore, in this embodiment of the present application, the first device can send the above second information to the second device so that the second device can determine whether the first device can serve as a device being trained later.

[0163] In some implementations, the second information includes one or more of the following: information indicating whether the first device supports generating data required for dual-end training; information indicating whether the first device supports transmitting data required for dual-end training.

[0164] In some implementations, the data required for the dual-end training may be data required for model training by a later-trained device in dual-end training method 2. For example, the data required for the dual-end training may include a training dataset.

[0165] Based on the aforementioned two-end training method 2, it can be seen that the device that is trained first needs to generate data for model training and send this data to the device that is trained later. In other words, a device that has one or more of the above capabilities can serve as the device that is trained first in two-end training method 2. Therefore, in this embodiment of the present application, the first device can send the above second information to the second device so that the second device can determine whether the first device can serve as the device that is trained first.

[0166] It should be noted that if the first device only supports generating the data required for dual-end training and transmitting the data required for dual-end training, and the first device does not support receiving the data required for dual-end training and processing the data required for dual-end training, then the first device cannot perform the dual-end training in method two, that is, even if the first device obtains the data required for training from other devices, it cannot complete the dual-end training task.

[0167] In some implementations, if dual-end training involves multiple devices independently training multiple models (i.e., the dual-end training described in Method 2 above), the second information includes one or more of the following: information indicating whether dual-end training is supported by the first device; information indicating whether dual-end training is supported by the second device. In other words, the second information may indicate the dual-end training initiation methods supported by the first device, where initiation methods include initiation by the first device and / or initiation by the second device.

[0168] In some implementations, initiating dual-end training by the first device may include the first device being a previously trained device as described above. For example, initiating dual-end training by the first device may include the first device performing model training, i.e., model training performed by the first device in dual-end training. For another example, initiating dual-end training by the first device may include the first device sending data required for dual-end training to the second device so that the second device can perform model training, i.e., model training performed by the second device in dual-end training.

[0169] Based on the above introduction, if the first device supports the first device to initiate dual-end training, it can be understood that the first device supports the capabilities required by the device for online training introduced above, for example, the first device supports generating data required for dual-end training, and / or the first device supports data transmission required for dual-end training.

[0170] In some implementations, initiating dual-end training by the second device may include the second device being a previously trained device as described above. For example, initiating dual-end training by the second device may include the second device performing model training, i.e., model training performed by the second device in dual-end training. For another example, initiating dual-end training by the second device may include the second device sending data required for dual-end training to the first device so that the first device can perform model training, i.e., model training performed by the first device in dual-end training.

[0171] Based on the above introduction, if the second device is supported to initiate dual-end training, it can be understood that the first device supports the capabilities required by the device for subsequent training introduced above, for example, the first device supports the data reception required for dual-end training, and / or the first device supports the data processing required for dual-end training.

[0172] It should be noted that the above only lists the information content of the second information by way of example, and the second information is not specifically limited in the embodiments of the present application. In some implementations, the second information may further indicate whether the first device supports distributed model training, wherein distributed model training refers to a method of performing model training on multiple devices simultaneously. It can speed up model training and can process large-scale data sets. In other implementations, the second information may further indicate whether the first device supports federated learning, wherein federated learning aims to achieve collective learning of the model by training and reasoning on local devices without transferring data from user devices to a central server.

[0173] In some implementations, the dual-end training is one of multiple types of dual-end training, and different types of dual-end training in the multiple types of dual-end training are associated with different second information.

[0174] In some implementations, the dual-end training mode 2 can be divided into multiple types according to the second information corresponding to the device that was trained first and the second information corresponding to the device that was trained later. For example, the dual-end training mode 2 may include type 1 to type 3. Among them, the device corresponding to type 1 can support the information of data reception required for dual-end training and the information of data processing required for dual-end training. The device corresponding to type 2 can support the information of data reception required for dual-end training and the information of data processing required for dual-end training. The device corresponding to type 3 can support the information of data reception required for dual-end training and the information of data processing required for dual-end training.

[0175] Accordingly, for type 1, the second information may include information indicating whether data reception required for dual-end training is supported, and information indicating whether data processing required for dual-end training is supported.

[0176] For type 2, the second information may include information indicating whether data reception required for dual-end training is supported, and information indicating whether data processing required for dual-end training is supported.

[0177] For type 3, the second information may include information for indicating whether data reception required for dual-end training is supported, information for indicating whether data processing required for dual-end training is supported, information for indicating whether data reception required for dual-end training is supported, and information for indicating whether data processing required for dual-end training is supported.

[0178] In other implementations, dual-end training mode 3 can be divided into multiple types based on whether it supports the data transmission required for dual-end training and the training device performing the dual-end training. For example, dual-end training mode 2 can include Type 1 and Type 2. Type 1 corresponds to a device that supports the data transmission required for dual-end training, and the training device is the device itself. Type 2 corresponds to a device that supports the data transmission required for dual-end training, and the training device is the target device associated with the device.

[0179] Accordingly, for type 1, the second information may include data transmission required to indicate support for dual-end training, and the training device is the device itself.

[0180] For type 2, the second information may include data transmission required to indicate support for dual-end training, and the training device is the target device.

[0181] In other implementations, the dual-end training described in Method 2 can be divided into multiple types based on the different dual-end training initiation methods supported by the first device. For example, the dual-end training described in Method 2 can include Types 1 to 3. Type 1 corresponds to a first device that can support dual-end training initiated by the first device itself. Type 2 corresponds to a first device that can support dual-end training initiated by a second device. Type 3 corresponds to a device that can support dual-end training initiated by both the first device and the second device.

[0182] Accordingly, for type 1, the second information may include information indicating that the first device supports dual-end training initiated by the first device. For type 2, the second information may include information indicating that the first device supports dual-end training initiated by the second device. For type 3, the second information may include information indicating that the first device supports dual-end training initiated by the second device, and information indicating that the first device supports dual-end training initiated by the first device.

[0183] In other implementations, the dual-end training described in Method 3 can be divided into multiple types based on the data transmission required for dual-end training supported by the first device and the training device for the dual-end training. For example, the dual-end training described in Method 3 can include Types 1 and 2. Type 1 corresponds to a first device that can support the data transmission required for dual-end training in Method 3, and the training device is the first device itself. Type 2 corresponds to a first device that can support the data transmission required for dual-end training in Method 3, and the training device is the target device.

[0184] Accordingly, for type 1, the second information may include information indicating that the first device supports the data transmission required for dual-end training in mode 3, and information indicating that the training device is the first device itself. For type 2, the second information may include information indicating that the first device supports the data transmission required for dual-end training in mode 3, and information indicating that the training device is the target device.

[0185] Example 1-3: The capability information includes third information.

[0186] In some implementations, the first condition indicated by the third information may be a restriction on the first device performing model training, or in other words, the first device supports model training that satisfies the first condition. Accordingly, the first device may be unable to perform model training that does not satisfy the first condition.

[0187] In some implementations, the first condition can be associated with one or more of the following: the use case type of the model use case; the model type; the model size; the amount of data used for model training; the computing power of the data used for model training; the time used for model training; and the software information used for model training.

[0188] Taking the association of the first condition with the use case type of the model use case as an example, in some implementations, the first condition can be used to indicate that the first device supports a restriction on the use case type for model training. Of course, in embodiments of the present application, the first condition can be used to indicate that the first device does not support a restriction on the use case type for model training.

[0189] The use case type of the above model use case is used to indicate the application scenario and function of the model. For example, the model use case can be used to describe the input and output of the model, as well as the application method and expected results of the model in a specific scenario.

[0190] In some implementations, the use case types of the model use case may include one or more of the following: model use case for CSI feedback; model use case for CSI prediction; model use case for beam management; model use case for beam prediction; model use case for positioning; model use case for channel estimation; model use case for symbol detection; model use case for mobility management; model use case for resource management; model use case for encoding and decoding; model use case for modulation and demodulation; and model use case for waveform adjustment.

[0191] For example, the first condition indicates that the first device supports model training for a model use case of a type that is a model use case for CSI feedback. In other words, the first device can perform model training for the model use case for CSI feedback, but cannot perform model training for other use case types other than the model use case for CSI feedback.

[0192] Taking the association of the first condition with the model type as an example, in some implementations, the first condition can be used to indicate that the first device supports model type restrictions for model training. Of course, in an embodiment of the present application, the first condition can be used to indicate that the first device does not support model type restrictions for model training.

[0193] In some implementations, the model type can be used to indicate a model development framework, where the model development framework is used to provide the tools and libraries required for model building, training, and deployment.

[0194] In some implementations, the model development framework can include one or more of the following: TensorFlow, PyTorch, open neural network exchange (ONNX), Keras, Caffe2, MXNet, ML.NET, Scikit-learn, Tensor, Theano, and computational network toolkit (CNTK).

[0195] For example, the first condition indicates that the development framework supported by the first device for model training is TensorFlow. In other words, the first device can perform model training on models developed based on TensorFlow, but the first device cannot perform model training on models developed based on other development frameworks.

[0196] In other implementations, the model type may be used to indicate a model structure of the model, wherein the model structure is used to indicate the overall architecture and components of the model.

[0197] In some implementations, the model structure may include one or more of the following: a fully connected network, a convolutional neural network, a recurrent neural network, a long short-term memory (LSTM) network structure, a Transformer structure, and a mixer structure.

[0198] For example, the first condition indicates that the model structure supported by the first device for model training is the LSTM structure. In other words, the first device can perform model training on models with the LSTM structure, but the first device cannot perform model training on models with other model structures.

[0199] Taking the association of the first condition with the model size as an example, in some implementations, the first condition can be used to indicate that the first device supports a model size limit for model training. For example, the first condition can be used to indicate an upper limit for the model size that the first device supports for model training. Of course, in an embodiment of the present application, the first condition can be used to indicate that the first device does not support a model size limit for model training. For example, the first condition can be used to indicate that the first device does not support a lower limit for the model size for model training.

[0200] In some implementations, the model size can be indicated by the storage space occupied by the model. For example, the model size can be measured in bytes. Generally, a larger model size means that the model requires more storage space to store and run.

[0201] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate the model size limit supported by the first device for model training, the first condition can indicate one of the following: the model size supported by the first device for model training is 10 bytes; the model size supported by the first device for model training is 100 bytes; the model size supported by the first device for model training is 1kB; the model size supported by the first device for model training is 1MB; the model size supported by the first device for model training is 10MB; the model size supported by the first device for model training is 100MB; the model size supported by the first device for model training is 200MB; the model size supported by the first device for model training is 500MB; the model size supported by the first device for model training is 1GB.

[0202] For example, the first condition may indicate that the model size supported by the first device for model training is 10MB. In this case, the upper limit of the model size supported by the first device for model training is 10MB, which means that the first device supports model training for models within 10MB. For another example, the first condition may indicate that the model size not supported by the first device for model training is 10MB. In this case, the lower limit of the model size not supported by the first device for model training is 10MB, which means that the first device does not support model training for models larger than 10MB.

[0203] In other implementations, the model size may also be indicated by the number of model parameters of the model to be trained. For example, the number of model parameters may be measured in bytes.

[0204] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate a limit on the amount of model parameters of the model that the first device supports for model training, the first condition can indicate one of the following: the amount of model parameters of the model that the first device supports for model training is 5; the amount of model parameters of the model that the first device supports for model training is 10; the amount of model parameters of the model that the first device supports for model training is 20; the amount of model parameters of the model that the first device supports for model training is 50; the amount of model parameters of the model that the first device supports for model training is 100; the amount of model parameters of the model that the first device supports for model training is 200; the amount of model parameters of the model that the first device supports for model training is 500; the amount of model parameters of the model that the first device supports for model training is 1k; the amount of model parameters of the model that the first device supports for model training is 2k; the amount of model parameters of the model that the first device supports for model training is The model parameter amount of the model is 5k; the model parameter amount of the model supported by the first device for model training is 10k; the model parameter amount of the model supported by the first device for model training is 100k; the model parameter amount of the model supported by the first device for model training is 1M; the model parameter amount of the model supported by the first device for model training is 10M; the model parameter amount of the model supported by the first device for model training is 100M; the model parameter amount of the model supported by the first device for model training is 1G; the model parameter amount of the model supported by the first device for model training is 10G; the model parameter amount of the model supported by the first device for model training is 100G; the model parameter amount of the model supported by the first device for model training is 1T; the model parameter amount of the model supported by the first device for model training is 10T; the model parameter amount of the model supported by the first device for model training is 100T.

[0205] For example, the first condition may indicate that the model parameter amount of the model supported by the first device for model training is 10G. In this case, the upper limit of the model parameter amount supported by the first device for model training is 10G, which means that the first device supports model training for models corresponding to model parameter amounts within 10G. For another example, the first condition may indicate that the model parameter amount of the model not supported by the first device for model training is 10G. In this case, the lower limit of the model parameter amount not supported by the first device for model training is 10G, which means that the first device does not support model training for models with model parameter amounts exceeding 10G.

[0206] In some other implementations, the model size may also be indicated by a model size level, where different model size levels correspond to different model sizes. In some implementations, the model size levels may include one or more of the following: a very small model level; a small model level; a medium model level; a large model level; and an extra large model level.

[0207] It should be noted that the above-mentioned model size levels can be divided based on the number of parameters of the model, or the above-mentioned model size levels can be divided based on the size of the storage space occupied by the model.

[0208] In addition, in the embodiment of the present application, the model size level can be determined by one or more of the following methods: pre-definition; pre-configuration; second device configuration; and first device configuration.

[0209] For example, the first condition may indicate that the first device supports model training at a large model size level. In this case, the upper limit of the model size level supported by the first device for model training is the large model level, that is, the first device supports model training for models below the large model level. For another example, the first condition may indicate that the first device does not support model training at a large model size level. In this case, the lower limit of the model size level supported by the first device for model training is the large model level, that is, the first device does not support model training for models above the large model level.

[0210] Taking the association of the first condition with the data amount as an example, in some implementations, the first condition can be used to indicate a data amount limit for data supported by the first device for model training. For example, the first condition can be used to indicate an upper limit on the data amount of data supported by the first device for model training. Of course, in an embodiment of the present application, the first condition can be used to indicate a data amount limit for data not supported by the first device for model training. For example, the first condition can be used to indicate a lower limit on the data amount of data not supported by the first device for model training.

[0211] In some implementations, the data used for model training may include one or more of data in a training set, data in a validation set, and data in a test set. Of course, in the embodiments of the present application, the data used for model training is not specifically limited. For example, the data used for model training may include feature data, where feature data is data obtained by extracting features from input data and is used to represent key features of the input data.

[0212] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate that the first device supports a data volume limit for data used for model training, the first condition can indicate one of the following: the data volume supported by the first device for model training is

[0213] For example, the first condition may indicate that the amount of data supported by the first device for model training is 10MB. In this case, the upper limit of the amount of data supported by the first device for model training is 10MB. In other words, the first device supports model training based on data within 10MB. For another example, the first condition may indicate that the amount of data not supported by the first device for model training is 10MB. In this case, the lower limit of the amount of data not supported by the first device for model training is 10MB. In other words, the first device does not support model training based on data exceeding 10MB.

[0214] In other implementations, the amount of data used for model training can be determined based on the number of samples of the training data. For example, the number of samples can be measured in bytes.

[0215] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate that the first device supports a limit on the number of samples of training data for model training, the first condition can indicate one of the following: the number of samples of training data for model training supported by the first device is 5; the number of samples of training data for model training supported by the first device is 10; the number of samples of training data for model training supported by the first device is 20; the number of samples of training data for model training supported by the first device is 50; the number of samples of training data for model training supported by the first device is 100; the number of samples of training data for model training supported by the first device is 200; the number of samples of training data for model training supported by the first device is 500; the number of samples of training data for model training supported by the first device is 1k; the number of samples of training data for model training supported by the first device is 2k; the number of samples of training data for model training supported by the first device is 500; The number of samples of training data supported for model training is 5k; the number of samples of training data supported by the first device for model training is 10k; the number of samples of training data supported by the first device for model training is 100k; the number of samples of training data supported by the first device for model training is 1M; the number of samples of training data supported by the first device for model training is 10M; the number of samples of training data supported by the first device for model training is 100M; the number of samples of training data supported by the first device for model training is 1G; the number of samples of training data supported by the first device for model training is 10G; the number of samples of training data supported by the first device for model training is 100G; the number of samples of training data supported by the first device for model training is 1T; the number of samples of training data supported by the first device for model training is 10T; the number of samples of training data supported by the first device for model training is 100T.

[0216] For example, the first condition may indicate that the number of samples of training data supported by the first device for model training is 10G. In this case, the upper limit of the number of samples of training data supported by the first device for model training is 10G, which means that the first device supports model training for models with a sample size of less than 10G. For another example, the first condition may indicate that the number of samples of training data not supported by the first device for model training is 10G. In this case, the lower limit of the number of samples of training data not supported by the first device for model training is 10G, which means that the first device does not support model training for models with a sample size of more than 10G.

[0217] In other implementations, the data volume may be indicated by a data volume level, where different data volume levels correspond to different data volumes. In some implementations, the data volume level may include one or more of the following: a very small data volume level; a small data volume level; a medium data volume level; a large data volume level; and an extremely large data volume level.

[0218] It should be noted that the above data volume levels can be divided based on the number of samples, or the above data volume levels can be divided based on the size of the data used to train the model.

[0219] In addition, in the embodiment of the present application, the data volume level can be determined by one or more of the following methods: pre-definition; pre-configuration; second device configuration; and first device configuration.

[0220] For example, the first condition may indicate that the data volume level supported by the first device for model training is a large data volume level. In this case, the upper limit of the data volume level supported by the first device for model training is the large data volume level, that is, the first device supports model training for models below the large data volume level. For another example, the first condition may indicate that the data volume level supported by the first device for model training is not a large model level. In this case, the lower limit of the data volume level supported by the first device for model training is not a large data volume level, that is, the first device does not support model training for models associated with data volumes above the large data volume level.

[0221] Taking the association of the first condition with the computing power used for model training as an example, in some implementations, the first condition can be used to indicate the computing power limit for model training supported by the first device. For example, the first condition can be used to indicate the upper limit of computing power for model training supported by the first device. Of course, in an embodiment of the present application, the first condition can be used to indicate a computing power limit for model training that is not supported by the first device, for example, the first condition can be used to indicate that the first device does not support the lower limit of computing power for model training.

[0222] In some implementations, the computing power used for model training can be used to indicate the computing resources and capabilities required to perform complex computing tasks. For example, the computing power used for model training can include one or more of the following: CPU computing power, GPU computing power, tensor processing power, and network bandwidth and storage capacity.

[0223] In some implementations, the unit of measurement for the computing power used for model training may be floating point operations per second (FLOPS). Of course, in the embodiment of the present application, the unit of measurement for the computing power used for model training may be tera operations per second (TOPS).

[0224] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate the computing power limit supported by the first device for model training, the first condition can indicate one of the following: the computing power supported by the first device for model training is 1K FLOPS; the computing power supported by the first device for model training is 1M FLOPS; the computing power supported by the first device for model training is 10M FLOPS; the computing power supported by the first device for model training is 100M FLOPS; the computing power supported by the first device for model training is 1G FLOPS; the computing power supported by the first device for model training is 10G FLOPS; the computing power supported by the first device for model training is 100G FLOPS; the computing power supported by the first device for model training is 1T FLOPS; the computing power supported by the first device for model training is 10T FLOPS.

[0225] In other implementations, if the first condition can be used to indicate the computing power limit supported by the first device for model training, the first condition may indicate one of the following: the computing power supported by the first device for model training is 1TOPS; the computing power supported by the first device for model training is 2TOPS; the computing power supported by the first device for model training is 5TOPS; the computing power supported by the first device for model training is 10TOPS; the computing power supported by the first device for model training is 20TOPS; the computing power supported by the first device for model training is 50TOPS; the computing power supported by the first device for model training is 100TOPS; the computing power supported by the first device for model training is 200TOPS; the computing power supported by the first device for model training is 500TOPS; the computing power supported by the first device for model training is 1000TOPS.

[0226] For example, the first condition may indicate that the computing power supported by the first device for model training is one billion FLOPS. In this case, the upper limit of the computing power supported by the first device for model training is one billion FLOPS. In other words, the first device supports model training with a computing power of less than one billion FLOPS. For another example, the first condition may indicate that the first device does not support a computing power of one billion FLOPS for model training. In this case, the lower limit of the computing power not supported by the first device for model training is one billion FLOPS. In other words, the first device does not support model training with a computing power greater than one billion FLOPS.

[0227] In other implementations, computing power may also be indicated by a computing power level, where different computing power levels correspond to different computing powers. In some implementations, computing power levels may include one or more of the following: minimal computing power level; small computing power level; medium computing power level; large computing power level; and ultra-large computing power level.

[0228] It should be noted that the above computing power levels can be divided based on computing power. In addition, in the embodiments of the present application, the computing power level can be determined by one or more of the following methods: pre-definition; pre-configuration; second device configuration; and first device configuration.

[0229] For example, the first condition may indicate that the computing power level supported by the first device is the high computing power level. In this case, the upper limit of the computing power level supported by the first device for model training is the high computing power level. In other words, the first device supports model training for models below the high computing power level. For another example, the first condition may indicate that the computing power level not supported by the first device is the high computing power level. In this case, the lower limit of the computing power level not supported by the first device is the high computing power level. In other words, the first device does not support model training for models associated with computing power above the high computing power level.

[0230] In other implementations, computing power can also be indicated by model computational complexity. The unit of measurement for model computational complexity can be FLOPS. Of course, in the embodiment of the present application, the unit of measurement for model computational complexity can be TOPS.

[0231] In the embodiments of the present application, the first condition is not specifically limited. In some implementations, if the first condition can be used to indicate the computing power limit supported by the first device for model training, the first condition can indicate one of the following: the model computing complexity supported by the first device is 1K FLOPS; the model computing complexity supported by the first device is 10M FLOPS; the model computing complexity supported by the first device is 100M FLOPS; the model computing complexity supported by the first device is 1G FLOPS; the model computing complexity supported by the first device is 10G FLOPS; the model computing complexity supported by the first device is 100G FLOPS; the model computing complexity supported by the first device is 1T FLOPS; the model computing complexity supported by the first device is 10T FLOPS.

[0232] For example, the first condition may indicate that the first device supports a model computation complexity of one billion FLOPS. In this case, the upper limit of the model computation complexity supported by the first device is one billion FLOPS. In other words, the first device supports model training with a required model computation complexity of less than one billion FLOPS. For another example, the first condition may indicate that the first device does not support a model computation complexity of one billion FLOPS. In this case, the lower limit of the model computation complexity not supported by the first device is one billion FLOPS. In other words, the first device does not support model training with a required model computation complexity greater than one billion FLOPS.

[0233] In other implementations, the computing power may also be indicated by the model calculation complexity level, where different model calculation complexity levels correspond to different computing powers.

[0234] In some implementations, the model computational complexity level may include one or more of the following: an extremely small model computational complexity level; a small model computational complexity level; a medium model computational complexity level; a large model computational complexity level; and an extremely large model computational complexity level.

[0235] It should be noted that the above model calculation complexity levels can be divided based on model calculation complexity. In addition, in the embodiment of the present application, the model calculation complexity level can be determined by one or more of the following methods: pre-definition; pre-configuration; second device configuration; and first device configuration.

[0236] For example, the first condition may indicate that the model calculation complexity level supported by the first device is the large model calculation complexity level. In this case, the upper limit of the model calculation complexity level supported by the first device for model training is the large model calculation complexity level. In other words, the first device supports model training for models below the large model calculation complexity level. For another example, the first condition may indicate that the model calculation complexity level not supported by the first device is the large model calculation complexity level. In this case, the lower limit of the model calculation complexity level not supported by the first device is the large model calculation complexity level. In other words, the first device does not support model training for models associated with model calculation complexities above the large model calculation complexity level.

[0237] Taking the association of the first condition with the time for model training as an example, in some implementations, the first condition can be used to indicate a time limit for the first device to be used for model training. For example, the first condition can be used to indicate an upper limit on the time available for model training supported by the first device. Of course, in an embodiment of the present application, the first condition can be used to indicate a time limit for model training that is not supported by the first device, for example, the first condition can be used to indicate that the first device does not support a lower limit on the time for model training.

[0238] In some implementations, if the first condition is used to indicate a time limit on the first device being available for model training, the time limit is used to indicate one or more of the following: a duration supported by the first device for model training; a time period supported by the first device for model training; or a time period supported by the first device for model training.

[0239] In some implementations, the above-mentioned time limit is used to indicate the duration of time supported by the first device for model training. It can be understood that the time limit is used to indicate the duration of time that the first device can participate in model training.

[0240] For example, the first condition may indicate that the duration of model training supported by the first device is 30 seconds. In this case, the upper limit of the duration of model training supported by the first device is 30 seconds. In other words, the first device supports model training with a duration of less than 30 seconds. For another example, the first condition may indicate that the first device does not support a duration of model training of 30 seconds. In this case, the lower limit of the time for model training not supported by the first device is 30 seconds. In other words, the first device does not support model training with a duration of more than 30 seconds.

[0241] In some implementations, the above-mentioned time limit is used to indicate the time period supported by the first device and available for model training. It can be understood that the time limit is used to indicate the time period in which the first device can participate in model training.

[0242] For example, the usage rate of the first device is usually low between 2:00 and 3:00. In this case, the first condition may indicate that the time period for model training supported by the first device is from 2:00 to 3:00. In this case, the first device may participate in the model training between 2:00 and 3:00. For another example, the usage rate of the first device is usually high between 8:00 and 20:00. In this case, the first condition may indicate that the first device does not support the time period for model training from 8:00 to 20:00. In this case, the first device cannot participate in the model training between 8:00 and 20:00.

[0243] In some implementations, the time limit is used to indicate a time period supported by the first device that can be used for model training, and the time limit may include parameters for determining the time period. For example, the time limit includes one or more of the following parameters: a start time of the time period, a number of periods in the time period, a length of the time period, and a time period within the time period that can be used for model training.

[0244] In some implementations, the time period available for model training within a time period can be determined based on a time offset value, where the time offset value is the time offset between a reference time and the start time of the time period available for model training. In the embodiments of the present application, the reference time is not limited. For example, the reference time can be the start time of the time period. For another example, the reference time can be the end time of the time period. For another example, the reference time can be a certain time within the time period.

[0245] Taking the association of the first condition with the software for model training supported by the first device as an example, in some implementations, the first condition can be used to indicate the software limitations of the model training supported by the first device. For example, the first condition can be used to indicate the software information that can be used for model training supported by the first device. Of course, in an embodiment of the present application, the first condition can be used to indicate software limitations for model training that are not supported by the first device. For example, the first condition can be used to indicate that the first device does not support software information for model training.

[0246] In some implementations, the software information may include one or more of the following: the model's software version; the model's development environment; the model's runtime environment; and the model's deployment tools. The model's development environment may include the software development environment used during model development, such as the programming language and programming framework. Model deployment tools may include model compilation tools, compilation format conversion tools, and model optimization tools.

[0247] For example, the first condition may indicate that the software for model training supported by the first device is limited to models developed based on model development environment 1. In this case, the models supported for model training by the first device are models developed based on model development environment 1. In other words, the first device can perform model training on models developed based on model development environment 1. For another example, the first condition may indicate that the software for model training not supported by the first device is limited to models developed based on model development environment 1. In this case, the models not supported for model training by the first device are models developed based on model development environment 1. In other words, the first device cannot perform model training on models developed based on model development environment 1.

[0248] In the embodiments of the present application, the carrying method of the capability information introduced in the above Examples 1-1 to 1-3 is not limited. In some implementations, the capability information introduced in Examples 1-1 to 1-3 can be indicated by means of a bitmap. For example, different bits in a bitmap with a length of W2 bits can represent different capability information. For example, the first m1 bits in the W2-bit bitmap correspond to the capability information of Example 1-1, the last m2 bits in the W2-bit bitmap correspond to the capability information of Example 1-2, and the remaining bits in the W2-bit bitmap correspond to the capability information of Example 1-3. Of course, in the embodiments of the present application, different indication information can be used to indicate the capability information introduced in Examples 1-1 to 1-3.

[0249] Example 2: Capability information is associated with model compilation.

[0250] Typically, before a model is deployed to hardware, it needs to be compiled into an instruction set suitable for a specific hardware platform. This typically involves converting a model representation in a high-level language to a low-level instruction set supported by the underlying hardware. Currently, not all devices support the above compilation, or some devices only support conversion of some hardware instruction sets. Therefore, in an embodiment of the present application, the first device can interact with the second device through the above capability information to indicate the capabilities of the first device associated with model compilation, which helps to increase the possibility of the first device deploying the model.

[0251] In some implementations, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

[0252] In some implementations, capability information may occupy one bit, which helps reduce the overhead required to transmit capability information. In this case, if the value of the bit is a first value, the bit may be used to indicate that the first device supports compiling the model into a hardware instruction set. If the value of the bit is a second value, the bit may be used to indicate that the first device does not support compiling the model into a hardware instruction set. The first value may be different from the second value, for example, the first value may be 1 and the second value may be 0. For another example, the first value may be 0 and the second value may be 1. Of course, in an embodiment of the present application, capability information may be carried by multiple bits.

[0253] In some other implementations, the capability information is used to indicate a hardware instruction set supported by the first device for model compilation. For example, the capability information may include an identifier of the hardware instruction set supported by the first device.

[0254] In the embodiments of the present application, the hardware instruction set is not specifically limited. For example, the hardware instruction set may include the Compute Unified Device Architecture (CUDA). For another example, the hardware instruction set may include the Open Computing Language (OpenCL). For another example, the hardware instruction set may include Vulkan.

[0255] For example, the hardware instruction set may include CUDA, OpenCL, and Vulkan, and the first device only supports model compilation based on CUDA. In this case, the capability information may carry a CUDA identifier to indicate that the hardware instruction set supported by the first device for model compilation is CUDA.

[0256] In some other implementations, the capability information is used to indicate that the first device does not support a hardware instruction set for model compilation. For example, the capability information may include an identifier of a hardware instruction set that the first device does not support.

[0257] For example, the hardware instruction set may include CUDA, OpenCL, and Vulkan, and the first device does not support model compilation based on OpenCL. In this case, the capability information may carry an OpenCL identifier to indicate that the hardware instruction set that the first device does not support for model compilation is CUDA.

[0258] It should be noted that the information indicating whether the first device supports compiling the model into a hardware instruction set, the information indicating the hardware instruction set that the first device supports for model compilation, and the information indicating that the first device does not support the hardware instruction set for model compilation can be independent of each other. Of course, in the embodiments of the present application, the above three types of information can also be used in combination with each other.

[0259] For example, if the capability information is used to indicate that the first device supports a hardware instruction set for model compilation, then the hardware instruction set used to indicate that the first device supports model compilation may implicitly indicate that the first device supports compiling the model into a hardware instruction set. To save the overhead of transmitting the capability information, the capability information may not require additional bits to carry the information indicating that the first device supports compiling the model into a hardware instruction set.

[0260] For another example, if the capability information indicates that the first device does not support a hardware instruction set for model compilation, and the hardware instruction set not supported by the first device is not all predefined hardware instruction sets, then the information indicating that the first device does not support the hardware instruction set for model compilation may implicitly indicate that the first device supports compiling the model into a hardware instruction set. To save the overhead of transmitting capability information, the capability information may not require additional bits to carry the information indicating that the first device supports compiling the model into a hardware instruction set.

[0261] Example 3: Capability information is associated with model deployment.

[0262] Typically, the model needs to be deployed on hardware for reasoning or training. This process can be achieved by loading the model onto the hardware device based on a software library or software framework. After the model is deployed to the hardware, the computing power of the hardware can be used for real-time reasoning or training. However, not all devices support the above-mentioned model deployment process. If this is not distinguished, the model may be transmitted to a device that does not support model deployment. At this time, the model transmission process becomes meaningless and will take up a lot of transmission resources. Therefore, in an embodiment of the present application, the first device can interact with the second device through the above-mentioned capability information to indicate the capability of the first device to associate with the model deployment, which helps to improve the rationality of model transmission.

[0263] In some implementations, the capability information is used to indicate whether the first device supports deploying the model to the hardware of the first device.

[0264] In some implementations, capability information may occupy one bit, which helps reduce the overhead required to transmit capability information. At this time, if the value of the bit is a first value, the bit can be used to indicate that the first device supports deploying the model to the hardware of the first device. If the value of the bit is a second value, the bit can be used to indicate that the first device does not support deploying the model to the hardware of the first device. The first value may be different from the second value, for example, the first value may be 1 and the second value may be 0. For another example, the first value may be 0 and the second value may be 1. Of course, in an embodiment of the present application, capability information may be carried by multiple bits.

[0265] In the embodiments of the present application, the carrying manner of the capability information introduced in Examples 1 to 3 above is not limited. In some implementations, the capability information introduced in Examples 1 to 3 may be indicated by means of a bitmap. For example, different bits in a bitmap with a length of W1 bits may represent different capability information. For example, the first n1 bits in the W1-bit bitmap correspond to the capability information of Example 1, the last n2 bits in the W1-bit bitmap correspond to the capability information of Example 2, and the remaining bits in the W1-bit bitmap correspond to the capability information of Example 3. Of course, in the embodiments of the present application, different indication information may be used to indicate the capability information introduced in Examples 1 to 3.

[0266] The above describes the capability information in the embodiment of the present application, and the following describes the transmission method of the capability information in the embodiment of the present application. The following describes the transmission methods 1 to 3.

[0267] Transmission mode 1: the reporting of the capability information may be autonomously triggered by the first device.

[0268] That is, the first device autonomously triggers the sending of the capability information of the first device to the second device. In the embodiment of the present application, the first device autonomously triggers the sending of the capability information, which helps to reduce the signaling overhead between the first device and the second device and simplifies the process of capability information reporting.

[0269] Transmission mode 2: the reporting of the capability information may be triggered by the second device.

[0270] 9 , before step S810 , the method further includes: S910 , the second device sends fourth information to the first device to trigger reporting of capability information. Accordingly, step S810 includes: in response to receiving the fourth information, the first device sends capability information to the second device.

[0271] In some implementations, the fourth information is used to instruct the first device to send capability information. In other implementations, the fourth information is used to request the first device to send capability information.

[0272] In an embodiment of the present application, the second device may trigger reporting of capability information, which helps to avoid unnecessary transmission of capability information and thus save transmission overhead of capability information.

[0273] Transmission mode 3: the information content carried in the capability information is indicated by the second device.

[0274] In some scenarios, the second device may only require partial capability information. In this case, transmitting all capability information from the first device may result in excessive transmission overhead. For example, the second device may be able to complete model training on its own and does not require the first device to have model training capabilities. In this case, reporting capability information associated with model training is unnecessary.

[0275] Therefore, in response to the above problem, in an embodiment of the present application, the second device can indicate the information content carried in the capability information by sending fourth information to the first device, which helps to reduce the overhead of transmitting the capability information.

[0276] As shown in FIG. 10 , before step S810, the method further includes: S1010, the second device sending fourth information to the first device, where the fourth information includes information indicating the content of information carried in the capability information sent by the first device. Accordingly, step S810 may include the first device sending capability information to the second device, where the content carried in the capability information matches the content of information carried in the capability information indicated in the fourth information.

[0277] In the embodiments of the present application, the specific manner of indicating the information content carried in the capability information is not limited. In some implementations, the fourth information may carry a target task, where the target task is associated with some or all of the information in the capability information. In this case, the first device may carry some or all of the information associated with the target task in the capability information.

[0278] For example, the target task carried in the fourth information is model training. In this case, the first device may carry capability information associated with model training in the capability information.

[0279] For another example, the target task carried in the fourth information is dual-end model training. In this case, the first device may carry second information associated with the dual-end model training in the capability information.

[0280] In other implementations, the fourth information may directly carry the index of the information content of the capability information that the capability information needs to carry. For example, if the index of the capability information associated with model training is 01, the index of the capability information associated with model deployment is 02, and the index of the capability information associated with model compilation is 03, then the fourth information may carry 01 to indicate that the information content carried in the capability information is capability information associated with model training.

[0281] It should be noted that, in some other implementations, the fourth information may also indicate information that does not need to be carried in the capability information sent by the first device, which may reduce the overhead of transmitting the capability information to a certain extent.

[0282] It should also be noted that the fourth information in implementation method 2 can be used independently of the fourth information in implementation method 3, or the fourth information in implementation method 2 can be used in combination with the fourth information in implementation method 3. At this time, if the fourth information contains information for indicating the information content carried in the capability information, then the information can implicitly indicate the triggering of the first device to report the capability information (i.e., the meaning of the fourth information in implementation method 2). At this time, no additional bit in the fourth information may be required to trigger the reporting of the capability information, which helps to save the transmission overhead of the fourth information. Of course, if this issue is not considered in the same way, the fourth information in implementation method 2 and the fourth information in implementation method 3 can be indicated by different bits.

[0283] In the embodiments of the present application, the first device and / or the second device are not limited. For example, the first device is a terminal device, and the second device is a network device. In another example, the first device is a network device, and the second device is a terminal device. In another example, the first device is a first terminal device, and the second device is a second terminal device. In another example, the first device is a first network device, and the second device is a second network device.

[0284] The network device is an access network device, a core network device, an AI / ML model-related information management device, or an operation administration and maintenance (OAM) device. Exemplarily, the access network device is any one of the following: a gNB, a centralized unit (CU), a distributed unit (DU), a centralized unit-control plane (CU-CP), or a centralized unit-user plane (CU-UP).

[0285] Exemplarily, the core network device is any one of the following: location management function (LMF) network element, network slice selection function (NSSF), authentication server function (AUSF), unified data management (UDM), access and mobility management function (AMF), session management function (SMF), policy control function (PCF), user plane function (UPF), sensing control function (SF), network data analysis (NWDAF) network element.

[0286] In some scenarios, any of the above information may include uplink information. For example, when the first device is a terminal device and the second device is a network device, the capability information may be uplink information. For another example, when the second device is a terminal device and the first device is a network device, the fourth information may be uplink information.

[0287] In some implementations, if the above information is uplink information, the information can be carried by one or more of the following: radio resource control (RRC) message, uplink control information (UCI) message, uplink message in the random access process (for example, message A (MsgA) and / or message 3 (Msg3)), physical uplink control channel (PUCCH), physical uplink shared channel (PUSCH), uplink channel dedicated to AI / ML, and UE capabilities.

[0288] For example, the first device is a UE and the second device is a network device, and the UE sends UE capabilities to the network device, where the UE capabilities carry the above capability information.

[0289] For another example, the first device is a terminal device and the second device is a network device. During the random access process, the terminal device sends an uplink message (e.g., MsgA and / or Msg3) to the network device. The uplink message may carry capability information, which helps to quickly transmit the capability information to the network device.

[0290] For another example, the first device is a terminal device and the second device is a network device. The terminal device sends UCI to the network device. The UCI can carry capability information, which helps to transmit the capability information to the network device more quickly.

[0291] For another example, the first device is a terminal device, the second device is a network device, and the terminal device sends an RRC message to the network device. The RRC message can carry capability information, which helps avoid occupying too many capability reporting resources and control resources such as UCI.

[0292] In other scenarios, any of the above information may include downlink information. For example, when the first device is a network device and the second device is a terminal device, the capability information may be downlink information. For another example, when the second device is a network device and the first device is a terminal device, the fourth information may be downlink information.

[0293] In some implementations, if the above information is downlink information, the information can be carried in one or more of the following: broadcast messages (e.g., MIB, SIB1, SIBx); RRC messages; media access control control element (MAC CE); downlink control information (DCI); downlink messages in the random access process (e.g., one or more of MsgB, Msg2, Msg4); physical downlink control channel (PDCCH), physical downlink shared channel (PDSCH); downlink channels dedicated to AI / ML; and network device capabilities.

[0294] For example, the first device is a network device and the second device is a terminal device, and the network device sends network device capabilities to the terminal device, where the network device capabilities carry the above capability information.

[0295] For another example, the first device is a network device and the second device is a terminal device. During the random access process, the network device sends a downlink message to the terminal device. The downlink message can carry capability information, which helps to transmit the capability information to the terminal device more quickly.

[0296] For another example, the first device is a network device and the second device is a terminal device. The network device sends a DCI to the terminal device. The DCI can carry capability information, which helps to transmit the capability information to the terminal device more quickly.

[0297] For another example, the first device is a network device and the second device is a terminal device. The network device sends an RRC message to the terminal device. The RRC message can carry capability information, which helps avoid occupying too many capability reporting resources and control resources such as DCI.

[0298] In some scenarios, any of the above information may include side information. For example, when the first device and the second device are both terminal devices, the capability information and the fourth information may be side information.

[0299] In some implementations, if the above information is sidelink information, the information can be carried on one or more of the following: sidelink control information (SCI); physical sidelink control channel (PSCCH), physical sidelink shared channel (PSSCH); UE capabilities.

[0300] The method embodiment of the present application is described in detail above in conjunction with Figures 1 to 10. The device embodiment of the present application is described in detail below in conjunction with Figures 11 to 13. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment. Therefore, for parts not described in detail, reference can be made to the above method embodiment.

[0301] FIG11 is a schematic diagram of a communication device according to an embodiment of the present application. The communication device 1100 shown in FIG11 is a first device, and the communication device 1100 includes a sending unit 1110 .

[0302] The sending unit 1110 is used to send capability information of the first device to the second device, where the capability information is associated with one or more of the following: model training; model compilation; model deployment.

[0303] In some implementations, if the capability information is associated with the model training, the capability information includes one or more of the following: first information for indicating whether the first device supports the model training; second information for indicating the training type of the model training supported by the first device; and third information for indicating a first condition, where the first condition is a restriction condition for the first device to perform the model training.

[0304] In some implementations, the training type of the model training includes online training, and the second information also includes one or more of the following: information indicating whether the first device supports the online training; information indicating the duration of the online training supported by the first device; information indicating the training complexity of the online training supported by the first device; information indicating the amount of data supported by the first device for the online training.

[0305] In some implementations, the online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

[0306] In some implementations, the training type of the model training includes dual-end training, and the second information includes one or more of the following: information for indicating the training device for performing the dual-end training; information for indicating whether the first device supports the dual-end training; information for indicating the training method of the dual-end training supported by the first device; information for indicating whether the first device supports the dual-end training initiated by the first device; information for indicating whether the first device supports the dual-end training initiated by the second device; information for indicating whether the first device supports the data transmission required for the dual-end training; information for indicating whether the first device supports the data generation required for the dual-end training; information for indicating whether the first device supports the data reception required for the dual-end training; information for indicating whether the first device supports the data processing required for the dual-end training.

[0307] In some implementations, if the second information includes information for indicating the training method of the dual-end training supported by the first device, the training method of the dual-end training includes one or more of the following: multiple devices complete model training for multiple models separately; multiple devices complete model training for multiple models simultaneously.

[0308] In some implementations, if the second information includes information indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

[0309] In some implementations, if the second information includes information indicating whether the first device supports the data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, then the data transmission required for the dual-end training includes transmitting one or more of the following data: model parameters required for the dual-end training; model gradient data required for the dual-end training; and model quantization data required for the dual-end training.

[0310] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models separately, the second information includes one or more of the following: information for indicating whether the first device supports data reception required for the dual-end training; information for indicating whether the first device supports data processing required for the dual-end training.

[0311] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models respectively, the second information includes one or more of the following: information for indicating whether the first device supports generating the data required for the dual-end training; information for indicating whether the first device supports transmitting the data required for the dual-end training.

[0312] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models respectively, the second information includes one or more of the following: information for indicating whether the dual-end training is supported to be initiated by the first device; information for indicating whether the dual-end training is supported to be initiated by the second device.

[0313] In some implementations, the dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

[0314] In some implementations, the capability information includes the third information, and the first condition is used to indicate one or more of the following: a use case type restriction of the model use case supported by the first device for the model training; a model type restriction supported by the first device for the model training; a model size restriction supported by the first device for the model training; a data volume restriction supported by the first device for the model training; a computing power restriction supported by the first device for the model training; a time restriction that the first device can be used for the model training; and a software restriction supported by the first device for the model training.

[0315] In some implementations, if the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

[0316] In some implementations, if the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deploying the model to hardware of the first device.

[0317] In some implementations, the communication device further includes: a receiving unit for receiving fourth information sent by the second device, wherein the fourth information includes one or more of the following: information for indicating that the first device sends the capability information; information for indicating the information content carried in the capability information sent by the first device.

[0318] FIG12 is a schematic diagram of a communication device according to another embodiment of the present application. The communication device 1200 shown in FIG12 is a second device. The communication device 1200 includes a receiving unit 1210.

[0319] The receiving unit 1210 is used to receive capability information of the first device sent by the first device, where the capability information is associated with one or more of the following: model training; model compilation; and model deployment.

[0320] In some implementations, if the capability information is associated with the model training, the capability information includes one or more of the following: first information for indicating whether the first device supports the model training; second information for indicating the training type of the model training supported by the first device; and third information for indicating a first condition, where the first condition is a restriction condition for the first device to perform the model training.

[0321] In some implementations, the training type of the model training includes online training, and the second information also includes one or more of the following: information indicating whether the first device supports the online training; information indicating the duration of the online training supported by the first device; information indicating the training complexity of the online training supported by the first device; information indicating the amount of data supported by the first device for the online training.

[0322] In some implementations, the online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

[0323] In some implementations, the training type of the model training includes dual-end training, and the second information includes one or more of the following: information for indicating the training device for performing the dual-end training; information for indicating whether the first device supports the dual-end training; information for indicating the training method of the dual-end training supported by the first device; information for indicating whether the first device supports the dual-end training initiated by the first device; information for indicating whether the first device supports the dual-end training initiated by the second device; information for indicating whether the first device supports the data transmission required for the dual-end training; information for indicating whether the first device supports the data generation required for the dual-end training; information for indicating whether the first device supports the data reception required for the dual-end training; information for indicating whether the first device supports the data processing required for the dual-end training.

[0324] In some implementations, if the second information includes information for indicating the training method of the dual-end training supported by the first device, the training method of the dual-end training includes one or more of the following: multiple devices complete model training for multiple models separately; multiple devices complete model training for multiple models simultaneously.

[0325] In some implementations, if the second information includes information indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

[0326] In some implementations, if the second information includes information indicating whether the first device supports the data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, then the data transmission required for the dual-end training includes transmitting one or more of the following data: model parameters required for the dual-end training; model gradient data required for the dual-end training; and model quantization data required for the dual-end training.

[0327] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models separately, the second information includes one or more of the following: information for indicating whether the first device supports data reception required for the dual-end training; information for indicating whether the first device supports data processing required for the dual-end training.

[0328] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models respectively, the second information includes one or more of the following: information for indicating whether the first device supports generating the data required for the dual-end training; information for indicating whether the first device supports transmitting the data required for the dual-end training.

[0329] In some implementations, if the dual-end training is for multiple devices to complete model training for multiple models respectively, the second information includes one or more of the following: information for indicating whether the dual-end training is supported to be initiated by the first device; information for indicating whether the dual-end training is supported to be initiated by the second device.

[0330] In some implementations, the dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

[0331] In some implementations, the capability information includes the third information, and the first condition is used to indicate one or more of the following: a use case type restriction of the model use case supported by the first device for the model training; a model type restriction supported by the first device for the model training; a model size restriction supported by the first device for the model training; a data volume restriction supported by the first device for the model training; a computing power restriction supported by the first device for the model training; a time restriction that the first device can be used for the model training; and a software restriction supported by the first device for the model training.

[0332] In some implementations, if the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

[0333] In some implementations, if the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deploying the model to hardware of the first device.

[0334] In some implementations, the communication device further includes: a sending unit for sending fourth information to the first device, wherein the fourth information includes one or more of the following: information for indicating that the first device sends the capability information; information for indicating the information content carried in the capability information sent by the first device.

[0335] In an optional embodiment, the sending unit 1110 may be a transceiver 1330. The communication device 1100 may further include a processor 1310 and a memory 1320, as specifically shown in FIG13 .

[0336] In an optional embodiment, the receiving unit 1210 may be a transceiver 1330. The communication device 1200 may further include a processor 1330 and a memory 1320, as specifically shown in FIG13 .

[0337] Figure 13 is a schematic block diagram of a communication device according to an embodiment of the present application. The dashed lines in Figure 13 indicate that the unit or module is optional. Apparatus 1300 may be used to implement the method described in the above method embodiment. Apparatus 1300 may be a chip, a terminal device, or a network device.

[0338] The device 1300 may include one or more processors 1310. The processor 1310 may support the device 1300 to implement the method described in the above method embodiment. The processor 1310 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0339] The apparatus 1300 may further include one or more memories 1320. The memories 1320 store programs that can be executed by the processor 1310, causing the processor 1310 to perform the methods described in the above method embodiments. The memories 1320 may be independent of the processor 1310 or integrated into the processor 1310.

[0340] The apparatus 1300 may further include a transceiver 1330. The processor 1310 may communicate with other devices or chips via the transceiver 1330. For example, the processor 1310 may transmit and receive data with other devices or chips via the transceiver 1330.

[0341] The present application also provides a computer-readable storage medium for storing a program. The computer-readable storage medium can be applied to a terminal or network device provided in the present application, and the program enables a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0342] The present application also provides a computer program product. The computer program product includes a program. The computer program product can be applied to a terminal or network device provided in the present application, and the program causes a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0343] The embodiments of the present application also provide a computer program. The computer program can be applied to the terminal or network device provided in the embodiments of the present application, and the computer program enables a computer to execute the method performed by the terminal or network device in each embodiment of the present application.

[0344] It should be understood that the terms "system" and "network" in this application can be used interchangeably. In addition, the terms used in this application are only used to explain the specific embodiments of this application and are not intended to limit this application. The terms "first", "second", "third", and "fourth" in the specification and claims of this application and the accompanying drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions.

[0345] In the embodiments of this application, the term "indication" may refer to a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" may refer to a direct indication of B, e.g., B can obtain information through A; it may refer to an indirect indication of B, e.g., A indicates C, e.g., B can obtain information through C; or it may refer to an association between A and B.

[0346] In the embodiment of the present application, "B corresponding to A" means that B is associated with A and B can be determined based on A. However, it should be understood that determining B based on A does not mean determining B based solely on A, but B can also be determined based on A and / or other information.

[0347] In the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and indication, configuration and configuration, etc.

[0348] In the embodiments of the present application, "pre-definition" or "pre-configuration" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., a terminal device and a network device). The present application does not limit the specific implementation method. For example, pre-definition may refer to information defined in a protocol.

[0349] In the embodiments of the present application, the “protocol” may refer to a standard protocol in the communications field, for example, it may include an LTE protocol, an NR protocol, and related protocols used in future communication systems, and the present application does not limit this.

[0350] In the embodiments of this application, the term "and / or" is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0351] In various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0352] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0353] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0354] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0355] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be read by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a digital versatile disc (DVD)), or a semiconductor medium (eg, a solid state disk (SSD)).

[0356] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A wireless communication method, characterized in that: include: A first device sends capability information of the first device to a second device, where the capability information is associated with one or more of the following: Model training; Model compilation; Model deployment.

2. The method according to claim 1, characterized in that If the capability information is associated with the model training, the capability information includes one or more of the following: First information used to indicate whether the first device supports the model training; second information for indicating a training type of the model training supported by the first device; The third information is used to indicate a first condition, where the first condition is a restriction condition for the first device to perform the model training.

3. The method according to claim 2, characterized in that The training type of the model training includes online training, and the second information further includes one or more of the following: Information used to indicate whether the first device supports the online training; Information used to indicate the training complexity of the online training supported by the first device; information indicating a duration supported by the first device and available for the online training; Information indicating a data volume supported by the first device and available for the online training.

4. The method according to claim 3, characterized in that The online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

5. The method according to claim 2, characterized in that The training type of the model training includes dual-end training, and the second information includes one or more of the following: Information for indicating a training device for performing the dual-end training; Information used to indicate whether the first device supports the dual-end training; Information used to indicate a training mode of the dual-end training supported by the first device; used to indicate whether the first device supports initiating the dual-end training by the first device; used to indicate whether the first device supports the dual-end training initiated by the second device; Information used to indicate whether the first device supports data transmission required for the dual-end training; Information used to indicate whether the first device supports data generation required for the dual-end training; Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

6. The method according to claim 5, characterized in that If the second information includes information for indicating a training mode of the dual-end training supported by the first device, the training mode of the dual-end training includes one or more of the following: Multiple devices complete model training for multiple models respectively; Multiple devices complete model training for multiple models simultaneously.

7. The method according to claim 5, characterized in that If the second information includes information for indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

8. The method according to claim 5, characterized in that If the second information includes information for indicating whether the first device supports data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, the data transmission required for the dual-end training includes transmitting one or more of the following data: Model parameters required for the dual-end training; Model gradient data required for the dual-end training; The quantization data of the model required for the dual-end training.

9. The method according to any one of claims 5 to 8, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

10. The method according to any one of claims 5 to 8, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports generating data required for the dual-end training; Information used to indicate whether the first device supports transmission of data required for the dual-end training.

11. The method according to any one of claims 5 to 8, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the dual-end training is supported to be initiated by the first device; Information used to indicate whether the dual-end training is supported to be initiated by the second device.

12. The method according to any one of claims 5 to 11, characterized in that The dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

13. The method according to any one of claims 2 to 12, characterized in that The capability information includes the third information, and the first condition is used to indicate one or more of the following: The first device supports use case type restrictions on model use cases for performing the model training; The first device supports restrictions on model types for performing the model training; The first device supports a model size limit for performing the model training; a data volume limit of data supported by the first device for the model training; A computing power limit supported by the first device for training the model; a time limit during which the first device can be used for training the model; The software limitations of the model training supported by the first device.

14. The method according to any one of claims 1 to 13, characterized in that If the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

15. The method according to any one of claims 1 to 14, characterized in that If the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deployment of the model to hardware of the first device.

16. The method according to any one of claims 1 to 15, characterized in that The method further comprises: The first device receives fourth information sent by the second device, where the fourth information includes one or more of the following: Information used to instruct the first device to send the capability information; Information used to indicate information content carried in the capability information sent by the first device.

17. A wireless communication method, characterized in that: include: The second device receives capability information of the first device sent by the first device, where the capability information is associated with one or more of the following: Model training; Model compilation; Model deployment.

18. The method according to claim 17, characterized in that If the capability information is associated with the model training, the capability information includes one or more of the following: First information used to indicate whether the first device supports the model training; second information for indicating a training type of the model training supported by the first device; The third information is used to indicate a first condition, where the first condition is a restriction condition for the first device to perform the model training.

19. The method according to claim 18, characterized in that The training type of the model training includes online training, and the second information further includes one or more of the following: Information used to indicate whether the first device supports the online training; information indicating a duration supported by the first device and available for the online training; Information used to indicate the training complexity of the online training supported by the first device; Information indicating a data volume supported by the first device and available for the online training.

20. The method of claim 19, wherein: The online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

21. The method of claim 18, wherein: The training type of the model training includes dual-end training, and the second information includes one or more of the following: Information for indicating a training device for performing the dual-end training; Information used to indicate whether the first device supports the dual-end training; Information used to indicate a training mode of the dual-end training supported by the first device; used to indicate whether the first device supports initiating the dual-end training by the first device; used to indicate whether the first device supports the dual-end training initiated by the second device; Information used to indicate whether the first device supports data transmission required for the dual-end training; Information used to indicate whether the first device supports data generation required for the dual-end training; Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

22. The method according to claim 21, characterized in that If the second information includes information for indicating a training mode of the dual-end training supported by the first device, the training mode of the dual-end training includes one or more of the following: Multiple devices complete model training for multiple models respectively; Multiple devices complete model training for multiple models simultaneously.

23. The method of claim 21, wherein: If the second information includes information for indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

24. The method of claim 21, wherein: If the second information includes information for indicating whether the first device supports data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, the data transmission required for the dual-end training includes transmitting one or more of the following data: Model parameters required for the dual-end training; Model gradient data required for the dual-end training; The quantization data of the model required for the dual-end training.

25. The method according to any one of claims 21 to 24, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

26. The method according to any one of claims 21 to 24, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports generating data required for the dual-end training; Information used to indicate whether the first device supports transmission of data required for the dual-end training.

27. The method according to any one of claims 21 to 24, characterized in that If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the dual-end training is supported to be initiated by the first device; Information used to indicate whether the dual-end training is supported to be initiated by the second device.

28. The method according to any one of claims 21 to 27, characterized in that The dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

29. The method according to any one of claims 18 to 28, characterized in that The capability information includes the third information, and the first condition is used to indicate one or more of the following: The first device supports use case type restrictions on model use cases for performing the model training; The first device supports restrictions on model types for performing the model training; The first device supports a model size limit for performing the model training; a data volume limit of data supported by the first device for the model training; A computing power limit supported by the first device for training the model; a time limit during which the first device can be used for training the model; The software limitations of the model training supported by the first device.

30. The method according to any one of claims 17 to 29, characterized in that If the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

31. The method according to any one of claims 17 to 30, characterized in that If the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deployment of the model to hardware of the first device.

32. The method according to any one of claims 17 to 31, characterized in that The method further comprises: The second device sends fourth information to the first device, where the fourth information includes one or more of the following: Information used to instruct the first device to send the capability information; Information used to indicate information content carried in the capability information sent by the first device.

33. A communication device, characterized in that: The communication device is a first device, comprising: A sending unit, configured to send capability information of the first device to a second device, where the capability information is associated with one or more of the following: Model training; Model compilation; Model deployment.

34. The communication device according to claim 33, characterized in that If the capability information is associated with the model training, the capability information includes one or more of the following: First information used to indicate whether the first device supports the model training; second information for indicating a training type of the model training supported by the first device; The third information is used to indicate a first condition, where the first condition is a restriction condition for the first device to perform the model training.

35. The communication device according to claim 34, characterized in that The training type of the model training includes online training, and the second information further includes one or more of the following: Information used to indicate whether the first device supports the online training; information indicating a duration supported by the first device and available for the online training; Information used to indicate the training complexity of the online training supported by the first device; Information indicating a data volume supported by the first device and available for the online training.

36. The communication device according to claim 35, characterized in that The online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

37. The communication device according to claim 34, characterized in that The training type of the model training includes dual-end training, and the second information includes one or more of the following: Information for indicating a training device for performing the dual-end training; Information used to indicate whether the first device supports the dual-end training; Information used to indicate a training mode of the dual-end training supported by the first device; used to indicate whether the first device supports initiating the dual-end training by the first device; used to indicate whether the first device supports the dual-end training initiated by the second device; Information used to indicate whether the first device supports data transmission required for the dual-end training; Information used to indicate whether the first device supports data generation required for the dual-end training; Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

38. The communication device according to claim 37, characterized in that If the second information includes information for indicating a training mode of the dual-end training supported by the first device, the training mode of the dual-end training includes one or more of the following: Multiple devices complete model training for multiple models respectively; Multiple devices complete model training for multiple models simultaneously.

39. The communication device according to claim 37, characterized in that If the second information includes information for indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

40. The communication device according to claim 37, characterized in that If the second information includes information for indicating whether the first device supports data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, the data transmission required for the dual-end training includes transmitting one or more of the following data: Model parameters required for the dual-end training; Model gradient data required for the dual-end training; The quantization data of the model required for the dual-end training.

41. The communication device according to any one of claims 37 to 40, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

42. The communication device according to any one of claims 37 to 40, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports generating data required for the dual-end training; Information used to indicate whether the first device supports transmission of data required for the dual-end training.

43. The communication device according to any one of claims 37 to 40, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the dual-end training is supported to be initiated by the first device; Information used to indicate whether the dual-end training is supported to be initiated by the second device.

44. The communication device according to any one of claims 37 to 43, characterized in that: The dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

45. The communication device according to any one of claims 34 to 44, characterized in that: The capability information includes the third information, and the first condition is used to indicate one or more of the following: The first device supports use case type restrictions on model use cases for performing the model training; The first device supports restrictions on model types for performing the model training; The first device supports a model size limit for performing the model training; a data volume limit of data supported by the first device for the model training; A computing power limit supported by the first device for training the model; a time limit during which the first device can be used for training the model; The software limitations of the model training supported by the first device.

46. ​​The communication device according to any one of claims 33 to 45, characterized in that: If the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

47. The communication device according to any one of claims 33 to 46, characterized in that: If the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deployment of the model to hardware of the first device.

48. The communication device according to any one of claims 33 to 47, characterized in that: The communication device further comprises: A receiving unit, configured to receive fourth information sent by the second device, wherein the fourth information includes one or more of the following: Information used to instruct the first device to send the capability information; Information used to indicate information content carried in the capability information sent by the first device.

49. A communication device, characterized in that: The communication device is a second device, comprising: A receiving unit is configured to receive capability information of the first device sent by a first device, where the capability information is associated with one or more of the following: Model training; Model compilation; Model deployment.

50. The communication device according to claim 49, characterized in that If the capability information is associated with the model training, the capability information includes one or more of the following: First information used to indicate whether the first device supports the model training; second information for indicating a training type of the model training supported by the first device; The third information is used to indicate a first condition, where the first condition is a restriction condition for the first device to perform the model training.

51. The communication device according to claim 50, characterized in that The training type of the model training includes online training, and the second information further includes one or more of the following: Information used to indicate whether the first device supports the online training; information indicating a duration supported by the first device and available for the online training; Information used to indicate the training complexity of the online training supported by the first device; Information indicating a data volume supported by the first device and available for the online training.

52. The communication device according to claim 51, characterized in that The online training is one of multiple types of online training, and different types of online training among the multiple types of online training are associated with different second information.

53. The communication device according to claim 50, characterized in that The training type of the model training includes dual-end training, and the second information includes one or more of the following: Information for indicating a training device for performing the dual-end training; Information used to indicate whether the first device supports the dual-end training; Information used to indicate a training mode of the dual-end training supported by the first device; used to indicate whether the first device supports initiating the dual-end training by the first device; used to indicate whether the first device supports the dual-end training initiated by the second device; Information used to indicate whether the first device supports data transmission required for the dual-end training; Information used to indicate whether the first device supports data generation required for the dual-end training; Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

54. The communication device according to claim 53, characterized in that If the second information includes information for indicating a training mode of the dual-end training supported by the first device, the training mode of the dual-end training includes one or more of the following: Multiple devices complete model training for multiple models respectively; Multiple devices complete model training for multiple models simultaneously.

55. The communication device according to claim 53, characterized in that If the second information includes information for indicating a training device for performing the dual-end training, the training device includes the first device or a target device associated with the first device.

56. The communication device according to claim 53, characterized in that If the second information includes information for indicating whether the first device supports data transmission required for the dual-end training, and the dual-end training is for multiple devices to simultaneously complete model training of multiple models, the data transmission required for the dual-end training includes transmitting one or more of the following data: Model parameters required for the dual-end training; Model gradient data required for the dual-end training; The quantization data of the model required for the dual-end training.

57. The communication device according to any one of claims 53 to 56, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports data reception required for the dual-end training; Information used to indicate whether the first device supports data processing required for the dual-end training.

58. The communication device according to any one of claims 53 to 56, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the first device supports generating data required for the dual-end training; Information used to indicate whether the first device supports transmission of data required for the dual-end training.

59. The communication device according to any one of claims 53 to 56, characterized in that: If the dual-end training is that multiple devices complete model training for multiple models respectively, the second information includes one or more of the following: Information used to indicate whether the dual-end training is supported to be initiated by the first device; Information used to indicate whether the dual-end training is supported to be initiated by the second device.

60. The communication device according to any one of claims 53 to 59, characterized in that: The dual-end training is one of multiple types of dual-end training, and different types of dual-end training among the multiple types of dual-end training are associated with different second information.

61. The communication device according to any one of claims 50 to 60, characterized in that: The capability information includes the third information, and the first condition is used to indicate one or more of the following: The first device supports use case type restrictions on model use cases for performing the model training; The first device supports restrictions on model types for performing the model training; The first device supports a model size limit for performing the model training; a data volume limit of data supported by the first device for the model training; A computing power limit supported by the first device for training the model; a time limit during which the first device can be used for training the model; The software limitations of the model training supported by the first device.

62. The communication device according to any one of claims 49 to 61, characterized in that: If the capability information is associated with the model compilation, the capability information is used to indicate whether the first device supports compiling the model into a hardware instruction set.

63. The communication device according to any one of claims 49 to 62, characterized in that: If the capability information is associated with the model deployment, the capability information is used to indicate whether the first device supports deployment of the model to hardware of the first device.

64. The communication device according to any one of claims 49 to 63, characterized in that: The communication device further comprises: A sending unit, configured to send fourth information to the first device, wherein the fourth information includes one or more of the following: Information used to instruct the first device to send the capability information; Information used to indicate information content carried in the capability information sent by the first device.

65. A communication device, characterized in that: It comprises a transceiver, a memory and a processor, wherein the memory is used to store programs, and the processor is used to call the programs in the memory and control the transceiver to receive or send signals so that the communication device executes the method as described in any one of claims 1 to 32.

66. A device, characterized in that The device comprises a processor, configured to call a program from a memory so as to cause the device to execute a method as claimed in any one of claims 1 to 32.

67. A chip, characterized in that: It comprises a processor, which is used to call a program from a memory, so that a device equipped with the chip executes a method as claimed in any one of claims 1 to 32.

68. A computer-readable storage medium, characterized in that A program is stored thereon, the program causing a computer to execute the method according to any one of claims 1 to 32.

69. A computer program product, characterized in that The method comprises a program which causes a computer to execute the method according to any one of claims 1 to 32.

70. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1 to 32.

Images