Method for model transmission and communication equipment

Abstract

The invention provides a method for model transmission and communication equipment. The method comprises: a first device receiving model information of a first model sent by a second device, the model information of the first model comprising one or more of the following: partial or all model parameters of the first model; the first information is used for determining a model structure of the first model; the second information is associated with the compiling format of the first model; and third information associated with the compression of the first model. In the embodiment of the invention, the second equipment can transmit the first model by sending the model information of the first model to the first equipment. Compared with a traditional model transmission scheme in which all the model data of the first model needs to be transmitted, the data size of the to-be-transmitted model information in the model transmission process is reduced, so that the transmission time needed in the model transmission process is shortened.

Description

Method and communication device for model transmission Technical Field

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

[0002] In many scenarios, models need to be transmitted between a first device and a second device. In some cases, the first device may urgently need the model to ensure communication continuity. Therefore, model transmission becomes extremely important. In traditional model transmission methods, the second device transmits the model as a whole to the first device. However, this model transmission method requires a large amount of model information to be transmitted, resulting in a long transmission time.

[0003] Summary of the Invention

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

[0005] In a first aspect, a method for model transmission is provided, comprising: a first device receives model information of a first model sent by a second device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information for determining the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[0006] In a second aspect, a method for model transmission is provided, comprising: a second device sends model information of a first model to a first device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information for determining the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[0007] According to a third aspect, a communication device is provided, which is a first device and includes: a receiving unit for receiving model information of a first model sent by a second device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information for determining the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[0008] In a fourth aspect, a communication device is provided, which is a second device and includes: a sending unit for sending model information of a first model to a first device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information for determining the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[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 second device can transmit the first model by sending the model information of the first model to the first device. Compared with traditional model transmission solutions that require the transmission of all model data of the first model, the provision of the model information of the first model helps reduce the amount of model information to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process. 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 a neural network applicable to an embodiment of the present application.

[0017] FIG3 is a schematic diagram of another neural network applicable to an embodiment of the present application.

[0018] FIG4 is a schematic flowchart of a method for model transmission according to an embodiment of the present application.

[0019] FIG5 is a schematic flowchart of a method for model transmission according to another embodiment of the present application.

[0020] FIG6 is a schematic diagram of a communication device according to an embodiment of the present application.

[0021] FIG7 is a schematic diagram of a communication device according to another embodiment of the present application.

[0022] FIG8 is a schematic structural diagram of a communication device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The technical solution in this application will be described below with reference to the accompanying drawings.

[0024] 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.

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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).

[0034] Artificial intelligence (AI) models

[0035] 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, as a type of AI model, 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, commonly referred to as model parameters. Each node performs a weighted summation of different input signals and outputs the result through a specific activation function.

[0036] Common neural networks include convolutional neural network (CNN), recurrent neural network (RNN), deep neural network (DNN), etc.

[0037] The following describes a neural network applicable to embodiments of the present application in conjunction with FIG2 . The neural network shown in FIG2 can be divided into three categories based on the location of different layers: input layer 210 , hidden layer 220 , and output layer 230 . Generally speaking, the first layer is the input layer 210 , the last layer is the output layer 230 , and the intermediate layers between the first and last layers are all hidden layers 220 .

[0038] The input layer 210 is used to input data, where the input data can be, for example, a received signal received by a receiver. The hidden layer 220 is used to process the input data, for example, decompress the received signal. The output layer 230 is used to output processed output data, for example, a decompressed signal.

[0039] As shown in Figure 2, 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.

[0040] 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, the more parameters a model has, the higher its complexity and "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.

[0041] As shown in Figure 3, a common neural network model structure may include a convolutional layer (denoted by "conv"), a fully connected layer (denoted by "fc"), and a softmax layer. In some implementations, the neural network model may also include a pooling layer. Each layer can be considered a model module.

[0042] Each convolution layer can include many convolution operators, also known as kernels, which can be regarded as a filter that extracts specific information from the input signal. The convolution operator can essentially be a parameter matrix, which is usually predefined.

[0043] The parameter values ​​in these parameter matrices need to be obtained through a lot of training in practical applications. The parameter matrices formed by the parameter values ​​obtained through training can extract information from the input signal, thereby helping CNN to make correct predictions.

[0044] 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.

[0045] Pooling layers 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 3, 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.

[0046] After being processed by the convolution layer and the pooling layer, the CNN is still not sufficient to output the required output information. Because as mentioned above, the convolution layer and the pooling layer 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), the CNN also needs to use the fully connected layer. Generally, a fully connected layer can include multiple hidden layers, and the parameters contained in the multiple hidden layers can be pre-trained according to the relevant training data of the specific task type. For example, the task type can include decoding the data signal received by the receiver. For example, the task type can also include channel estimation based on the pilot signal received by the receiver.

[0047] After the multiple hidden layers in the fully connected layer, the final layer of the CNN is the output layer, which is used to output the results. Typically, this output layer is configured with a loss function (for example, a loss function similar to categorical cross entropy) to calculate the prediction error, or 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).

[0048] In order to minimize the loss function, the CNN model needs to be trained. In some implementations, the backpropagation algorithm (BP) can be used to train the CNN model. The BP training process consists of a forward propagation process and a backpropagation process. During the forward propagation process (as shown in Figure 3, the propagation from the input layer to the output layer is forward propagation), the input data is input into the above-mentioned layers of the CNN model, processed layer by layer and transmitted to the output layer. If the result output at the output layer is significantly different from the ideal result, the minimization of the above-mentioned loss function is used as the optimization goal, and backpropagation is entered (as shown in Figure 3, the propagation from the output layer to the input layer is backpropagation). The partial derivatives of the optimization target with respect to the weights of each neuron are calculated layer by layer to form the gradient of the optimization target with respect to the weight vector, which is used as the basis for modifying the model parameters. The CNN training process is completed during the parameter modification process. When the above-mentioned error reaches the expected value, the CNN training process ends.

[0049] It should be noted that the CNN shown in Figure 3 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.

[0050] 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.

[0051] Training an RNN is similar to training a traditional artificial neural network (ANN). 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, it is necessary to propagate three steps backward, and the gradients of the three subsequent steps must be added. This learning algorithm is called backpropagation through time (BPTT).

[0052] 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 input and output 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 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. To put it simply, an RNN is a unit structure that is reused.

[0053] Model segmentation

[0054] Taking the AI ​​model introduced above as an example, as shown in Figure 2, the model can be composed of multiple layers, each layer containing multiple nodes (e.g., neurons). The number, arrangement, and connection method of these layers can be referred to as the model structure. In addition, the nodes and edges in the model have corresponding weights that are used to adjust the model's processing power. These weights are called model parameters. In other words, a model can be divided into two parts: model structure and model parameters.

[0055] In some implementations, the model structure can be further divided into different modules to take into account the layering of the model structure. As shown in Figure 2, each model module can include one or more layers of the model. Model modules can be replaced as needed to achieve different model performance. Currently, common models include long short-term memory (LSTM) networks and transformer models, and model modules typically include fully connected layers and / or convolutional layers.

[0056] Model Transfer

[0057] In some scenarios, models can be transmitted between communication devices, for example, models can be transmitted over the air interface between communication devices. Currently, the communication protocol specifies seven ways for network devices to transmit models to terminal devices. The following describes model transmission methods 1 to 7, where the network device can be any of the following: access network equipment; core network equipment other than location management function (LMF); LMF; server (for example, over-the-top OTT server, information management equipment, or operation administration and maintenance (OAM) equipment).

[0058] Model transmission mode 1: The access network device transmits the model to the terminal device through a radio resource control (RRC) message.

[0059] Model transmission mode 2: The core network device (except LMF) transmits the model to the terminal device through NAS signaling.

[0060] Model transmission mode 3: LMF transmits the model to the terminal device through the long term evolution positioning protocol (LPP) message.

[0061] Model transmission method 4: The access network device transmits the model to the terminal device through the user.

[0062] Model transmission mode 5: The core network device (except LMF) transmits the model to the terminal device through the user.

[0063] Model transmission mode 6: LMF transmits the model to the terminal device through the user. The message used to transmit the model is not defined in 3GPP.

[0064] Model transmission mode 7: The server transmits the model to the terminal device, where the message used to transmit the model is not defined in 3GPP.

[0065] In some implementations, the models described above can be identified by a model identifier (or "model ID"). In some implementations, different model IDs can be assigned to models with the same structure, taking into account differences in deployment locations, implementation scenarios, and so on. Typically, the model ID is understood by both the transmitter and receiver in a communication system, meaning that either party can determine the model's functionality, structure, and parameters based on the model ID.

[0066] In some implementations, if the model functionality of the model cannot be determined based on the model identifier, the model functionality may be determined based on the model identifier and auxiliary meta information (denoted as "meta info").

[0067] In other implementations, if the communication system supports model transmission, the model identifier (model ID) can be used for lifecycle management of the model.

[0068] Currently, the application scenarios of models are relatively limited. Typically, models can only be applied to a specific area (such as a single cell). The performance gain brought by model-based communication usually depends on the training data associated with the specific area. In other words, if the model is trained based on the training data associated with the specific area, then the model performance is generally better when used in the specific area. However, when the specific area changes or the service to be transmitted changes, the first device needs to obtain a new model from the second device.

[0069] In the embodiments of the present application, the second device can be, for example, any of the network devices described above, and the first device can be a terminal device. Of course, in the embodiments of the present application, both the first device and the second device can be terminal devices, or the first device can be a network device and the second device can be a terminal device. The following description uses the first device as a terminal device and the second device as a network device as an example.

[0070] Taking the change of a specific area as an example, in some scenarios, the terminal device is in motion. When the terminal device moves from cell 1 to cell 2, that is, the specific area changes, the model of cell 1 may no longer be applicable. At this time, the terminal device needs to re-obtain the model corresponding to cell 2 from the network device to communicate.

[0071] Taking the change of the service to be transmitted as an example, in some scenarios, if the terminal device temporarily needs to perform beam management based on the model, but the terminal device does not have a model that can be used for beam management, at this time, the network device can send the model for beam management to the terminal device so that the terminal device can perform beam management.

[0072] As mentioned above, models need to be transmitted between a first device and a second device in many scenarios. In particular, in some scenarios, the first device may urgently need the model to ensure communication continuity. Therefore, model transmission becomes extremely important. In traditional model transmission methods, the second device transmits the model as a whole to the first device. However, this model transmission method requires a long transmission time.

[0073] Therefore, to address the above issues, embodiments of the present application provide a method for model transmission that helps reduce the amount of model information data required to be transmitted during the transmission of a model (hereinafter referred to as the "first model"), thereby reducing the transmission time required for the model transmission process. Therefore, the model transmission process of embodiments of the present application can also be referred to as a "lightweight model transmission process."

[0074] In some scenarios, the lightweight model transmission process can support the application of specialized models, that is, lightweight model transmission can make model updates more convenient. Accordingly, the model design can be more targeted at specific scenarios. Using a model targeted at that scenario in a specific scenario can achieve better performance than a general model.

[0075] In other scenarios, the lightweight model transmission process avoids occupying a large amount of resources of the first device. That is, fast model transmission means that the first device does not need to store a large number of models, and the second device can dynamically update the model for the first device in real time, thereby reducing the hardware requirements for the first device and bringing better technical backward compatibility.

[0076] In other scenarios, the lightweight model transmission process helps improve mobile user service continuity. For example, when determining whether to switch cells based on a model, and updating the model when the signal quality of the source service cell is below a specific threshold, the transmission of the lightweight model can reduce the switching decision delay and thus reduce service interruption time.

[0077] In other scenarios, the lightweight model transmission process helps reduce the impact of model transmission on traditional communication services. That is, model transmission will occupy the original control plane / user plane / non-access layer resources and affect the progress of other communication services. The lightweight model transmission process can reduce the occupation of the above resources by model transmission to a certain extent.

[0078] The following describes a method for model transmission according to an embodiment of the present application in conjunction with FIG4 . In the embodiment of the present application, the first model is not limited. For example, the first model may be an AI model, such as any of the neural network models described above, or other neural network models. For another example, the first model may be a machine learning (ML) model.

[0079] Fig. 4 is a schematic flow chart of a method for model transmission according to an embodiment of the present application. The method shown in Fig. 4 includes S410.

[0080] In S410 , the second device sends model information of the first model to the first device.

[0081] In some implementations, the model information of the first model is used for transmission of the first model, or in other words, the model information of the first model is associated with the transmission of the first model. In other implementations, the model information of the first model is used for lightweight transmission of the first model, or in other words, the model information of the first model is associated with the lightweight transmission of the first model.

[0082] In some implementations, the model information of the first model includes one or more of the following: some or all model parameters of the first model; first information used to determine the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model. These are described below with reference to Examples 1 to 4.

[0083] Example 1: The model information of the first model includes part or all of the model parameters of the first model.

[0084] In the embodiments of the present application, the applicant has found that there are certain similarities between different models. For example, some models may have similar model structures. For another example, the businesses processed by some models are similar. This means that some model information is repeated between different models. For example, for some models with similar model structures, their model structure information is repeated. For another example, for some models with similar businesses processed by the models, the model parameters are repeated. In this way, the applicant believes that during the model transmission process, the second device does not actually need to transmit the complete model information every time the model is transmitted. It can only transmit the non-repeated part of the model information, such as part or all of the model parameters of the first model, which helps to reduce the amount of model data to be transmitted during the model transmission process, so as to reduce the time required for the model transmission process.

[0085] Therefore, in an embodiment of the present application, the first model can be parameterized, that is, as mentioned above, the model information of the first model can include part or all of the model parameters of the first model to achieve the purpose of transmitting the first model, which helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, so as to reduce the transmission time required for the model transmission process.

[0086] In some implementations, model parameters may also be referred to as model weights. For details, please refer to the above introduction on model segmentation. Taking the example that the model information of the first model includes the partial model weights of the first model, the model information of the first model may include the model weights of some layers of the first model. Taking the model structure shown in Figure 2 as an example, the model weights of some layers of the first model may be the model parameters of the input layer.

[0087] In other implementations, the partial model parameters may be those that have a greater impact on the performance of the first model among all the model parameters of the first model, or in other words, the partial model parameters may be associated with the performance of the first model. In this way, transmitting the partial model parameters of the first model via the model parameters of the first model helps reduce the impact of transmitting the partial model parameters on the performance of the first model, and also helps reduce the amount of model information data required to be transmitted during the model transmission process.

[0088] Taking the example that the model information of the first model includes all model weights of the first model, the model information of the first model may include the model weights of all layers of the first model.

[0089] Example 2: The model information of the first model includes first information for determining the model structure of the first model.

[0090] In the embodiments of the present application, the applicant has found that there are certain similarities between different models. For example, some models may have similar model structures. For another example, some models may handle similar businesses. This means that some model information is repeated between different models. For example, for some models with similar model structures, their model structure information is repeated. For another example, for some models with similar businesses handled by the models, the model parameters are repeated. In this way, the applicant believes that during the model transmission process, the second device does not actually need to transmit the complete model information every time the model is transmitted. It can only transmit the non-repeated part of the model information, such as the first information used to determine the model structure of the first model, which helps to reduce the amount of model data to be transmitted during the model transmission process, so as to reduce the time required for the model transmission process.

[0091] Therefore, in an embodiment of the present application, the first information can be carried by the model information of the first model to achieve the purpose of transmitting the first model, which helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0092] In some implementations, the first information may include one or more of the following: a first identifier; a second identifier; a third identifier; fourth information; or fifth information.

[0093] Taking the first information including the first identifier as an example, the first identifier can be associated with the second model. In some implementations, associating the first identifier with the second model can include the first identifier being an identifier of the second model, also called a model identifier of the second model.

[0094] In an embodiment of the present application, the purpose of transmitting the first model can be achieved by carrying the first identifier through the model information of the first model. Compared with the traditional model transmission scheme, transmitting the model as a whole helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0095] In some implementations, the second model may be the same model as the first model. In this case, the first identifier may alternatively be associated with the first model, for example, the first identifier being the model identifier of the first model. In other implementations, the first model may be determined based on the second model, or in other words, the first model may be obtained by modifying the second model. For example, the first model may be obtained by adding a model module to the second model. In another example, the first model may be obtained by deleting a model module from the second model.

[0096] Taking the example of the first information including the second identifier, the second identifier can be associated with the model structure of the second model. In some implementations, associating the second identifier with the model structure of the second model can include the second identifier being an identifier of the model structure of the second model, also known as the model structure ID of the second model. For an introduction to the model structure, please refer to the previous introduction to model segmentation.

[0097] In an embodiment of the present application, the purpose of transmitting the first model can be achieved by carrying the second identifier in the model information of the first model. Compared with the traditional model transmission scheme, transmitting the model as a whole helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0098] In some implementations, the model structure of the second model may be the same as the model structure of the first model. Alternatively, the second identifier may be associated with the model structure of the first model. For example, the second identifier may be the model structure identifier of the first model. In other implementations, the model structure of the first model may be determined based on the model structure of the second model, or in other words, the model structure of the first model may be obtained by modifying the model structure of the second model.

[0099] For example, the model structure of the first model can be obtained by adding a model module to the model structure of the second model. For another example, the model structure of the first model can be obtained by deleting a model module from the model structure of the second model.

[0100] Taking the example of the first information including the third identifier, the third identifier can be associated with a model module in the first model. In some implementations, associating the third identifier with the model module in the first model can include the third identifier being an identifier of the model module in the first model, also referred to as the model module ID in the first model. For an introduction to model modules, please refer to the previous introduction to model segmentation.

[0101] In an embodiment of the present application, the purpose of transmitting the first model can be achieved by carrying a third identifier in the model information of the first model. Compared with the traditional model transmission scheme, transmitting the model as a whole helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0102] In some implementations, the model structure of the first model may be determined based on the model structure of the second model, or in other words, the model structure of the first model may be obtained by modifying the model structure of the second model. For example, the model structure of the first model may be obtained by adding a model module associated with the third identifier to the model structure of the second model. For another example, the model structure of the first model may be obtained by deleting the model module associated with the third identifier from the model structure of the second model.

[0103] In some implementations, in addition to indicating the model module in the first model, the above-mentioned third identifier may also indicate a modification method associated with the model module in the first model. For example, the model structure of the first model can be obtained by adding a model module associated with the third identifier on the basis of the model structure of the second model. In this case, the third identifier is used to indicate the addition of the model module. For another example, the model structure of the first model can be obtained by deleting the model module associated with the third identifier on the basis of the model structure of the second model. In this case, the third identifier is used to indicate the deletion of the model module. Of course, in the embodiment of the present application, the above-mentioned model module and the modification method associated with the model module can be indicated by different information. At this time, the third identifier may only indicate the model module in the first model.

[0104] Taking the example of the first information including the fourth information, the fourth information is associated with the arrangement position of the model module in the first model, or the fourth information can be used to indicate the arrangement position of the model module in the first model, wherein the arrangement position can be used to determine the layer of the model module in the first model.

[0105] In an embodiment of the present application, the fourth information can be carried by the model information of the first model to achieve the purpose of transmitting the first model. Compared with the traditional model transmission scheme, transmitting the model as a whole helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0106] In some implementations, the fourth information may include an identifier of the layer of the model module in the first model. For example, if the fourth information is the identifier of the second layer in the first model, it can be determined that the model module is located in the second layer in the first model. In other implementations, the fourth information may indicate the identifier of the previous layer of the model module in the first model. For example, if the fourth information is the identifier of the third layer in the first model, it can be determined that the model module is located in the layer below the third layer in the first model. Of course, in an embodiment of the present application, the fourth information may indicate the identifier of the next layer of the model module in the first model. For example, if the fourth information is the identifier of the fourth layer in the first model, it can be determined that the model module is located in the layer above the fourth layer in the first model.

[0107] In an embodiment of the present application, the first information may not carry the fourth information. In some scenarios, if the arrangement position of the model module in the first model can be directly determined based on the model module, the first information may not carry the fourth information. For example, the number of input nodes of the model module is n, and accordingly, only the number of output nodes of the second layer in the first model is n, and the number of output nodes of other layers is greater than n. At this time, the number of input nodes of the model module does not match the number of output nodes of other layers. Therefore, the model module can only be located after the second layer in the first model. In this way, the position of the model module in the first model can be determined directly based on the number of inputs of the model module, and no longer based on the fourth information.

[0108] Taking the example that the first information includes the fifth information, the fifth information may include module data of some model modules in the first model.

[0109] In some implementations, the module data may include model parameters of the model module, and / or the model structure of the model module.

[0110] In some implementations, the model module associated with the fifth information may be the module data of the model module that needs to be modified in the second model as described above. In other implementations, the model module associated with the fifth information may be the module data of the model module to be updated in the second model, and the updated second model is the first model.

[0111] The above describes the first information in the embodiments of the present application. It should be understood that the above first information can be used alone or in combination with each other in the embodiments of the present application. The following describes the scheme of combining the above first information with each other in conjunction with Implementation Methods 1 to Implementation Method 4, wherein the schemes of combining with each other in the embodiments of the present application are not limited to this.

[0112] Implementation method 1: The model information of the first model includes the first identifier and part or all of the model parameters of the first model.

[0113] In some implementations, the first model may be obtained based on the second model. In this case, the model information of the first model may include a first identifier associated with the second model and some or all model parameters of the first model.

[0114] For example, the first device may store multiple models. Accordingly, the second device may send a first identifier and some or all model parameters of the first model to the first device. In this case, the second device may select a second model from the multiple models based on the first identifier, and combine the second model with some or all model parameters of the received first model to obtain the first model.

[0115] In an embodiment of the present application, the first model can be transmitted by including a first identifier and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission schemes that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0116] Implementation method 2: The model information of the first model includes the second identifier and part or all of the model parameters of the first model.

[0117] In some implementations, the model structure of the first model may be the same as the model structure of the second model. In this case, the model information of the first model may include a second identifier associated with the model structure of the first model and some or all model parameters of the first model.

[0118] For example, the first device may store the model structures of multiple models. Accordingly, the second device may send a second identifier and some or all model parameters of the first model to the first device. In this case, the second device may select the model structure of the first model from the model structures of the multiple models based on the second identifier, and combine the model structure of the first model with the received model parameters of the first model or some or all of the model parameters of the first model to obtain the first model.

[0119] In an embodiment of the present application, by carrying a second identifier and part or all of the model parameters of the first model in the model information of the first model, compared with the traditional model transmission scheme, all the model information of the first model is transmitted, which helps to reduce the amount of data of the model information required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0120] Implementation method 3: The model information of the first model includes the third identifier and the fourth information.

[0121] In some implementations, the first model is obtained based on the second model. In this case, the model information of the first model may include a third identifier and an arrangement position of the model module associated with the third identifier in the first model.

[0122] For example, the first device may store the model structure of the second model. Accordingly, the second device may send the third identifier and the fourth information to the first device. In this case, the second device may determine the arrangement position of the model module associated with the third identifier in the first model based on the fourth information, and combine the model module associated with the third identifier with the model structure of the second model based on the fourth information to obtain the first model.

[0123] In an embodiment of the present application, the first model can be transmitted by including the third identifier and the fourth information in the model information of the first model. Compared to traditional model transmission solutions that transmit all the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0124] Implementation method 4: The model information of the first model includes the third identifier and part or all of the model parameters of the first model.

[0125] In some implementations, the first model is obtained based on the second model. In this case, the model information of the first model may include the third identifier and part or all of the model parameters of the first model.

[0126] For example, the first device may store the model structure of the second model. Accordingly, the second device may send the third identifier and some or all model parameters of the first model to the first device. In this case, the second device may modify the model structure of the second model based on the model module associated with the third identifier to obtain the model structure of the first model, and combine the model structure of the first model with some or all model parameters of the first model to obtain the first model.

[0127] In an embodiment of the present application, the first model can be transmitted by including the third identifier and the fourth information in the model information of the first model. Compared to traditional model transmission solutions that transmit all the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0128] In an embodiment of the present application, the model information of the first model may include other model information in addition to the third identifier and some or all of the model parameters of the first model. The following describes the model information of the first model in an embodiment of the present application, taking the case where the first model is determined based on the second model as an example, in conjunction with implementations 4-1 to 4-4.

[0129] Implementation method 4-1: The model information of the first model also includes a first identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the second model.

[0130] That is to say, the model information of the first model includes the first identifier, the third identifier and part or all of the model parameters of the first model, wherein the first identifier is associated with the second model, and the model module associated with the third identifier is the model module that needs to be deleted in the second model.

[0131] For example, a first device may store multiple models. If a second device sends model information of a first model to the first device, the first device may select the second model from the multiple models based on the first identifier. Furthermore, the first device may determine the model modules that need to be deleted from the second model based on a third identifier, and delete the model modules associated with the third identifier from the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0132] In an embodiment of the present application, the first model can be transmitted by carrying the first identifier, the third identifier, and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0133] Implementation method 4-2: The model information of the first model also includes a first identifier, and the model module associated with the third identifier is the model module that needs to be added to the second model.

[0134] That is to say, the model information of the first model includes the first identifier, the third identifier and part or all of the model parameters of the first model, wherein the first identifier is associated with the second model, and the model module associated with the third identifier is the model module that needs to be added to the second model.

[0135] In some implementations, the first device may store multiple models. If the second device sends model information of the first model to the first device, the first device may select the second model from the multiple models based on the first identifier. Furthermore, the first device may determine the model modules that need to be added to the second model based on the third identifier, and add the model modules associated with the third identifier to the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0136] For example, assuming that the first identifier in the model information of the first model is the model ID of model 2, the third identifier in the model information of the first model indicates the addition of model module 2, and the model information of the first model also includes the model parameters of the first model. After receiving the model information of the first model, the first device can select model 2 from multiple models based on the first identifier, and the model modules included in model 2 are model module 1, model module 3, and model module 4. In addition, the first device can add model module 2 to model 2 based on the third identifier to obtain the model structure of the first model. At this time, the model structure of the first model includes model module 1, model module 2, model module 3, and model module 4. The first device combines part or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0137] In an embodiment of the present application, the first model can be transmitted by carrying the first identifier, the third identifier, and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0138] Implementation method 4-3: The model information of the first model also includes a second identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model.

[0139] That is to say, the model information of the first model includes the second identifier, the third identifier and part or all of the model parameters of the first model, wherein the second identifier is associated with the model structure of the second model, and the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model.

[0140] In some implementations, the first device may store the model structures of multiple models. If the second device sends the model information of the first model to the first device, the first device may select the model structure of the second model from the multiple models based on the second identifier. In addition, the first device may determine the model module that needs to be deleted from the model structure of the second model based on the third identifier, and delete the model module associated with the third identifier from the model structure of the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0141] For example, assuming that the second identifier in the model information of the first model is the model structure ID of model 2, the third identifier in the model information of the first model indicates deletion of model module 2, and the model information of the first model also includes model parameters of the first model. After receiving the model information of the first model, the first device can select the model structure of model 2 from multiple models based on the first identifier, and the model modules included in the model structure of model 2 are model module 1, model module 2, model module 3 and model module 4. In addition, the first device can delete model module 2 from model 2 based on the third identifier to obtain the model structure of the first model. At this time, the model structure of the first model includes model module 1, model module 3 and model module 4. The first device combines part or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0142] In an embodiment of the present application, the first model can be transmitted by carrying the second identifier, the third identifier, and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0143] Implementation method 4-4: The model information of the first model also includes a second identifier, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

[0144] That is to say, the model information of the first model includes the second identifier, the third identifier and part or all of the model parameters of the first model, wherein the second identifier is associated with the model structure of the second model, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

[0145] In some implementations, the first device may store the model structures of multiple models. If the second device sends model information of the first model to the first device, the first device may select the model structure of the second model from the multiple models based on the second identifier. In addition, the first device may determine the model modules that need to be added to the model structure of the second model based on the third identifier, and add the model modules associated with the third identifier to the model structure of the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0146] For example, assuming that the second identifier in the model information of the first model is the model structure ID of model 2, the third identifier in the model information of the first model indicates the addition of model module 2, and the model information of the first model also includes the model parameters of the first model. After receiving the model information of the first model, the first device can select the model structure of model 2 from multiple models based on the second identifier, and the model modules included in the model structure of model 2 are model module 1, model module 3, and model module 4. In addition, the first device can add model module 2 to the model structure of model 2 based on the third identifier to obtain the model structure of the first model. At this time, the model structure of the first model includes model module 1, model module 2, model module 3, and model module 4. The first device combines part or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0147] In an embodiment of the present application, the first model can be transmitted by carrying the second identifier, the third identifier, and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0148] Based on the above introduction, it can be seen that the above solution can be applied to the scenario where the first device can store multiple models. In this case, the first identifier can be carried in the model information of the first model to indicate the second model from the multiple stored models, and / or the second identifier can be carried in the model information of the first model to indicate the model structure of the second model from the multiple stored model structures. In other scenarios, if the first device only stores one second model, the model information of the first model may no longer carry the first identifier and / or the second identifier, which helps to reduce the overhead required to transmit the model information of the first model. Of course, if the above problem is not taken into account, if the first device only stores one second model, the model information of the first model may also carry the first identifier and / or the second identifier.

[0149] The following describes the model information of the first model of the embodiment of the present application in combination with implementation methods 4-5 to 4-8, taking the example that the first model is determined based on the second model and the first device only stores one second model.

[0150] Implementation method 4-5: The model module associated with the third identifier is the model module that needs to be deleted in the second model.

[0151] That is, the model information of the first model includes the third identifier and part or all of the model parameters of the first model, wherein the model module associated with the third identifier is the model module that needs to be deleted in the second model.

[0152] In some implementations, a second model may be stored in the first device. If the second device sends model information of the first model to the first device, the first device may determine, based on the third identifier, a model module to be deleted from the second model, and delete the model module associated with the third identifier from the second model to obtain a model structure of the first model. The first device combines some or all model parameters of the first model with the model structure of the first model to obtain the first model.

[0153] In an embodiment of the present application, the first model can be transmitted by carrying a third identifier and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0154] Implementation method 4-6: The model module associated with the third identifier is the model module that needs to be added to the second model.

[0155] That is, the model information of the first model includes the third identifier and part or all of the model parameters of the first model, wherein the model module associated with the third identifier is the model module that needs to be added to the second model.

[0156] For example, a first device may store a second model. If the second device sends model information of the first model to the first device, the first device may determine the model modules that need to be added to the second model based on the third identifier, and add the model modules associated with the third identifier to the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0157] In an embodiment of the present application, the first model can be transmitted by carrying a third identifier and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0158] Implementation method 4-7: The model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model.

[0159] That is, the model information of the first model includes the third identifier and part or all of the model parameters of the first model, wherein the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model.

[0160] For example, the first device may store a model structure of a second model. If the second device sends model information of the first model to the first device, the first device may determine the model module that needs to be deleted from the model structure of the second model based on the third identifier, and delete the model module associated with the third identifier from the model structure of the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0161] In an embodiment of the present application, the first model can be transmitted by carrying a third identifier and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0162] Implementation method 4-8: The model module associated with the third identifier is a model module that needs to be added to the model structure of the second model.

[0163] That is, the model information of the first model includes the third identifier and part or all of the model parameters of the first model, wherein the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

[0164] For example, the first device may store a model structure of a second model. If the second device sends model information of the first model to the first device, the first device may determine the model modules that need to be added to the model structure of the second model based on the third identifier, and add the model modules associated with the third identifier to the model structure of the second model to obtain the model structure of the first model. The first device combines some or all of the model parameters of the first model with the model structure of the first model to obtain the first model.

[0165] In an embodiment of the present application, the first model can be transmitted by carrying a third identifier and some or all of the model parameters of the first model in the model information of the first model. Compared to traditional model transmission solutions that transmit all of the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0166] In some scenarios, the first device may not store some model modules in the first model, or in other scenarios, the first device may store some model modules in the first model, but the first device does not understand the model modules (for example, the first device cannot identify the function of the model modules). In this case, the first device may not be able to determine the first model based on the model information of the first model. For example, assuming that the model information of the first model only includes the third identifier, the first device may not be able to determine the model module based on the third identifier.

[0167] Therefore, to address the above issue, the second device can include fifth information in the model information of the first model, where the fifth information can include module data for some model modules in the first model. Accordingly, the first device can determine the model module based on the module data. In other words, the model information of the first model includes a target identifier and the fifth information, where the target identifier includes one or more of the first identifier, the second identifier, and the third identifier.

[0168] In some implementations, the model information of the first model may further include any of the information described above. For example, the model information of the first model may further include some or all of the model parameters of the first model. Taking the example where the model information of the first model includes the model parameters of the first model, if the fifth information includes the model parameters of the associated model module, the model parameters of the first model may include the model parameters of other model modules in addition to the model module associated with the fifth information. Of course, in the embodiment of the present application, the model parameters of the first model may also be all of the model parameters of the first model.

[0169] For ease of understanding, the fifth information of the embodiment of the present application is introduced below using the first identifier, the second identifier, and the third identifier as examples. It should be understood that the fifth information can be used in combination with the model information of any of the first models introduced above. The usage of the fifth information in different combinations is similar, and for the sake of brevity, it will not be repeated here.

[0170] Implementation method 5-1: The model information of the first model includes the first identifier and the fifth information.

[0171] In some implementations, the first identifier is associated with the second model, and the fifth information is used to indicate module data of some model modules in the first model.

[0172] For example, the first device may store multiple models. If the second device sends model information of the first model to the first device, the first device may select the second model from the multiple models based on the first identifier. Alternatively, the first device may combine the module data with the second model based on the fifth information to obtain the first model.

[0173] It should be noted that the combination of the model module associated with the fifth information and the second model is not limited. For example, the position of the model module associated with the fifth information in the second model can be determined based on the third identifier. In this case, the model information of the first model can carry the third identifier. Of course, in an embodiment of the present application, the position of the model module associated with the fifth information in the second model can be determined based on the model module associated with the fifth information, for example, based on the number of output nodes and / or the number of input nodes of the model module associated with the fifth information, etc. For details, please refer to the above introduction to the fourth information.

[0174] In an embodiment of the present application, the first model can be transmitted by carrying the first identifier and the fifth information in the model information of the first model. Compared with traditional model transmission solutions that transmit all the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0175] Implementation method 5-2: The model information of the first model also includes the second identifier and the fifth information.

[0176] In some implementations, the second identifier is associated with the model structure of the second model, and the fifth information is used to indicate module data of some model modules in the first model.

[0177] For example, the first device may store the model structures of multiple models. If the second device sends model information of the first model to the first device, the first device may select the model structure of the second model from the multiple models based on the second identifier. Alternatively, the first device may combine the module data of some model modules in the first model indicated by the fifth information with the model structure of the second model to obtain the first model.

[0178] It should be noted that the combination of the model module associated with the fifth information and the model structure of the second model is not limited. For example, the position of the model module associated with the fifth information in the model structure of the second model can be determined based on the third identifier. In this case, the model information of the first model can carry the third identifier. Of course, in an embodiment of the present application, the position of the model module associated with the fifth information in the model structure of the second model can be determined based on the model module associated with the fifth information, for example, based on the number of output nodes and / or the number of input nodes of the model module associated with the fifth information, etc. For details, please refer to the above introduction to the fourth information.

[0179] In an embodiment of the present application, the first model can be transmitted by carrying the second identifier and the fifth information in the model information of the first model. Compared with the traditional model transmission scheme that transmits all the model information of the first model, this helps to reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0180] Implementation method 5-3: The model information of the first model includes the third identifier and the fifth information.

[0181] In some implementations, the model module associated with the third identifier is a model module that needs to be added to the model structure of the second model. The fifth information is used to indicate module data of the model module associated with the third identifier.

[0182] For example, if the first device only stores the model structure of a second model and the second device sends the model information of the first model to the first device, the first device can determine the model module associated with the third identifier based on the fifth information. The first device can then determine the model module to add to the model structure of the second model based on the third identifier to obtain the first model.

[0183] In an embodiment of the present application, the first model can be transmitted by including the third identifier and the fifth information in the model information of the first model. Compared to traditional model transmission solutions that transmit all the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0184] Implementation method 5-4: The model information of the first model includes the second identifier, the third identifier and the fifth information.

[0185] In some implementations, the second identifier is used to indicate the model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model, and the fifth information is used to indicate the module data of the model module associated with the third identifier.

[0186] For example, the first device only stores the model structure of multiple models. If the second device sends the model information of the first model to the first device, the first device can determine the model module associated with the third identifier based on the fifth information. In addition, the first device can select the model structure of the second model from the model structures of multiple models based on the second identifier. After that, the first device can determine the model module added to the model structure of the second model based on the third identifier to obtain the first model.

[0187] In an embodiment of the present application, the first model can be transmitted by including the third identifier and the fifth information in the model information of the first model. Compared to traditional model transmission solutions that transmit all the model information of the first model, this helps reduce the amount of model information data required to be transmitted during the model transmission process, thereby reducing the transmission time required for the model transmission process.

[0188] In the above embodiment, in the process of determining the model structure of the first model, the model module associated with the third identifier and the arrangement position of the model module associated with the third identifier in the first model can be determined, wherein the arrangement position of the model module associated with the third identifier in the first model can be determined based on the model information of the first model. In this case, the model information of the first model can also include fourth information. Of course, the arrangement position of the model module associated with the third identifier in the first model can also be determined based on the number of input nodes of the model module and / or the number of output nodes of the model module. For details, please refer to the above description of the fourth information.

[0189] Example 3: The model information of the first model includes second information associated with the compilation format of the first model.

[0190] In an embodiment of the present application, the applicant discovered that the reason why the amount of model data required to be transmitted during the model transmission process is large is due to not considering the compilation format of the first model. The amount of model data of the first model varies significantly between different compilation formats. Therefore, in an embodiment of the present application, the compilation format of the first model can be selected, and the model information of the first model can be used to carry second information associated with the compilation format of the first model. This helps to reduce the amount of model information required to be transmitted during the transmission of the first model, thereby reducing the transmission time required for the model transmission process.

[0191] In some implementations, the compilation format of the first model may indicate the compilation format used to compile the model data of the first model. Taking the first model as an AI model as an example, the compilation format of the first model refers to converting the trained AI model into a specific binary format for execution on a specific hardware or software platform. Among them, the compilation format can improve the operating efficiency and performance of the AI ​​model, and can fully utilize the acceleration capabilities of the underlying hardware. Different hardware or software platforms may require different compilation formats. Common compilation formats include TensorFlow, open neural network exchange (ONNX), convolutional architecture for fast feature embedding (Caffe), Theano, PyTorch, etc. Through the compilation format, the AI ​​model can run efficiently in different environments such as embedded devices, mobile devices or cloud servers.

[0192] In an embodiment of the present application, the model data of the first model may include one or more of the following: model module data of the first model; model structure data of the first model; and model parameters of the first model. The model module data of the first model may include part or all of the model module data of the first model. The model structure data of the first model may include part or all of the model structure data of the first model. The model parameters of the first model may include part or all of the model parameters of the first model.

[0193] In some implementations, the second information may include information indicating the compilation format of the first model, or in other words, the second information is used to indicate the compilation format of the first model. Accordingly, in other embodiments, the model data of the first model includes a first model that conforms to the above-mentioned compilation format. In other words, the first model that conforms to the compilation format can be transmitted with the second information through the model information of the first model. Of course, in the embodiments of the present application, the first model that conforms to the compilation format can also be transmitted with the second information through different information.

[0194] In some implementations, the second information includes information for indicating the compiled format of the model data of the first model, or in other words, the second information is used to indicate the compiled format of the model data of the first model.

[0195] In some implementations, the second information may include an index of the coding format of the first model. In other implementations, the second information may include an identifier of the coding format of the first model.

[0196] In an embodiment of the present application, the second information can be carried by multiple bits in the model information of the first model. For example, if the value of multiple bits is 001, it means that the compilation format is PyTorch. For another example, if the value of multiple bits is 0001, it means that the compilation format is TensorFlow. In an embodiment of the present application, the number of multiple bits can be determined based on the total number of compilation formats to be indicated. Generally speaking, the number of bits is set so that the total number of compilation formats that can be indicated by the bits is greater than or equal to the total number of compilation formats to be indicated. For example, if the total number of compilation formats to be indicated is 5, the number of bits can be 3, so that the 3 bits can indicate 8 compilation formats.

[0197] In other embodiments, if the compilation format of the first model can also be a default compilation format, for example, the default compilation format is PyTorch, if the model information of the first model does not carry the second information, then the compilation format of the first model can also be PyTorch.

[0198] In some implementations, the compilation format associated with the second information is the first compilation format, and the model information of the first model further includes a first model that conforms to the first compilation format. The first compilation format may be one of one or more compilation formats supported by the second device. In other words, the first model that conforms to the first compilation format can be transmitted along with the second information via the model information of the first model.

[0199] In other implementations, the compilation format associated with the second information is the first compilation format, and the model information of the first model also includes model data of the first model that conforms to the first compilation format, where the first compilation format can be one of one or more compilation formats supported by the second device. In other words, the model data that conforms to the first compilation format can be transmitted along with the second information via the model information of the first model. The model data of the first model can include one or more of the following: some or all parameters of the first model; model structure information of the first model; and module data of some model modules in the first model.

[0200] Typically, different compilation formats may correspond to different amounts of model data. For example, if the compilation format is PyTorch, the model data amount of the first model that complies with the compilation format is 5M. If the compilation format is TensorFlow, the model data amount of the first model that complies with the compilation format is 10M. Therefore, in order to reduce the amount of model data of the first model, if the first device supports one or more compilation formats, the first compilation format can be the compilation format with the smallest amount of corresponding model data among the one or more compilation formats. Taking TensorFlow and PyTorch introduced above as an example, assuming that the first device supports TensorFlow and PyTorch, the first compilation format can be PyTorch.

[0201] It should be noted that the first compilation format is the compilation format with the smallest data size of the corresponding model data among the one or more compilation formats supported by the first device. However, the first compilation format may not be the compilation format with the smallest data size of the corresponding model data among the one or more compilation formats supported by the second device.

[0202] For example, the amount of model data corresponding to compilation format 1 is greater than the amount of model data corresponding to compilation format 2, and the amount of model data corresponding to compilation format 2 is greater than the amount of model data corresponding to compilation format 3. Furthermore, the second device supports compilation format 1, compilation format 2, and compilation format 3, but the first device only supports compilation format 1 and compilation format 2. In this case, the first compilation format is compilation format 2, that is, the compilation format supported by the first device that corresponds to the smaller amount of model data.

[0203] Of course, in this embodiment of the present application, the first compilation format can be the compilation format with the smallest amount of model data among the one or more compilation formats supported by the second device. Continuing with the above example, assuming the first device also supports compilation formats 1, 2, and 3, the first compilation format is compilation format 3, i.e., the first compilation format is the compilation format with the smallest amount of model data among the one or more compilation formats supported by the second device.

[0204] In some scenarios, the second device may perform compilation format conversion on the model data of the first model, wherein the compilation format conversion may be understood as conversion of the compilation format, that is, the model data of the first model is the model data of the first model after compilation format conversion.

[0205] In an embodiment of the present application, converting the compilation format of the model data of the first model helps to transmit the model data of the first model based on a suitable compilation format, so as to reduce the data volume of the model data of the first model.

[0206] Of course, in this embodiment of the present application, to reduce the complexity of the second device performing the coding format conversion, the second device can pre-store models corresponding to each coding format. In this way, the second device can directly transmit the model of the required coding format without performing the coding format conversion. For example, the second device can store a first model that conforms to coding format 1 and a first model that conforms to coding format 2. If the first coding format is coding format 1, the second device can directly send the first model that conforms to coding format 1 to the first device.

[0207] In some implementations, the second information is associated with the conversion of the compiled format of the first model. For example, the second information is used to indicate that the first model is a model that has undergone compilation format conversion. Accordingly, in other implementations, the second information also includes the first model after the compilation format conversion. In other words, the second information and the first model after the compilation format conversion can be transmitted through the model information of the first model. Of course, in an embodiment of the present application, the second information and the first model after the compilation format conversion can be transmitted through different information.

[0208] For another example, the second information is used to indicate that the model data of the first model is a model that has been converted through a compiled format. Accordingly, in some other implementations, the second information also includes the model data of the first model after the compiled format conversion. In other words, the second information and the model data of the first model after the compiled format conversion can be transmitted through the model information of the first model. Of course, in an embodiment of the present application, the second information and the model data of the first model after the compiled format conversion can be transmitted through different information.

[0209] In some implementations, the second information includes one or more of the following: information indicating whether the model data of the first model has been converted to a compiled format; information indicating the compiled format used by the model data of the first model before the compiled format conversion; information indicating the compiled format used by the model data of the first model after the compiled format conversion; information indicating the compiled format that the model data of the first model should use when running on the first device.

[0210] If the second information includes information for indicating whether the model data of the first model has been converted into a compiled format, then in this embodiment of the present application, the second information may also be referred to as a "format conversion indication."

[0211] Assuming that the second information can occupy one bit, if the value of this bit is a first value, it can indicate that the coding format of the first model has undergone coding format conversion. If the value of this bit is a second value, it can indicate that the coding format of the first model has not undergone coding format conversion. The first value and the second value are different values. For example, the first value can be 0 and the second value can be 1. For another example, the first value can be 1 and the second value can be 0. Of course, in the embodiment of the present application, the second information can occupy multiple bits.

[0212] If the second information includes information used to indicate the compilation format used by the model data of the first model before the compilation format conversion, or in other words, the second information is used to indicate the original compilation format of the first model before the compilation format conversion.

[0213] In some implementations, the second information may indicate the original coding format by carrying indication information (eg, an index) of the original coding format. Of course, in the embodiment of the present application, the second information may carry the name of the original coding format.

[0214] In an embodiment of the present application, indicating the original coding format through the second information helps the first device to know the original coding format of the first model. In some scenarios, the first device can instruct the second device not to perform coding format conversion based on the original coding format of the first model. For example, when the first device runs the first model that conforms to the original coding format, it generally has better performance, or when the first device runs the first model that conforms to the original coding format, it generally has a faster running speed. In this case, the first device can instruct the first device not to perform coding format conversion based on the second information.

[0215] If the second information includes information for indicating the compilation format used by the model data of the first model after the compilation format conversion, in other words, the second information can be used to indicate the compilation format of the first model after the compilation format conversion.

[0216] In some implementations, the second information may carry information indicating the coding format of the first model after the coding format conversion (eg, an index). Of course, in the embodiment of the present application, the second information may carry the name of the above coding format.

[0217] In the embodiment of the present application, the second information indicates the compiled format of the first model after the compiled format conversion, which helps the first device to know the compiled format of the first model after the compiled format conversion. In some scenarios, the first device can run the first model based on the compiled format of the first model after the compiled format conversion.

[0218] Of course, in the embodiments of the present application, the second information can be used alone or in combination. For example, the second information is used to indicate the compilation format of the first model before the compilation format conversion and the compilation format of the first model after the conversion. In other words, the second information is used to indicate the conversion of the compilation format of the first model from the original compilation format to the target compilation format (i.e., the compilation format after the compilation format conversion).

[0219] If the second information includes information for indicating the compilation format that the model data of the first model should use when running on the first device, or in other words, the second information can be used to indicate the compilation format that the model data of the first model should use when running on the first device.

[0220] In some implementations, the compilation format that should be used when the first device is running may be a compilation format that performs better when other devices are running the first model. For example, the compilation format that should be used when the first device is running may be a compilation format that runs faster when other devices are running the first model. For another example, the compilation format that should be used when the first device is running may be a compilation format that requires less storage space when other devices are running the first model.

[0221] In some implementations, the compilation format that the model data of the first model should use when the first device is running (also called the "running compilation format") may be different from the compilation format used when transmitting the model data of the first model (also called the "transmission compilation format"). In this case, the first device can convert the compilation format of the received model data of the first model based on the second information. Therefore, the second information can be used to assist the first device in realizing the conversion between the transmission compilation format and the running compilation format. Of course, in an embodiment of the present application, the running compilation format can be the same as the transmission compilation format. In this case, the first device may not perform the compilation format conversion after receiving the model data of the first model.

[0222] In some implementations, the second information may be information indicating the above-mentioned compilation format (eg, an index). Of course, in the embodiment of the present application, the second information may carry the name of the above-mentioned compilation format.

[0223] Example 4: The model information of the first model includes third information associated with compression of the first model.

[0224] In an embodiment of the present application, the original model can be compressed to obtain a first model, and the third information associated with the compression of the first model can be carried through the model information of the first model. Compared with the traditional model transmission scheme of directly transmitting the original model, it helps to reduce the amount of data of the model information required to be transmitted during the transmission of the first model, so as to reduce the transmission time required for the model transmission process.

[0225] In some implementations, the first model included in the model information of the first model is a compressed model. In other words, the model information of the first model may include the third information as well as the first model. Of course, in the embodiments of the present application, the third information and the first model may also be transmitted via different information.

[0226] In some implementations, the model data of the first model included in the model information of the first model is compressed model data. In other words, the model information of the first model may include the third information as well as the model data of the first model. Of course, in embodiments of the present application, the third information and the model data of the first model may also be transmitted via different information.

[0227] In some implementations, the compression format of the first model includes bitstream compression, wherein the bitstream compression is intended to utilize redundant information in the model data of the model for compression to reduce the number of bits occupied by transmitting the model data of the first model. Common bitstream compression methods may include lossless compression, Huffman coding, predictive coding, and ARITHMETIC algorithm, among others.

[0228] In some implementations, part of the original model information of the first model may be compressed based on bitstream compression to obtain model data of the first model. For example, the original model structure information of the first model may be compressed based on bitstream compression to obtain model structure information of the first model. Of course, in embodiments of the present application, all of the original model information of the first model may be compressed based on bitstream compression.

[0229] In some implementations, different bitstream compression algorithms can be used for different data in the original model information of the first model. For example, Huffman coding can be used to compress the original model structure information of the first model, and the ARITHMETIC algorithm can be used to compress the original model parameters of the first model, which helps to improve the flexibility of information compression. Of course, in the embodiment of the present application, the same bitstream compression algorithm can be used for different data in the original model information of the first model to simplify the complexity of data compression.

[0230] In some implementations, the compression format of the first model includes model compression, where model compression refers to compressing the original model into a smaller first model through knowledge distillation, network pruning, etc. The compressed model (i.e., the "first model") and the original model usually have different model structures. The first model can achieve functions similar to those of the original model, but the performance of the first model is lower than that of the original model.

[0231] In embodiments of the present application, model compression and bitstream compression can be used independently of each other, or they can be used in combination. For example, model compression can be performed on the original model to obtain a compressed model, and then bitstream compression can be continued on the compressed model. In other words, the compression format of the first model is bitstream compression of the first model after model compression.

[0232] In some implementations, the third information includes one or more of the following: information indicating that the model data of the first model is compressed model data; information indicating the performance of the first model (i.e., the compressed model); and information indicating the compression format of the first model.

[0233] If the third information includes information for indicating that the model data of the first model is compressed model data, or in other words, the third information includes information for indicating whether the model data of the first model is compressed model data.

[0234] If the third information includes information for indicating the compression format of the first model, or in other words, the third information includes information for indicating the compression format adopted by the model data of the first model.

[0235] In some implementations, the compression format of the first model may include bitstream compression or model compression. Accordingly, the third information may include information indicating that the compression format corresponding to the model data of the first model is bitstream compression. Alternatively, the third information may include information indicating that the compression format corresponding to the model data of the first model is model compression. Of course, in the embodiment of the present application, the third information may directly carry the name of the compression format.

[0236] In the embodiment of the present application, the indication of the compression format is not limited. In some implementations, the above information can indicate two compression formats through one or more bits. Taking the above information as an example of indicating two compression formats through one bit, if the value of the bit is the first value, it indicates that the compression format corresponding to the model data of the first model is bit stream compression. If the value of the bit is the second value, it indicates that the compression format corresponding to the model data of the first model is model compression. The first value and the second value are different values. For example, the first value can be 1, and for another example, the second value can be 0.

[0237] Taking the above information as an example where two compression formats are indicated by multiple bits, if the values ​​of the multiple bits are 001, it indicates that the model data of the first model corresponds to the model compression format. If the values ​​of the multiple bits are 100, it indicates that the model data of the first model corresponds to bitstream compression.

[0238] In an embodiment of the present application, the information indicating the compression format and the information indicating whether the model data of the first model is compressed may be the same information. For example, if the information indicates that the compression format of the model data of the first model is model compression, then the information may also indicate that the model data of the first model is compressed. Of course, in an embodiment of the present application, the information indicating the compression format and the information indicating whether the model data of the first model is compressed may be different information.

[0239] As previously mentioned, multiple compression algorithms are currently supported for bitstream compression. Therefore, in order to simplify the complexity of decompression by the first device, the compression format of the above-mentioned first model may include the compression algorithm used for bitstream compression. For example, the third information may include information for indicating that the compression format of the first model is lossless compression. For another example, the third information may include information for indicating that the compression format of the first model is Huffman coding. For another example, the third information may include information for indicating that the compression format of the first model is predictive coding. For another example, the third information may include information for indicating that the compression format of the first model is ARITHMETIC algorithm.

[0240] In some implementations, the third information may include information indicating the compression format of the first model. For example, the third information may include an index of the compression format of the first model. For another example, the third information may include an identifier of the compression format of the first model.

[0241] In an embodiment of the present application, the information of the compression format of the first model can be indicated by multiple bits in the model information of the first model. For example, if the value of multiple bits is 001, it means that the compression format is lossless compression. For another example, if the value of multiple bits is 0001, it means that the compression format is predictive coding. In an embodiment of the present application, the number of multiple bits can be determined based on the total number of compression formats to be indicated. Generally speaking, the number of bits is set so that the total number of compression formats that can be indicated by the bits is greater than or equal to the total number of compression formats to be indicated. For example, if the total number of compression formats to be indicated is 5, the number of bits can be 3, so that the 3 bits can indicate 8 compression formats.

[0242] In other embodiments, the compression format of the first model may also be a default compression format, where the default compression format may be, for example, Huffman coding. For example, if the model information of the first model does not indicate the compression format of the first model, the compression format of the first model is the default compression format, i.e., Huffman coding.

[0243] If the third information includes information indicating the performance of the first model, the first device can determine the performance of the first model (ie, the compressed model).

[0244] In some implementations, the information indicating the performance of the first model may include one or more of the following: accuracy, recall, precision, mean square error (MSE), training time, loss function, memory usage, and inference time, etc.

[0245] Accuracy indicates the overall classification or regression accuracy of the first model on the test dataset. Generally speaking, a higher accuracy indicates better performance of the first model.

[0246] Recall and precision. For classification problems, recall refers to the ratio of the number of samples correctly predicted as positive by the first model to the total number of actual positive samples; precision refers to the ratio of the number of samples correctly predicted as positive by the first model to the total number of predicted positive samples. Generally speaking, higher recall and precision indicate better performance of the first model.

[0247] Mean Squared Error (MSE) is a metric used to measure the average difference between the predicted values ​​of the first model and the true values ​​for regression problems. Generally speaking, the smaller the MSE, the better the performance of the first model.

[0248] A loss function is a function used to measure the difference between the first model's prediction and the true value. Common loss functions include mean squared error loss function and cross entropy loss function. The performance of the first model can be improved by optimizing the loss function.

[0249] Training time refers to the time required to train the first model. Generally speaking, the shorter the training time, the better the performance of the first model.

[0250] Memory usage refers to the amount of memory required by the first model during runtime. Generally speaking, the smaller the memory usage, the better the performance of the first model.

[0251] Inference time refers to the time it takes for the first model to perform inference or prediction. Generally speaking, shorter inference time indicates better performance of the first model.

[0252] In the embodiment of the present application, the scheme of converting the model data of the first model into a compiled format and the scheme of compressing the first model can be used independently or in combination. If the above two schemes are used in combination with each other, the model data of the first model involved in the model compression scheme above can be converted into a compiled format, wherein the compiled format conversion can refer to the introduction of the second information above.

[0253] In some scenarios, before the second device sends the model information of the first model, the first device may send sixth information to the second device to assist the second device in sending the model information of the first model to the first device, thereby helping to improve the rationality of the second device sending the model information of the first model to the first device. That is, as shown in FIG5 , before S410 , the method further includes: S510 , the first device sends the sixth information to the second device.

[0254] In some implementations, the sixth information is used to indicate one or more of the following: capability information associated with the first model; time information when the first device requires the first model; model information stored in the first device; performance requirements of the first model for the first device; whether the first device supports model compression; information associated with data arrival in the first device; location information of the first device; information about the serving cell of the first device; information about the neighboring cell of the first device; and buffer status report (BSR) information of the first device.

[0255] Taking the sixth information as an example, which indicates capability information associated with the first model, the capability information associated with the first model can also be referred to as capability information of the first device, wherein the capability information of the first device can be associated with the first model. The capability information of the first device in the embodiment of the present application is described below.

[0256] In some implementations, the capability information of the first device may indicate whether the first device has dedicated hardware that can be used to accelerate model inference, where the dedicated hardware may include, for example, a computing unit such as a GPU or an ASIC.

[0257] In some implementations, the capability information of the first device may indicate the memory space that the first device can allocate to storing models. Generally speaking, the larger the memory space that can be allocated to models, the more models the first device can store.

[0258] In some implementations, the capability information of the first device may indicate the cache space that the first device can allocate to model inference. The larger the cache space, the larger the model size and the greater the number of models that the first device can run.

[0259] In some implementations, the capability information of the first device may indicate the computing power allocated by the first device to model inference. Generally speaking, greater computing power that can be allocated to the model indicates that the first device can run larger models and a greater number of models.

[0260] In some implementations, the capability information of the first device may indicate the amount of power that the first device can allocate to model inference. Generally speaking, more power that can be allocated to model inference indicates that the first device can run larger models or a greater number of models.

[0261] In some implementations, the capability information of the first device may indicate the remaining memory space available for model storage on the first device. In some scenarios, part of the memory space allocated by the first device for model storage may already be occupied by models, leaving the remaining memory space available for new models. Therefore, indicating the remaining memory space helps the second device indicate the first model that matches the remaining memory space.

[0262] In some implementations, the capability information of the first device may indicate the remaining cache space available for model inference on the first device. In some scenarios, a portion of the cache space allocated by the first device for model inference may already be occupied by a model, leaving the remaining cache space available for use by a new model. Therefore, indicating the remaining cache space helps the second device indicate a first model that matches the remaining cache space.

[0263] In some implementations, the capability information of the first device may indicate the remaining computing power available for model inference on the first device. In some scenarios, some of the computing power allocated to model inference by the first device may already be used by models, leaving the remaining computing power available for new models. Therefore, indicating the remaining computing power helps the second device indicate a first model that matches the remaining computing power.

[0264] In some implementations, the capability information of the first device may indicate the remaining power available for model inference on the first device. In some scenarios, some of the power allocated by the first device for model inference may already be used by the model, leaving the remaining power available for the new model. Therefore, indicating the remaining power helps the second device indicate the first model that matches the remaining power.

[0265] In some implementations, the capability information of the first device may indicate the compilation formats supported by the first device, which helps the second device indicate to the first device a first model that conforms to the compilation format supported by the first device. As previously mentioned, the compilation format is the basic operating environment of the model. If the compilation format required by the model is not among the compilation formats supported by the first device, the first device cannot infer the model. For example, if the first device only supports the PyTorch compilation format, but the first model indicated by the second device is in the TensorFlow format, the first model cannot be run on the first device.

[0266] In some scenarios, the sixth information may include stored model-related information, which will be described below with reference to specific examples.

[0267] Taking the sixth information used to indicate the time information when the first device needs the first model as an example, in some implementations, the above time information may be the time when the first device needs to obtain the model inference result. Accordingly, the time information when the first device needs the first model can be determined based on the time when the first device needs to obtain the model inference result.

[0268] The embodiments of the present application do not limit the manner in which the above-mentioned time information is indicated. In some implementations, the above-mentioned time information may be represented by universal time. For example, the above-mentioned time information is 23:12:34:23 milliseconds on July 25, 2023. In other implementations, the above-mentioned time information may be represented by time units, where time units may include, for example, frames, subframes, time slots, OFDM symbols, etc. For example, the above-mentioned time information is the 5th OFDM symbol in the 2nd time slot in subframe 1.

[0269] In other implementations, the time information may be time information requiring the first model, wherein the representation of the time information may refer to the above description.

[0270] In some other implementations, the sixth information may include indication information of a time offset, where the time offset is a time offset between a time when the first device needs the first model information and a current time. The current time may be a time when the first device generates the sixth information.

[0271] The embodiments of the present application do not limit the manner in which the time offset is indicated. In some implementations, the time offset may be expressed as time. For example, the time offset may be 1 second. In other implementations, the time offset may be expressed as a time unit, where a time unit may include, for example, a frame, a subframe, a time slot, or an OFDM symbol. For example, the time offset may be 10 OFDM symbols. In another example, the time offset may be 1 subframe.

[0272] Taking the sixth information used to indicate the model-related information stored in the first device as an example, the stored model-related information may include one or more of the following: identification information of the stored model; the number of stored models; the stored model structure identification; the number of stored model structures; and the stored model module information.

[0273] In some implementations, the stored model module information may include one or more of the following: stored model module identifiers; and the number of stored model modules.

[0274] In an embodiment of the present application, the first device can send stored model module information to the second device, which helps the second device determine whether the stored model information can be reused to achieve the purpose of transmitting the first model, so as to reduce the amount of model data transmitted for the first model.

[0275] For example, the stored model information is 001, which indicates a hidden layer corresponding to 3 layers with 100 nodes in each layer. For another example, the stored model information is 010, which indicates a hidden layer corresponding to 2 layers with 50 nodes in each layer.

[0276] For example, if the model structure information of a stored model is 001, it indicates an LSTM structure. For another example, if the model structure information of a stored model is 010, it indicates a transformer structure.

[0277] In an embodiment of the present application, the first device can send stored model-related information to the second device, which helps the second device determine whether the stored model-related information can be reused to achieve the purpose of transmitting the first model, so as to reduce the amount of model data transmitted for the first model.

[0278] For example, the sixth information is used to indicate the model performance requirement of the first model by the first device. For the model performance of the first model, please refer to the previous description. For example, if the model performance is accuracy, assuming the first model is used for positioning, the sixth information is used to indicate that the model performance requirement of the first model by the first device is a positioning error of less than or equal to 0.5m.

[0279] Taking the sixth information used to indicate whether the first device supports model compression as an example, as mentioned above, model compression may cause the performance of the model to degrade. At this time, if the first device cannot accept the performance degradation caused by model compression, the first device does not support model compression.

[0280] In other scenarios, the sixth information may include information related to the first device, which will be described below with reference to specific examples.

[0281] Take the example of the sixth information being used to indicate information associated with the arrival of data in the first device, where the data may be data to be processed by the first model.

[0282] In some implementations, the information associated with data arrival in the first device may include information associated with data arrival at the current moment and / or information associated with data arrival at a historical moment. The information associated with data arrival may include data arrival time and / or data arrival amount.

[0283] For example, information associated with data arrival may include one or more of the following information: the amount of data arrival at the current moment; the amount of data arrival at M consecutive sampling points before the current moment, where M is a positive integer; the data arrival time at M consecutive sampling points before the current moment, where M is a positive integer; the amount of data arrival in the second time window before the current moment; the data arrival time in the second time window before the current moment.

[0284] Taking the sixth information indicating the location information of the first device as an example, in some implementations, the location information of the first device can be used to indicate the location of the first device. For example, the location information of the first device can include one or more of the following: the location coordinates of the first device, and the relative position with respect to a reference point (e.g., a network device).

[0285] In some implementations, the location information of the first device can be used to indicate the movement of the first device. For example, the location information of the first device can include one or more of the following information: multiple locations of the first device over a period of time; N consecutive locations of the first device, where N is a positive integer; the movement speed of the first device; or the displacement of the first device over a period of time.

[0286] In some implementations, if the location information of the first device includes multiple location information, the location information of the first device can be used to indicate the movement of the first device, for example, the movement speed of the first device, or for example, the movement direction of the first device.

[0287] In some implementations, the location information of the first device may include the current location information of the first device and / or the location information of the first device at a historical moment. For example, the location information of the first device may include one or more of the following information of the first device: the current location; N consecutive location points before the current moment, where N is a positive integer; location information within a second time window before the current moment, etc.

[0288] Taking the sixth information as an example of information indicating the serving cell of the first device, in some implementations, the serving cell information may be related to the serving cell. For example, the serving cell information may include one or more of the following: an identifier of the serving cell, a signal quality of the serving cell. Among them, the identifier of the serving cell may include a physical cell identifier (physical cell ID, PCI). The signal quality of the serving cell may include one or more of the following: reference signal receiving power (RSRP), reference signal receiving quality (RSRQ), and signal-to-interference-plus-noise ratio (SINR).

[0289] In an embodiment of the present application, the serving cell information may include current information of the serving cell and / or historical information of the serving cell. Taking signal quality as an example, the serving cell information may include one or more of the following information of the serving cell: signal quality at the current moment; signal quality of P consecutive sampling points before the current moment, where P is a positive integer; and signal quality within a second time window before the current moment.

[0290] Taking the sixth information indicating information about a neighboring cell of the first device as an example, in some implementations, the neighboring cell information may be related to the neighboring cell of the first device. For example, the neighboring cell information may include one or more of the following information: a neighboring cell identifier; and a neighboring cell signal quality. The neighboring cell identifier may include the PCI of the neighboring cell, and the neighboring cell signal quality may include one or more of the following: RSRP, RSRQ, and SINR.

[0291] In the embodiments of the present application, neighboring cell information may include current information and / or historical information of neighboring cells. Taking signal quality as an example, neighboring cell information may include one or more of the following information about the neighboring cell: signal quality at the current moment; signal quality at Q consecutive sampling points before the current moment, where Q is a positive integer; and signal quality within a second time window before the current moment.

[0292] Taking the sixth information indicating the buffer status report (BSR) information of the first device as an example, in some implementations, the BSR-related information may include current BSR information and / or historical BSR information. For example, the BSR-related information may include one or more of the following information: the BSR at the current moment; the BSRs of R consecutive sampling points before the current moment, where R is a positive integer; and the BSR within a second time window before the current moment.

[0293] It should be noted that if the sixth information may include multiple pieces of information, these multiple pieces of information may be reported together or separately. For example, the information about the compilation formats supported by the first device and the time information when the first device requires the first model may be included in a single RRC message for transmission. For another example, the information about the compilation formats supported by the first device and the time information when the first device requires the first model may be stored in two separate RRC messages and transmitted separately.

[0294] In some scenarios, the sixth information may be autonomously reported by the first device. For example, after the first device switches to a new cell, the first device may send the sixth information to the second device. In other scenarios, the sixth information may be sent by the second device via the seventh information. Alternatively, the sixth information may be requested by the second device via the seventh information.

[0295] 5 , before step S510 , the method further includes step S520 , where the second device sends seventh information to the first device, wherein the seventh information is associated with the transmission of the sixth information.

[0296] In some implementations, the second device may proactively send the seventh information, for example, the second device may periodically send the seventh information. In other implementations, the seventh information may be triggered based on the first event, that is, in response to the occurrence of the first event, the second device may send the seventh information to the first device.

[0297] In some implementations, the first event may include synchronization between the first device and the second device. For example, taking the second device as a network device and the first device as a terminal device, synchronization between the first device and the second device may include downlink synchronization between the network device and the terminal device. In other implementations, the first event may include the second device receiving a reporting request sent by the first device, the reporting request being used to request transmission of the sixth information.

[0298] In some implementations, the seventh information includes one or more of the following: information for indicating the sending of the sixth information; information for indicating the sending method of the sixth information; and information for indicating the information content carried in the sixth information.

[0299] Taking the example that the seventh information includes information for indicating the sending of the sixth information, the information for indicating the sending of the sixth information can be replaced by information for indicating whether to send the sixth information. In some implementations, the information for indicating the sending of the sixth information can occupy one bit in the seventh information. If the value of the bit is the first value, it is used to indicate the information for sending the sixth information. If the value of the bit is the second value, it is used to indicate the information for not sending the sixth information. The first value and the second value are different. 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 information for indicating the sending of the sixth information can occupy multiple bits in the seventh information.

[0300] Taking the example of the seventh information including information for indicating the sending method of the sixth information, the sending method of the sixth information may include single reporting and / or periodic reporting. Accordingly, the information for indicating the sending method of the sixth information is used to indicate whether the sixth information is a single reporting or a periodic reporting. Among them, single reporting can be understood as the first device sending the sixth information to the second device once each time it receives the seventh information, which helps to improve the flexibility of transmitting the sixth information. Periodic reporting can be understood as the first device periodically reporting the sixth information, which helps to reduce the resources used to transmit the sixth information.

[0301] In some implementations, if the sixth information is periodically reported, the seventh information may also carry parameters for determining the periodic reporting. For example, the seventh information may include the period length, the number of periodic transmissions, the start time of the period, etc. Of course, the parameters for determining the periodic reporting may also be preconfigured or predefined.

[0302] Taking the example of the seventh information including information indicating the information content carried in the sixth information, the information content carried in the sixth information can be referred to as described above. In an embodiment of the present application, the seventh information can indicate the information content carried in the sixth information, which helps to improve the rationality of transmitting the sixth information. Of course, in an embodiment of the present application, the information content carried in the sixth information can also be predefined or preconfigured, in which case the seventh information may not include information indicating the information content carried in the sixth information.

[0303] It should be noted that if the seventh information may include multiple pieces of information, these multiple pieces of information may be reported together or separately. For example, the information indicating the content of the sixth information and the information indicating the method for sending the sixth information may be included in a single message for transmission. For another example, the information indicating the content of the sixth information and the information indicating the method for sending the sixth information may be stored in two separate messages for transmission.

[0304] In an embodiment of the present application, any information involved above (for example, the model information of the first model, the sixth information) can be carried by one or more of the following message types: NR Positioning Protocol A (NRPPa) message, LPP message, non-access stratum (NAS) message, RRC message, media access control control element (MAC CE), downlink control information (DCI), uplink control information (UCI), physical uplink control channel (PUCCH), physical uplink shared channel (PUSCH), inter-node message, Xn interface message, F1 interface message, E1 interface message, NG interface message, core network service architecture-based message or AI dedicated message.

[0305] In an embodiment of the present application, any of the information mentioned above can be carried by one or more of unicast messages, multicast messages and broadcast messages.

[0306] Unicast messages can be understood as one-to-one transmissions, meaning they are sent from one sender to one receiver. The source delivers unicast messages over a unicast channel, and only devices or network equipment allocated the corresponding unicast resources can attempt to receive them. Unicast messages are also called dedicated signaling.

[0307] The multicast message described above can be understood as a one-to-many transmission, meaning it's sent from a single sender to multiple receivers. The source transmits the message via a multicast channel. Terminal devices or network devices within the multicast signal's coverage area and that are members of the group can attempt to receive the message. When a terminal device or network device joins a group, it acquires multicast channel resources.

[0308] The broadcast message can be understood as a message transmitted from one sender to any receiver. The source transmits the message via a broadcast channel, and any terminal or network device within the coverage area of ​​the broadcast signal can attempt to receive it.

[0309] As previously described, the second device can be a network device. In some implementations, 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 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).

[0310] Exemplarily, the core network device is any one of the following: 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 analytics function (NWDAF) network element.

[0311] The method embodiment of the present application is described in detail above in conjunction with Figures 1 to 5 , and the device embodiment of the present application is described in detail below in conjunction with Figures 6 to 8 . It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, and therefore, for portions not described in detail, reference can be made to the above method embodiment.

[0312] FIG6 is a schematic diagram of a communication device according to an embodiment of the present application. The communication device 600 shown in FIG6 is a first device. The communication device 600 includes a receiving unit 610 .

[0313] The receiving unit 610 is used to receive model information of the first model sent by the second device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information used to determine the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[0314] In some implementations, the model information of the first model includes the first information, and the first information includes one or more of the following: a first identifier, the first identifier is the model identifier of the second model; a second identifier, the second identifier is the model structure identifier of the second model; a third identifier, the third identifier is the model module identifier of the first model; fourth information is associated with the arrangement position of the model module in the first model; fifth information includes module data of some model modules in the first model; wherein, the second model is the first model, or the first model is determined based on the second model.

[0315] In some implementations, the model information of the first model includes the first identifier and some or all model parameters of the first model.

[0316] In some implementations, the model information of the first model includes the second identifier and some or all model parameters of the first model.

[0317] In some implementations, the model information of the first model includes the third identifier and some or all model parameters of the first model.

[0318] In some implementations, the model information of the first model also includes the fourth information.

[0319] In some implementations, the first model is determined based on the second model, and the model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the second model; or, the model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module that needs to be added in the second model; or, the model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model; or, the model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added in the model structure of the second model.

[0320] In some implementations, the first model is determined based on the second model. If the first device only stores one second model, the model module associated with the third identifier is the model module that needs to be deleted from the second model; or, if the first device only stores one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, if the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be deleted from the model structure of the second model; or, if the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

[0321] In some implementations, the model information of the first model includes a target identifier, some or all model parameters of the first model, and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier, and the third identifier.

[0322] In some implementations, the model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating the compilation format of the model data of the first model; information for indicating whether the model data of the first model has undergone compilation format conversion; information for indicating the compilation format used by the model data of the first model before compilation format conversion; information for indicating the compilation format used by the model data of the first model after compilation format conversion; information for indicating the compilation format that the model data of the first model should use when the first device is running.

[0323] In some implementations, the second information is used to indicate that the model data of the first model is model data converted through a compiled format, and the model data of the first model is the model data of the first model after the compiled format conversion.

[0324] In some implementations, the second information is used to indicate that the first model is a model that has been converted from a compiled format, and the second information also includes the first model after the conversion from the compiled format.

[0325] In some implementations, the compilation format associated with the second information is a first compilation format, and the first compilation format belongs to multiple compilation formats supported by the first device. Among the multiple compilation formats, the first compilation format corresponds to the smallest amount of model data.

[0326] In some implementations, the model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

[0327] In some implementations, the model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

[0328] In some implementations, the model data of the first model is converted into a compiled format.

[0329] In some implementations, the third information includes one or more of the following: information indicating that the model data of the first model is compressed model data; information indicating the performance of the first model; and information indicating the compression format of the first model.

[0330] In some implementations, the compression format of the first model includes bitstream compression and / or model compression.

[0331] In some implementations, the compression format of the first model is bitstream compression of the first model after model compression.

[0332] In some implementations, the model data of the first model includes one or more of the following model information of the first model: model structure data of the first model; model module data of the first model; and part or all of the model parameters of the first model.

[0333] In some implementations, the communication device further includes: a sending unit for sending sixth information to the second device, wherein the sixth information is used to indicate one or more of the following: capability information associated with the first model; time information when the first device requires the first model; model-related information stored in the first device; performance requirements of the first model by the first device; whether the first device supports model compression; information associated with data arrival in the first device; location information of the first device; information of the service cell of the first device; information of the neighboring cell of the first device; and buffer status report BSR information of the first device.

[0334] In some implementations, the sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: whether the first device has or does not have the capability to accelerate model reasoning; the storage space available for model reasoning in the first device; the computing power available for model reasoning in the first device; the amount of electricity available for model reasoning in the first device; and the model compression format supported by the first device.

[0335] In some implementations, the sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

[0336] In some implementations, the receiving unit is configured to receive seventh information sent by the second device, where the seventh information is associated with the transmission of the sixth information.

[0337] In some implementations, the seventh information includes one or more of the following: information for indicating the sending of the sixth information; information for indicating the sending method of the sixth information; and information for indicating the information content carried in the sixth information.

[0338] In some implementations, the first device is a terminal device, and the second device is a network device.

[0339] FIG7 is a schematic diagram of a communication device according to an embodiment of the present application. The communication device 700 shown in FIG7 is a second device, and the communication device 700 includes a sending unit 710 .

[0340] The sending unit 710 is used to send model information of the first model to the first device, wherein the model information of the first model includes one or more of the following: part or all of the model parameters of the first model; first information for determining the model structure of the first model; second information associated with the compilation format of the first model; and third information associated with the compression of the first model.

[0341] In some implementations, the model information of the first model includes the first information, and the first information includes one or more of the following: a first identifier, the first identifier is the model identifier of the second model; a second identifier, the second identifier is the model structure identifier of the second model; a third identifier, the third identifier is the model module identifier of the first model; fourth information is associated with the arrangement position of the model module in the first model; fifth information includes module data of some model modules in the first model; wherein, the second model is the first model, or the first model is determined based on the second model.

[0342] In some implementations, the model information of the first model includes the first identifier and some or all model parameters of the first model.

[0343] In some implementations, the model information of the first model includes the second identifier and some or all model parameters of the first model.

[0344] In some implementations, the model information of the first model includes the third identifier and some or all model parameters of the first model.

[0345] In some implementations, the model information of the first model also includes the fourth information.

[0346] In some implementations, the first model is determined based on the second model, and the model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the second model; or, the model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module that needs to be added in the second model; or, the model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model; or, the model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added in the model structure of the second model.

[0347] In some implementations, the first model is determined based on the second model. If the first device only stores one second model, the model module associated with the third identifier is the model module that needs to be deleted from the second model; or, if the first device only stores one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, if the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be deleted from the model structure of the second model; or, if the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

[0348] In some implementations, the model information of the first model includes a target identifier, some or all model parameters of the first model, and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier, and the third identifier.

[0349] In some implementations, the model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating the compilation format of the model data of the first model; information for indicating whether the model data of the first model has undergone compilation format conversion; information for indicating the compilation format used by the model data of the first model before compilation format conversion; information for indicating the compilation format used by the model data of the first model after compilation format conversion; information for indicating the compilation format that the model data of the first model should use when the first device is running.

[0350] In some implementations, the second information is used to indicate that the model data of the first model is model data converted through a compiled format, and the model data of the first model is the model data of the first model after the compiled format conversion.

[0351] In some implementations, the second information is used to indicate that the first model is a model that has been converted from a compiled format, and the second information also includes the first model after the conversion from the compiled format.

[0352] In some implementations, the compilation format associated with the second information is a first compilation format, and the first compilation format belongs to multiple compilation formats supported by the first device. Among the multiple compilation formats, the first compilation format corresponds to the smallest amount of model data.

[0353] In some implementations, the model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

[0354] In some implementations, the model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

[0355] In some implementations, the model data of the first model is converted into a compiled format.

[0356] In some implementations, the third information includes one or more of the following: information indicating that the model data of the first model is compressed model data; information indicating the performance of the first model; and information indicating the compression format of the first model.

[0357] In some implementations, the compression format of the first model includes bitstream compression and / or model compression.

[0358] In some implementations, the compression format of the first model is bitstream compression of the first model after model compression.

[0359] In some implementations, the model data of the first model includes one or more of the following model information of the first model: model structure data of the first model; model module data of the first model; and part or all of the model parameters of the first model.

[0360] In some implementations, the communication device further includes: a receiving unit for receiving sixth information sent by the first device, wherein the sixth information is used to indicate one or more of the following: capability information associated with the first model; time information when the first device requires the first model; model-related information stored in the first device; performance requirements of the first model by the first device; whether the first device supports model compression; information associated with data arrival in the first device; location information of the first device; information on the service cell of the first device; information on the neighboring cell of the first device; and buffer status report BSR information of the first device.

[0361] In some implementations, the sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: whether the first device has or does not have the capability to accelerate model reasoning; the storage space available for model reasoning in the first device; the computing power available for model reasoning in the first device; the amount of electricity available for model reasoning in the first device; and the model compression format supported by the first device.

[0362] In some implementations, the sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

[0363] In some implementations, the sending unit is configured to send seventh information to the first device, where the seventh information is associated with the transmission of the sixth information.

[0364] In some implementations, the seventh information includes one or more of the following: information for indicating the sending of the sixth information; information for indicating the sending method of the sixth information; and information for indicating the information content carried in the sixth information.

[0365] In some implementations, the first device is a terminal device, and the second device is a network device.

[0366] In an optional embodiment, the receiving unit 610 may be a transceiver 830. The communication device 600 may further include a transceiver 830 and a processor 810, as specifically shown in FIG8 .

[0367] In an optional embodiment, the sending unit 710 may be a transceiver 830. The communication device 700 may further include a transceiver 830 and a processor 810, as specifically shown in FIG8 .

[0368] Figure 8 is a schematic block diagram of a communication device according to an embodiment of the present application. The dashed lines in Figure 8 indicate that the unit or module is optional. The device 800 may be used to implement the method described in the above method embodiment. The device 800 may be a chip, a terminal device, or a network device.

[0369] The device 800 may include one or more processors 810. The processor 810 may support the device 800 to implement the method described in the method embodiment above. The processor 810 may be a general-purpose processor or a special-purpose processor. For example, the processor may be a central processing unit (CPU). Alternatively, the processor may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0370] The apparatus 800 may further include one or more memories 820. The memories 820 store programs that can be executed by the processor 810, causing the processor 810 to perform the methods described in the above method embodiments. The memories 820 may be independent of the processor 810 or integrated into the processor 810.

[0371] The apparatus 800 may further include a transceiver 830. The processor 810 may communicate with other devices or chips via the transceiver 830. For example, the processor 810 may transmit and receive data with other devices or chips via the transceiver 830.

[0372] 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.

[0373] 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.

[0374] 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.

[0375] 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.

[0376] 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.

[0377] 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.

[0378] 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.

[0379] 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.

[0380] 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.

[0381] 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.

[0382] 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.

[0383] 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.

[0384] 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.

[0385] 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.

[0386] 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)).

[0387] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for model transmission, characterized in that: include: The first device receives model information of a first model sent by the second device, where the model information of the first model includes one or more of the following: Some or all of the model parameters of the first model; first information for determining a model structure of the first model; second information associated with the compiled format of the first model; Third information associated with the compression of the first model.

2. The method according to claim 1, characterized in that The model information of the first model includes the first information, and the first information includes one or more of the following: A first identifier, where the first identifier is a model identifier of the second model; A second identifier, where the second identifier is a model structure identifier of a second model; A third identifier, wherein the third identifier is a model module identifier of the first model; Fourth information, associated with the arrangement position of the model module in the first model; fifth information, including module data of some model modules in the first model; The second model is the first model, or the first model is determined based on the second model.

3. The method according to claim 2, characterized in that The model information of the first model includes the first identifier and part or all of the model parameters of the first model.

4. The method according to claim 2, characterized in that: The model information of the first model includes the second identifier and part or all of the model parameters of the first model.

5. The method according to claim 2, characterized in that: The model information of the first model includes the third identifier and part or all of the model parameters of the first model.

6. The method according to claim 5, characterized in that The model information of the first model also includes the fourth information.

7. The method according to claim 5 or 6, characterized in that: The first model is determined based on the second model, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be deleted in the second model; or, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be added to the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module to be deleted in the model structure of the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

8. The method according to claim 5 or 6, characterized in that: The first model is determined based on the second model, If the first device stores only one second model, the model module associated with the third identifier is the model module in the second model that needs to be deleted; or, If the first device stores only one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, If the first device stores only one model structure of the second model, the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model; or, If the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

9. The method according to claim 2, characterized in that: The model information of the first model includes a target identifier, part or all of the model parameters of the first model and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier and the third identifier.

10. The method according to any one of claims 1 to 9, characterized in that The model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating a compilation format of model data of the first model; Information indicating whether the model data of the first model has been converted into a compilation format; Information used to indicate a compilation format used by the model data of the first model before the compilation format is converted; Information used to indicate a compilation format used by the model data of the first model after the compilation format conversion; Information used to indicate a compilation format that should be used when the model data of the first model is run on the first device.

11. The method according to claim 10, characterized in that The second information is used to indicate that the model data of the first model is model data converted through a compilation format, and the model data of the first model is model data of the first model after conversion through a compilation format.

12. The method according to claim 1, characterized in that The second information is used to indicate that the first model is a model that has been converted into a compiled format, and the second information also includes the first model after the conversion into a compiled format.

13. The method according to any one of claims 1 to 12, characterized in that The compilation format associated with the second information is a first compilation format, the first compilation format belongs to multiple compilation formats supported by the first device, and the model data amount corresponding to the first compilation format is the smallest among the multiple compilation formats.

14. The method according to claim 1, characterized in that The model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

15. The method according to any one of claims 1 to 13, characterized in that The model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

16. The method according to claim 14 or 15, characterized in that The model data of the first model is converted into a compiled format.

17. The method according to claim 15 or 16, characterized in that The third information includes one or more of the following: used to indicate that the model data of the first model is compressed model data; information indicating performance of the first model; Information indicating a compression format of the first model.

18. The method according to any one of claims 14 to 17, characterized in that: The compression format of the first model includes bit stream compression and / or model compression.

19. The method according to claim 18, characterized in that The compression format of the first model is to perform bit stream compression on the first model after model compression.

20. The method according to any one of claims 10-11 and 15-17, characterized in that: The model data of the first model includes one or more of the following among the model information of the first model: Model structure data of the first model; Model module data of the first model; Some or all of the model parameters of the first model.

21. The method according to any one of claims 1 to 20, characterized in that Before the first device receives the model information of the first model sent by the second device, the method further includes: The first device sends sixth information to the second device, where the sixth information is used to indicate one or more of the following: capability information associated with the first model; The first device requires time information of the first model; Model-related information stored in the first device; performance requirements of the first device for the first model; whether the first device supports model compression; information associated with the arrival of data in the first device; location information of the first device; Information about a serving cell of the first device; Information about neighboring cells of the first device; The first device sends buffer status report (BSR) information.

22. The method according to claim 21, characterized in that The sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: The first device has or does not have a capability for accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; The model compression format supported by the first device.

23. The method according to claim 21 or 22, characterized in that The sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

24. The method according to any one of claims 21 to 23, characterized in that Before the first device sends sixth information to the second device, the method further includes: The first device receives seventh information sent by the second device, where the seventh information is associated with transmission of the sixth information.

25. The method according to claim 24, characterized in that The seventh information includes one or more of the following: Information used to instruct sending the sixth information; Information used to indicate the sixth information sending method; Information used to indicate the information content carried in the sixth information.

26. The method according to any one of claims 1 to 25, characterized in that The first device is a terminal device, and the second device is a network device.

27. A method for model transmission, characterized in that: include: The second device sends model information of the first model to the first device, where the model information of the first model includes one or more of the following: Some or all of the model parameters of the first model; first information for determining a model structure of the first model; second information associated with the compiled format of the first model; Third information associated with the compression of the first model.

28. The method according to claim 27, characterized in that The model information of the first model includes the first information, and the first information includes one or more of the following: A first identifier, where the first identifier is a model identifier of the second model; A second identifier, where the second identifier is a model structure identifier of a second model; A third identifier, wherein the third identifier is a model module identifier of the first model; Fourth information, associated with the arrangement position of the model module in the first model; fifth information, including module data of some model modules in the first model; The second model is the first model, or the first model is determined based on the second model.

29. The method according to claim 27, characterized in that The model information of the first model includes the first identifier and part or all of the model parameters of the first model.

30. The method according to claim 27, characterized in that The model information of the first model includes the second identifier and part or all of the model parameters of the first model.

31. The method according to claim 27, characterized in that The model information of the first model includes the third identifier and part or all of the model parameters of the first model.

32. The method according to claim 31, characterized in that The model information of the first model also includes the fourth information.

33. The method according to claim 31 or 32, characterized in that The first model is determined based on the second model, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be deleted in the second model; or, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be added to the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module to be deleted in the model structure of the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

34. The method according to claim 31 or 32, characterized in that The first model is determined based on the second model, If the first device stores only one second model, the model module associated with the third identifier is the model module in the second model that needs to be deleted; or, If the first device stores only one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, If the first device stores only one model structure of the second model, the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model; or, If the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

35. The method according to claim 28, characterized in that The model information of the first model includes a target identifier, part or all of the model parameters of the first model and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier and the third identifier.

36. The method according to any one of claims 27 to 35, characterized in that The model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating a compilation format of model data of the first model; Information indicating whether the model data of the first model has been converted into a compilation format; Information used to indicate a compilation format used by the model data of the first model before the compilation format is converted; Information used to indicate a compilation format used by the model data of the first model after the compilation format conversion; Information used to indicate a compilation format that should be used when the model data of the first model is run on the first device.

37. The method according to claim 36, characterized in that The second information is used to indicate that the model data of the first model is model data converted through a compilation format, and the model data of the first model is model data of the first model after conversion through a compilation format.

38. The method according to claim 27, characterized in that The second information is used to indicate that the first model is a model that has been converted into a compiled format, and the second information also includes the first model after the conversion into a compiled format.

39. The method according to any one of claims 27 to 38, characterized in that The compilation format associated with the second information is a first compilation format, the first compilation format belongs to multiple compilation formats supported by the first device, and the model data amount corresponding to the first compilation format is the smallest among the multiple compilation formats.

40. The method according to claim 27, characterized in that The model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

41. The method according to any one of claims 27 to 39, characterized in that The model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

42. The method according to claim 40 or 41, characterized in that The model data of the first model is converted into a compiled format.

43. The method according to claim 41 or 42, characterized in that The third information includes one or more of the following: used to indicate that the model data of the first model is compressed model data; information indicating performance of the first model; Information indicating a compression format of the first model.

44. The method according to any one of claims 40 to 43, characterized in that The compression format of the first model includes bit stream compression and / or model compression.

45. The method according to claim 44, characterized in that The compression format of the first model is to perform bit stream compression on the first model after model compression.

46. ​​The method according to any one of claims 36-37 and 41-43, characterized in that The model data of the first model includes one or more of the following among the model information of the first model: Model structure data of the first model; Model module data of the first model; Some or all of the model parameters of the first model.

47. The method according to any one of claims 27 to 46, characterized in that Before the first device receives the model information of the first model sent by the second device, the method further includes: The second device receives sixth information sent by the first device, where the sixth information is used to indicate one or more of the following: capability information associated with the first model; The first device requires time information of the first model; Model-related information stored in the first device; performance requirements of the first device for the first model; whether the first device supports model compression; information associated with the arrival of data in the first device; location information of the first device; Information about a serving cell of the first device; Information about neighboring cells of the first device; The first device sends buffer status report (BSR) information.

48. The method according to claim 47, characterized in that The sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: The first device has or does not have a capability for accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; The model compression format supported by the first device.

49. The method according to claim 47 or 48, characterized in that The sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

50. The method according to any one of claims 47 to 49, characterized in that Before the first device sends sixth information to the second device, the method further includes: The second device sends seventh information to the first device, where the seventh information is associated with the transmission of the sixth information.

51. The method according to claim 50, characterized in that The seventh information includes one or more of the following: Information used to instruct sending the sixth information; Information used to indicate the sixth information sending method; Information used to indicate the information content carried in the sixth information.

52. The method according to any one of claims 27 to 51, characterized in that The first device is a terminal device, and the second device is a network device.

53. A communication device, characterized in that: The communication device is a first device, and the communication device includes: A receiving unit, configured to receive model information of a first model sent by a second device, wherein the model information of the first model includes one or more of the following: Some or all of the model parameters of the first model; first information for determining a model structure of the first model; second information associated with the compiled format of the first model; Third information associated with the compression of the first model.

54. The communication device according to claim 53, characterized in that The model information of the first model includes the first information, and the first information includes one or more of the following: A first identifier, where the first identifier is a model identifier of the second model; A second identifier, where the second identifier is a model structure identifier of a second model; A third identifier, wherein the third identifier is a model module identifier of the first model; Fourth information, associated with the arrangement position of the model module in the first model; fifth information, including module data of some model modules in the first model; The second model is the first model, or the first model is determined based on the second model.

55. The communication device according to claim 54, characterized in that The model information of the first model includes the first identifier and part or all of the model parameters of the first model.

56. The communication device according to claim 54, characterized in that The model information of the first model includes the second identifier and part or all of the model parameters of the first model.

57. The communication device according to claim 54, characterized in that The model information of the first model includes the third identifier and part or all of the model parameters of the first model.

58. The communication device according to claim 57, characterized in that The model information of the first model also includes the fourth information.

59. The communication device according to claim 57 or 58, characterized in that The first model is determined based on the second model, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be deleted in the second model; or, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be added to the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module to be deleted in the model structure of the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

60. The communication device according to claim 57 or 58, characterized in that: The first model is determined based on the second model, If the first device stores only one second model, the model module associated with the third identifier is the model module in the second model that needs to be deleted; or, If the first device stores only one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, If the first device stores only one model structure of the second model, the model module associated with the third identifier is the The model module to be deleted in the model structure of the second model; or, If the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

61. The communication device according to claim 54, characterized in that The model information of the first model includes a target identifier, part or all of the model parameters of the first model and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier and the third identifier.

62. The communication device according to any one of claims 53 to 61, characterized in that: The model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating a compilation format of model data of the first model; Information indicating whether the model data of the first model has been converted into a compilation format; Information used to indicate a compilation format used by the model data of the first model before the compilation format is converted; Information used to indicate a compilation format used by the model data of the first model after the compilation format conversion; Information used to indicate a compilation format that should be used when the model data of the first model is run on the first device.

63. The communication device according to claim 62, characterized in that The second information is used to indicate that the model data of the first model is model data converted through a compilation format, and the model data of the first model is model data of the first model after conversion through a compilation format.

64. The communication device according to claim 53, characterized in that The second information is used to indicate that the first model is a model that has been converted into a compiled format, and the second information also includes the first model after the conversion into a compiled format.

65. The communication device according to any one of claims 53 to 64, characterized in that: The compilation format associated with the second information is a first compilation format, the first compilation format belongs to multiple compilation formats supported by the first device, and the model data amount corresponding to the first compilation format is the smallest among the multiple compilation formats.

66. The communication device according to claim 53, characterized in that The model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

67. The communication device according to any one of claims 53 to 65, characterized in that: The model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

68. The communication device according to claim 66 or 67, characterized in that: The model data of the first model is converted into a compiled format.

69. The communication device according to claim 67 or 68, characterized in that The third information includes one or more of the following: used to indicate that the model data of the first model is compressed model data; information indicating performance of the first model; Information indicating a compression format of the first model.

70. The communication device according to any one of claims 66 to 69, characterized in that: The compression format of the first model includes bit stream compression and / or model compression.

71. The communication device according to claim 70, characterized in that The compression format of the first model is to perform bit stream compression on the first model after model compression.

72. The communication device according to any one of claims 62-63 and 67-69, characterized in that: The model data of the first model includes one or more of the following among the model information of the first model: Model structure data of the first model; Model module data of the first model; Some or all of the model parameters of the first model.

73. The communication device according to any one of claims 53 to 72, characterized in that: The communication device further comprises: A sending unit, configured to send sixth information to the second device, where the sixth information is used to indicate one or more of the following: capability information associated with the first model; The first device requires time information of the first model; Model-related information stored in the first device; performance requirements of the first device for the first model; whether the first device supports model compression; information associated with the arrival of data in the first device; location information of the first device; Information about a serving cell of the first device; Information about neighboring cells of the first device; The first device sends buffer status report (BSR) information.

74. The communication device according to claim 73, characterized in that The sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: The first device has or does not have a capability for accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; The model compression format supported by the first device.

75. The communication device according to claim 73 or 74, characterized in that The sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

76. The communication device according to any one of claims 73 to 75, characterized in that: The receiving unit is used for: Receive seventh information sent by the second device, where the seventh information is associated with the transmission of the sixth information.

77. The communication device according to claim 76, characterized in that The seventh information includes one or more of the following: Information used to instruct sending the sixth information; Information used to indicate the sixth information sending method; Information used to indicate the information content carried in the sixth information.

78. The communication device according to any one of claims 53 to 77, characterized in that: The first device is a terminal device, and the second device is a network device.

79. A communication device, characterized in that: The communication device is a second device, comprising: A sending unit, configured to send model information of a first model to a first device, wherein the model information of the first model includes one or more of the following: Some or all of the model parameters of the first model; first information for determining a model structure of the first model; second information associated with the compiled format of the first model; Third information associated with the compression of the first model.

80. The communication device according to claim 79, characterized in that The model information of the first model includes the first information, and the first information includes one or more of the following: A first identifier, where the first identifier is a model identifier of the second model; A second identifier, where the second identifier is a model structure identifier of a second model; A third identifier, wherein the third identifier is a model module identifier of the first model; Fourth information, associated with the arrangement position of the model module in the first model; fifth information, including module data of some model modules in the first model; The second model is the first model, or the first model is determined based on the second model.

81. The communication device according to claim 79, characterized in that The model information of the first model includes the first identifier and part or all of the model parameters of the first model.

82. The communication device according to claim 79, characterized in that The model information of the first model includes the second identifier and part or all of the model parameters of the first model.

83. The communication device according to claim 79, characterized in that The model information of the first model includes the third identifier and part or all of the model parameters of the first model.

84. The communication device according to claim 83, characterized in that The model information of the first model also includes the fourth information.

85. The communication device according to claim 83 or 84, characterized in that The first model is determined based on the second model, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be deleted in the second model; or, The model information of the first model also includes the first identifier, and the model module associated with the third identifier is the model module to be added to the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module to be deleted in the model structure of the second model; or, The model information of the first model also includes the second identifier, and the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

86. The communication device according to claim 83 or 84, characterized in that The first model is determined based on the second model, If the first device stores only one second model, the model module associated with the third identifier is the model module in the second model that needs to be deleted; or, If the first device stores only one second model, the model module associated with the third identifier is the model module that needs to be added to the second model; or, If the first device stores only one model structure of the second model, the model module associated with the third identifier is the model module that needs to be deleted in the model structure of the second model; or, If the first device only stores one model structure of the second model, the model module associated with the third identifier is the model module that needs to be added to the model structure of the second model.

87. The communication device according to claim 80, characterized in that The model information of the first model includes a target identifier, part or all of the model parameters of the first model and the fifth information, wherein the target identifier includes one or more of the first identifier, the second identifier and the third identifier.

88. The communication device according to any one of claims 79 to 87, characterized in that: The model information of the first model includes the second information, and the second information includes one or more of the following: information for indicating a compilation format of model data of the first model; Information indicating whether the model data of the first model has been converted into a compilation format; Information used to indicate a compilation format used by the model data of the first model before the compilation format is converted; Information used to indicate a compilation format used by the model data of the first model after the compilation format conversion; Information used to indicate a compilation format that should be used when the model data of the first model is run on the first device.

89. The communication device according to claim 88, characterized in that The second information is used to indicate that the model data of the first model is model data converted through a compilation format, and the model data of the first model is model data of the first model after conversion through a compilation format.

90. The communication device according to claim 79, characterized in that The second information is used to indicate that the first model is a model that has been converted into a compiled format, and the second information also includes the first model after the conversion into a compiled format.

91. The communication device according to any one of claims 79 to 90, characterized in that: The compilation format associated with the second information is a first compilation format, the first compilation format belongs to multiple compilation formats supported by the first device, and the model data amount corresponding to the first compilation format is the smallest among the multiple compilation formats.

92. The communication device according to claim 79, characterized in that The model information of the first model includes the third information, and the model information of the first model includes the compressed first model.

93. The communication device according to any one of claims 79 to 91, characterized in that: The model information of the first model includes the third information, and the model data of the first model included in the model information of the first model is compressed model data.

94. The communication device according to claim 92 or 93, characterized in that: The model data of the first model is converted into a compiled format.

95. The communication device according to claim 93 or 94, characterized in that: The third information includes one or more of the following: used to indicate that the model data of the first model is compressed model data; information indicating performance of the first model; Information indicating a compression format of the first model.

96. The communication device according to any one of claims 92-95, characterized in that: The compression format of the first model includes bit stream compression and / or model compression.

97. The communication device according to claim 96, characterized in that The compression format of the first model is to perform bit stream compression on the first model after model compression.

98. The communication device according to any one of claims 88-89 and 93-95, characterized in that: The model data of the first model includes one or more of the following among the model information of the first model: Model structure data of the first model; Model module data of the first model; Some or all of the model parameters of the first model.

99. The communication device according to any one of claims 79 to 98, characterized in that: The communication device further comprises: A receiving unit, configured to receive sixth information sent by the first device, where the sixth information is used to indicate one or more of the following: capability information associated with the first model; The first device requires time information of the first model; Model-related information stored in the first device; performance requirements of the first device for the first model; whether the first device supports model compression; information associated with the arrival of data in the first device; location information of the first device; Information about a serving cell of the first device; Information about neighboring cells of the first device; The first device sends buffer status report (BSR) information.

100. The communication device according to claim 99, characterized in that The sixth information is used to indicate the capability information, and the capability information is used to indicate one or more of the following: The first device has or does not have a capability for accelerating model reasoning; Storage space in the first device that can be used for model reasoning; The computing power available for model reasoning in the first device; The amount of power available for model inference in the first device; The model compression format supported by the first device.

101. The communication device according to claim 99 or 100, characterized in that: The sixth information is used to indicate the model-related information stored in the first device, and the stored model-related information includes one or more of the following: the number of stored models; the stored model identification; the stored model structure identification; the number of stored model structures; and the stored model module information.

102. The communication device according to any one of claims 99 to 101, characterized in that: The sending unit is used for: Seventh information is sent to the first device, where the seventh information is associated with the transmission of the sixth information.

103. The communication device according to claim 102, characterized in that The seventh information includes one or more of the following: Information used to instruct sending the sixth information; Information used to indicate the sixth information sending method; Information used to indicate the information content carried in the sixth information.

104. The communication device according to any one of claims 79 to 103, characterized in that: The first device is a terminal device, and the second device is a network device.

105. 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 1-26 or 27-52.

106. A device, characterized in that: It comprises a processor, which is used to call a program from a memory so that the device executes the method as described in any one of claims 1-26 or 27-52.

107. A chip, characterized in that: It comprises a processor, which is used to call a program from a memory, so that a device equipped with the chip executes the method as described in any one of claims 1-26 or 27-52.

108. A computer-readable storage medium, characterized in that A program is stored thereon, the program causing a computer to execute the method as claimed in any one of claims 1-26 or 27-52.

109. A computer program product, characterized in that A program is included, which causes a computer to execute the method as described in any one of claims 1-26 or 27-52.

110. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1-26 or 27-52.

Images