Information configuration method and device

Through information configuration methods and equipment, model training information is flexibly configured, and the problem of model training information configuration in the communication system is solved, automated management and independent optimization are realized, and the needs of complex communication systems are met.

WO2024234386A1PCT designated stage expired Publication Date: 2024-11-21GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD

Patent Information

Application Number
PCT/CN2023/095150
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-18
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

When the existing communication systems introduce artificial intelligence models, it is difficult to flexibly configure the information required for model training, resulting in the model being unable to meet the needs of complex and diverse communication systems, and relying on human engineers for maintenance and optimization, and lacking the ability to independently optimize.

Method used

Provide an information configuration method and equipment to flexibly configure the information required for model training by obtaining and sending model training configuration information, including input parameter type range, data source constraints, output parameter type range, operation capability requirements, etc., to automatically implement model training process to achieve independent optimization.

Benefits of technology

It realizes automatic management of model training, avoids artificial setting errors, improves the flexibility and efficiency of model training, supports autonomous optimization of communication systems, and meets the needs of diverse communication systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2023095150_21112024_PF_FP_ABST
    Figure CN2023095150_21112024_PF_FP_ABST
Patent Text Reader

Abstract

The present application relates to an information configuration method. The method comprises: a first device acquiring model training configuration information, wherein the model training configuration information is used for indicating information related to the training of a first model. In the embodiments of the present application, model training configuration information is designed according to actually required model characteristics, a model is trained according to the model training configuration information, a model training report is generated, and a model identifier is requested for a generated target model, such that the automated management of model training is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Information configuration method and device Technical Field

[0001] The present application relates to the field of communications, and more specifically, to an information configuration method and device. Background Art

[0002] The model training process typically uses a large amount of collected data as algorithm input and continuously optimizes the algorithm to achieve the desired output. When the output meets the expectations, the algorithm is called a model. Communication system scenarios are complex and diverse. When introducing AI models, it is necessary to consider how to ensure that the model meets the needs of the communication system.

[0003] Summary of the Invention

[0004] The embodiments of the present application provide an information configuration method and device that can flexibly configure the information required for model training.

[0005] This embodiment of the present application provides an information configuration method, including:

[0006] The first device obtains model training configuration information, where the model training configuration information is used to indicate information related to training of the first model.

[0007] This embodiment of the present application provides an information configuration method, including:

[0008] The second device sends model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0009] An embodiment of the present application provides a first device, including:

[0010] An acquisition unit is used to acquire model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0011] An embodiment of the present application provides a second device, including:

[0012] The first sending unit is used to send model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0013] An embodiment of the present application provides a first device, comprising a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory, so that the first device executes the above-mentioned information configuration method.

[0014] An embodiment of the present application provides a second device, comprising a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to call and run the computer program stored in the memory, so that the second device executes the above-mentioned information configuration method.

[0015] The embodiment of the present application provides a chip for implementing the above-mentioned information configuration method. Specifically, the chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the above-mentioned information configuration method.

[0016] An embodiment of the present application provides a computer-readable storage medium for storing a computer program. When the computer program is executed by a device, the device executes the above-mentioned information configuration method.

[0017] An embodiment of the present application provides a computer program product, including computer program instructions, which enable a computer to execute the above-mentioned information configuration method.

[0018] An embodiment of the present application provides a computer program, which, when executed on a computer, enables the computer to execute the above-mentioned information configuration method.

[0019] In the embodiment of the present application, the information required for model training can be flexibly configured through model training configuration information. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] FIG1 is a schematic diagram of an application scenario according to an embodiment of the present application.

[0021] FIG2 is a schematic flowchart of an information configuration method according to an embodiment of the present application.

[0022] FIG3 is a schematic flowchart of an information configuration method according to another embodiment of the present application.

[0023] FIG4 is a schematic flowchart of an information configuration method according to another embodiment of the present application.

[0024] FIG5 is a schematic flowchart of an information configuration method according to another embodiment of the present application.

[0025] FIG6 is a schematic block diagram of a first device according to an embodiment of the present application.

[0026] FIG7 is a schematic block diagram of a first device according to another embodiment of the present application.

[0027] FIG8 is a schematic block diagram of a second device according to an embodiment of the present application.

[0028] FIG9 is a schematic block diagram of a second device according to another embodiment of the present application.

[0029] FIG10 is a schematic block diagram of a communication device according to an embodiment of the present application.

[0030] FIG11 is a schematic block diagram of a chip according to an embodiment of the present application.

[0031] FIG12 is a schematic block diagram of a communication system according to an embodiment of the present application. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0033] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Global System of Mobile communication (GSM) system, Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, Advanced Long Term Evolution (LTE-A) system, New Radio (NR) system, NR system evolution system, LTE on unlicensed spectrum (LTE-U) system, NR on unlicensed spectrum (NR-U) system, Non-Terrestrial Networks (NTN) system, Universal Mobile Telecommunication System (UMTS), Wireless Local Area Networks (WLAN), Wireless Fidelity (Wireless Fidelity) system. Fidelity, WiFi), fifth-generation communication (5th-Generation, 5G) system or other communication systems, etc.

[0034] Generally speaking, traditional communication systems support a limited number of connections and are easy to implement. However, with the development of communication technology, mobile communication systems will not only support traditional communications, but will also support, for example, device-to-device (D2D) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), vehicle-to-vehicle (V2V) communication, or vehicle-to-everything (V2X) communication, etc. The embodiments of the present application can also be applied to these communication systems.

[0035] In one embodiment, the communication system in the embodiment of the present application can be applied to a carrier aggregation (CA) scenario, a dual connectivity (DC) scenario, and a standalone (SA) networking scenario.

[0036] In one embodiment, the communication system in the embodiment of the present application can be applied to an unlicensed spectrum, wherein the unlicensed spectrum can also be considered as a shared spectrum; or, the communication system in the embodiment of the present application can also be applied to an authorized spectrum, wherein the authorized spectrum can also be considered as an unshared spectrum.

[0037] The embodiments of the present application describe various embodiments in conjunction with network devices and terminal devices, wherein the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, etc.

[0038] The terminal device can be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a wearable device, a terminal device in a next-generation communication system such as an NR network, or a terminal device in a future evolved Public Land Mobile Network (PLMN) network, etc.

[0039] In an embodiment of the present application, the terminal device can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; it can also be deployed on the water surface (such as ships, etc.); it can also be deployed in the air (such as airplanes, balloons and satellites, etc.).

[0040] In an embodiment of the present application, the terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, or a wireless terminal device in a smart home, etc.

[0041] As an example and not a limitation, in the embodiment of the present application, the terminal device may also be a wearable device. Wearable devices may also be called wearable smart devices, which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are fully functional, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0042] In an embodiment of the present application, the network device may be a device for communicating with a mobile device. The network device may be an access point (AP) in WLAN, a base station (BTS) in GSM or CDMA, a base station (NodeB, NB) in WCDMA, an evolved base station (eNB or eNodeB) in LTE, or a relay station or access point, or a vehicle-mounted device, a wearable device, and a network device (gNB) in an NR network, or a network device in a future evolved PLMN network or a network device in an NTN network, etc.

[0043] As an example and not a limitation, in an embodiment of the present application, the network device may have a mobile feature, for example, the network device may be a mobile device. Alternatively, the network device may be a satellite or a balloon station. For example, the satellite may be a low earth orbit (LEO) satellite, a medium earth orbit (MEO) satellite, a geostationary earth orbit (GEO) satellite, a high elliptical orbit (HEO) satellite, etc. Optionally, the network device may also be a base station set up in a location such as land or water.

[0044] In an embodiment of the present application, the network device can provide services for a cell, and the terminal device communicates with the network device through the transmission resources used by the cell (for example, frequency domain resources, or spectrum resources). The cell can be a cell corresponding to the network device (for example, a base station). The cell can belong to a macro base station or a base station corresponding to a small cell. The small cells here may include: metro cells, micro cells, pico cells, femto cells, etc. These small cells have the characteristics of small coverage and low transmission power, and are suitable for providing high-speed data transmission services.

[0045] FIG1 exemplarily illustrates a communication system 100. The communication system includes a network device 110 and two terminal devices 120. In one embodiment, the communication system 100 may include multiple network devices 110, and each network device 110 may include a different number of terminal devices 120 within its coverage area, which is not limited in this embodiment of the present application.

[0046] In one embodiment, the communication system 100 may further include other network entities such as a Mobility Management Entity (MME) and an Access and Mobility Management Function (AMF), which is not limited in this embodiment of the present application.

[0047] Among them, the network equipment may include access network equipment and core network equipment. That is, the wireless communication system also includes multiple core networks for communicating with the access network equipment. The access network equipment can be an evolutionary base station (evolutional node B, abbreviated as eNB or e-NodeB) macro base station, micro base station (also called "small base station"), pico base station, access point (AP), transmission point (TP) or new generation base station (new generation Node B, gNodeB), etc. in a long-term evolution (LTE) system, a next-generation (mobile communication system) (next radio, NR) system or an authorized auxiliary access long-term evolution (LAA-LTE) system.

[0048] It should be understood that in the embodiments of the present application, a device having a communication function in a network / system may be referred to as a communication device. Taking the communication system shown in Figure 1 as an example, the communication device may include a network device and a terminal device having a communication function. The network device and the terminal device may be specific devices in the embodiments of the present application and will not be described in detail here. The communication device may also include other devices in the communication system, such as a network controller, a mobility management entity, and other network entities, which are not limited in the embodiments of the present application.

[0049] It should be understood that the terms "system" and "network" are often used interchangeably herein. The term "and / or" is simply a description of an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " generally indicates that the related objects are in an "or" relationship.

[0050] It should be understood that the "indication" mentioned in the embodiments of this application can be a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also mean that there is an association between A and B.

[0051] In the description of 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 being indicated, configuration and being configured, etc.

[0052] To facilitate understanding of the technical solutions of the embodiments of the present application, the relevant technologies of the embodiments of the present application are described below. The following relevant technologies can be arbitrarily combined with the technical solutions of the embodiments of the present application as optional solutions, and they all fall within the protection scope of the embodiments of the present application.

[0053] The model training process is generally an industry-wide practice. Even for the same optimization goal, different vendors will likely implement it differently, and the reference benchmarks for each vendor's trained models are likely to differ. The situation is slightly different when artificial intelligence (AI) is applied to communication systems. Communication systems have stringent requirements for many system metrics (such as latency, throughput, and bit error rate), which requires trained models to meet many common communication system metrics. Furthermore, existing communication systems are still largely manually maintained, requiring human engineers to participate from development to ongoing maintenance. The future development trend of communication systems is towards self-optimizing systems, which requires these systems to possess further autonomous optimization capabilities. AI, as an advanced optimization tool, can be a valuable tool for achieving self-optimizing communication systems. Consideration can be given to enabling communication systems to autonomously generate models that meet these requirements.

[0054] FIG2 is a schematic flow chart of an information configuration method according to an embodiment of the present application. The method 200 can optionally be applied to the system shown in FIG1 , but is not limited thereto. The method includes at least part of the following contents.

[0055] S210. The first device obtains model training configuration information, where the model training configuration information is used to indicate information related to training of the first model.

[0056] In an embodiment of the present application, it is necessary to obtain model training configuration information before formally training a model so as to train a model that meets the requirements according to the model training configuration information requirements. Obtaining model training configuration information can also be determining model training configuration information. Models may include but are not limited to AI models and / or ML models. For example, the model may include a model with functions related to the communication field, such as a CSI prediction model, a beam prediction model, a UE positioning model, a channel estimation model, and the like. The first device may obtain various information related to the training of the first model. This information may be collectively referred to as model training configuration information of the first model. The first model may include an expected model, where the expected model may also be referred to as a planned model. The training configuration information of the expected model may include one or more of information related to parameters associated with the expected model, information related to performance requirements associated with the expected model, and the like.

[0057] In one embodiment, the model training configuration information is generated by the first device. In this case, the first device can obtain the training configuration information of the desired model from the training configuration information of one or more models generated by the first device. Optionally, when the model training configuration information is generated by the first device, the method further includes the first device determining the model training configuration information after negotiating with the second device through a first negotiation process.

[0058] In one embodiment, the model training configuration information is received by the first device from the second device, and the model training configuration information is generated by the second device. In the case where the model training configuration information is sent by the second device to the first device, the model training configuration information can be generated by the second device and sent to the first device. In other words, the first device can obtain the training configuration information of the desired model from the training configuration information of one or more models generated by the second device.

[0059] In one embodiment, the model training configuration information is received by the first device from the second device, the model training configuration information is received by the second device from the third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device receives the model training configuration information directly from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices. In the case where the model training configuration information is sent from the second device to the first device, the model training configuration information can be generated by the third device and sent to the first device through the second device, and the third device and the second device are different devices. In this case, the second device and / or the third device can be understood as relay devices, which are used to forward the model training configuration information generated by other devices to the first device.

[0060] The above-mentioned model training configuration information can be generated from a variety of sources, which can meet the model training needs of any node in the communication system, allowing artificial intelligence tools to flexibly serve various functional units of the communication network.

[0061] In one embodiment, the first device is a terminal device or a network device. For example, the first device is a terminal device, and the second device and / or the third device are network devices. For another example, the first device is a network device, and the second device and / or the third device are terminal devices. For another example, the first device, the second device, and the third device are different terminal devices. For another example, the first device, the second device, and the third device are different network devices.

[0062] In one embodiment, the model training configuration information includes one or more of the following information: the input parameter type range of the first model, the data source constraint information of the input parameters of the first model, the output parameter type range of the first model, the operating capability requirements of the first model, the output accuracy requirements of the first model, the complexity requirements of the first model, the size requirements of the first model, the algorithm structure requirements of the first model, the contribution rate of the input parameters of the first model or the contribution rate level threshold requirements, and the prediction accuracy rate of the output parameters of the first model or the prediction accuracy level threshold requirements. In the example of the above configuration information, the input parameter type range may include an input parameter type list and / or an input parameter type set, etc. The output parameter type range may include an output parameter type list and / or an output parameter type set, etc.

[0063] Among them, the first model can also be called the expected model, and the model training configuration information corresponding to the expected model may include one or more of the following information: expected model input parameter type range, expected model input parameter data source constraint information, expected model output parameter type range, expected model operation capability requirements, expected model output accuracy requirements, expected model complexity requirements, expected model size requirements, expected model algorithm structure requirements, expected model input parameter contribution rate or contribution rate level threshold requirements (also referred to as effective model input parameter contribution rate or contribution rate level threshold requirements), expected model output parameter prediction accuracy or prediction accuracy level threshold requirements (also referred to as effective model output parameter contribution rate or contribution rate level threshold requirements).

[0064] In one embodiment, the input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

[0065] In one embodiment, the input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

[0066] In an embodiment of the present application, the identification information and / or label information corresponding to the determined model input parameter type may be the identification information and / or label information corresponding to the required model input parameter type. The identification information and / or label information corresponding to the recommended model input parameter type may be the identification information and / or label information corresponding to the optional model input parameter type. In an embodiment of the present application, the identification information and label information corresponding to the parameter type may be the same or different. For example, the identification information corresponding to the position parameter may include Po, and the label information may include <position>.

[0067] Example 1-1, Expected Model Input Parameter Type Range: There may be differences between the model input parameters recommended during model training and the model input parameters actually confirmed by the model after model training. Typically, the parameter set included in the latter is a subset of the parameter set included in the former (including scenarios where the sets are the same). This is because, at the beginning of model training, the model training strategy developer cannot determine which parameters are suitable as input parameters for the target model (hereinafter referred to as the second model), so an expected model input parameter type range is provided. In the first implementation, the expected model input parameter type range includes identification and / or label information corresponding to at least one recommended model input parameter type. The first approach may result in slower convergence during model training, but its advantages include not having to prejudge which parameters are necessarily useful for model input; the final model input parameters of the target model are fully determined by the model training results. In the second implementation, the expected model input parameter type range includes two types of parameters. The first type of parameters are mandatory parameters, and contain identifiers and / or label information corresponding to at least one determined model input parameter type. That is, the first type of parameters must be included as model input parameters during the model training process (or the model input parameters actually confirmed by the model after model training must include the first type of parameters). The second type of parameters are optional parameters, and contain identifiers and / or label information corresponding to at least one recommended model input parameter type. That is, the second type of parameters can be included as model input parameters during the model training process (or whether the model input parameters actually confirmed by the model after model training include the second type of parameters depends on the model training results). Under similar conditions, the second implementation method may result in faster convergence of the model training process than the first implementation method.

[0068] In one embodiment, the data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

[0069] Example 1-2, data source constraint information for expected model input parameters: Once the data source constraint information for expected model input parameters is configured, all model input parameter data used in the model training process must meet the data source constraint information for the expected model input parameters. The data source constraint information for the expected model input parameters is related to the generalization of the model. The so-called model generalization can be the scope of application of the model. Generally, models with better generalization have a wider scope of application, while models with poorer generalization have a smaller scope of application. The generalization of the model is usually directly related to the source of the model input parameters during model training. Generally, if a model is expected to be applicable to multiple scenarios after training, the source of the model input parameters provided during model training should include these multiple scenarios that are expected to be applicable. In order for the target model obtained through training to meet a certain level of model generalization, it is limited by the expected model input parameter source constraint information.

[0070] Optionally, the data source constraint information of the expected model input parameter includes one or more of the following information: source country information of the data association of the expected model input parameter, source geographic area information of the data association of the expected model input parameter, source operator information of the data association of the expected model input parameter, source logical area information of the data association of the expected model input parameter, source scenario information of the data association of the expected model input parameter, and source configuration information of the data association of the expected model input parameter.

[0071] Optionally, the operator information includes one or more of Public Land Mobile Network (PLMN) identification information, Stand-alone Non-public Networks (SNPN) identification information, and Closed Access Group (CAG) identification information.

[0072] Optionally, the logical area information includes one or more of the following information: one or more frequency identification information, one or more physical cell identification (PCI) identification information, one or more cell identification information, at least one tracking area identification tracking area code (TAC) information and at least one access network notification area identification (Radio Access Network Area Code, RANAC) information.

[0073] The above cell identification information is composed of PCI+frequency information or represented by Cell Global Identity (CGI) information or represented by Cell Index (Cell Index).

[0074] Optionally, the scene information is related to the actual model type. For example, for a model used for positioning, the scene information may include a Line of Sight (LOS) scene, a None-Line of Sight (NLOS) scene, and a mixed scene (i.e., both LOS and NLOS scenes). For a model used for beam prediction, the scene information may include a FR1 frequency range (FR) scene, a FR2-1 scene, a FR2-2 scene, and a mixed scene (without distinguishing between FR scenes).

[0075] Optionally, the configuration information includes one or more of physical layer configuration, L2 configuration, L3 configuration, non-access stratum (NAS) configuration, and application layer configuration.

[0076] In one embodiment, the output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

[0077] In one embodiment, the output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

[0078] Example 1-3, Expected Model Output Parameter Type Range: There may be differences between the model output parameters recommended during model training and the model output parameters actually confirmed after model training. Typically, the parameter set included in the latter is a subset of the parameter set included in the former (including scenarios where the sets are identical). This is because, at the beginning of model training, the model training strategy developer cannot determine which parameters are suitable as output parameters for the target model, and therefore provides an expected model output parameter type range. In the first implementation, the expected model output parameter type range includes identifiers and / or label information corresponding to at least one suggested model output parameter type. This may result in slower model training convergence, but its advantages include not needing to prejudge which parameters are most suitable as model outputs; the final model output parameters of the target model are fully determined by the model training results. In the second implementation, the expected model output parameter type range includes two types of parameters. The third type of parameter is mandatory and includes identifiers and / or label information corresponding to at least one determined model output parameter type. This means that the third type of parameter must be included as part of the model output parameters during model training (or, the model output parameters actually confirmed after model training must include the third type of parameter). The fourth type of parameters are optional parameters, which contain identification and / or label information corresponding to at least one suggested model output parameter type, that is, the fourth type of parameters can be selected as model output parameters during model training (or whether the model output parameters actually confirmed by the model after model training include the fourth type of parameters depends on the confirmation of the model training results).

[0079] Examples 1-4, expected model operation capability requirements: Using a model generally requires meeting certain capabilities, such as computing power and memory capacity. Models trained using different implementations will also require different capabilities. To ensure that the trained target model meets usage requirements, constraints are imposed by the expected model operation capability requirements. Of course, the expected model operation capability requirements can be formulated separately for different capability items, and this is not limited in the present embodiment.

[0080] Examples 1-5, Expected Model Output Accuracy Requirements: Models trained using different implementations typically have different corresponding model output accuracy requirements. To ensure that the trained target model meets the model output accuracy requirements, the expected model output accuracy requirements are used as constraints. Of course, the expected model output accuracy requirements can be set separately for different model output items, and this is not a limitation in the present embodiments.

[0081] Example 1-6, Expected Model Complexity Requirements: Models trained using different implementations will have different corresponding model complexities. For example, model complexity can be expressed using the total number of neurons and / or the total number of neuron layers. To ensure that the trained target model meets the model complexity requirements, the expected model complexity requirements are used as constraints.

[0082] Example 1-7, Expected Model Size Requirements: The sizes of trained models vary depending on the implementation. For example, for the same CSI compression feedback model, some implementations yield models of only tens of KB, while others can reach several MB. To ensure that the trained target model meets the required model size, use the expected model size requirement as a constraint.

[0083] Example 1-8, Expected Model Algorithm Structure Requirements: Models implementing the same functionality may use different model algorithm structures. For example, some algorithms use a fully connected network algorithm structure, some use a convolutional network algorithm structure, some use a neural network algorithm structure, and some use a hybrid network algorithm structure. To ensure that the target model obtained through training meets the model algorithm structure requirements, the expected model algorithm structure requirements are used as constraints.

[0084] In one embodiment, there is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold;

[0085] Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model. All or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

[0086] The contribution rate level is associated with a contribution rate range, and each contribution rate level is associated with a contribution rate range.

[0087] Example 1-9, effective model input parameter contribution rate or contribution rate level threshold requirement: The expected model input parameter type range may include a variety of input parameter types, and not every expected model input parameter contributes to the accuracy of the model's output parameters. Even if they do contribute, different model input parameters may contribute differently to the accuracy of the model's output parameters. To ensure that the trained target model is concise and effective, that is, to minimize the model input parameters while ensuring that the model output meets the accuracy requirements, effective model input parameter contribution rate or contribution rate level threshold requirements can be used to constrain the model. For example, the expected model input parameters include four types: a, b, c, and d, and the model output is parameter type e. If the model training process finds that parameters a and b are essential for obtaining the model output, then the contribution rates of parameters a and b to the model output can be considered to be 100%. If it is found that under the premise that the model input includes parameters a and b, the addition of parameter c can only obtain an additional 5% performance gain, while the addition of parameter d actually reduces the overall performance by 10%, and assuming that the effective model input parameter contribution rate threshold is set to 6% or the effective model input parameter contribution rate level threshold is set to level 1 (where level 1 corresponds to an effective model input parameter contribution rate greater than or equal to 10%), then the input parameters of the target model obtained after training only include parameters a and b. Although the addition of parameter c can improve the accuracy of the model output parameters by 5%, it still does not meet the 6% contribution rate threshold requirement or the level 1 threshold requirement. The contribution rate of parameter d is negative, which obviously does not meet the model input contribution rate threshold or contribution rate level threshold requirements.

[0088] It should be noted that the input parameter types included in the expected model input parameter type range and the effective model input parameter contribution rate or contribution rate level threshold requirements can be a one-to-one mapping relationship or a many-to-one mapping relationship. Among them, the one-to-one mapping relationship means that each input parameter type included in the expected model input parameter type range is associated with a valid model input parameter contribution rate or contribution rate level threshold. The many-to-one mapping relationship means that all or a group of input parameter types included in the expected model input parameter type range are associated with a valid model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different valid model input parameter contribution rates or contribution rate level thresholds.

[0089] In one embodiment, there is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold;

[0090] Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirements of the first model. All or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and the effective model output parameter prediction accuracy or prediction accuracy level threshold associated with different groups of output parameter types is different.

[0091] The prediction accuracy level is associated with a prediction accuracy range, and each prediction accuracy level is associated with a prediction accuracy range.

[0092] Example 1-10, effective model output parameter prediction accuracy or accuracy level threshold requirement: The expected model output parameter type range may include a variety of output parameter types, and not every expected model output parameter can be accurately predicted or inferred. Even if these expected model output parameters are predictable, the model's prediction accuracy or accuracy level for different output parameters may vary. To ensure that the model output parameters associated with the trained target model are concise and the prediction results have high accuracy or accuracy level, effective model output parameter prediction accuracy or accuracy level threshold requirements can be used to constrain the prediction accuracy or accuracy level. For example, the expected model output parameters include four types: m, n, p, and q, and the model input parameters are w1 and w2. During model training, it was found that when the model input parameters w1 and w2 were all the same, the prediction accuracy of the four output parameters m, n, p, and q was 90%, 85%, 30%, and 10%, respectively. At the same time, assuming that the effective model output parameter prediction accuracy threshold is set to 80% or the effective model output parameter prediction accuracy level threshold is set to level 2 (where level 2 corresponds to an effective model output parameter prediction accuracy greater than or equal to 80%), then the model's prediction of output parameters m and parameter n meets the prediction accuracy or accuracy level threshold requirements, but the prediction of output parameters p and parameter q does not meet the prediction accuracy or accuracy level threshold requirements.

[0093] It should be noted that there is a one-to-one mapping relationship or a many-to-one mapping relationship between the output parameter types included in the expected model output parameter type range and the effective model output parameter prediction accuracy or accuracy level threshold requirements. Among them, the one-to-one mapping relationship means that each output parameter type included in the expected model output parameter type range is associated with a valid model output parameter prediction accuracy or accuracy level threshold. The many-to-one mapping relationship means that all or a group of output parameter types included in the expected model output parameter type range are associated with a valid model output parameter prediction accuracy or accuracy level threshold, and different groups of output parameter types are associated with different valid model output parameter prediction accuracy or accuracy level thresholds.

[0094] FIG3 is a schematic flow chart of an information configuration method according to another embodiment of the present application. The method 300 may include one or more features of the information configuration method of the above embodiment. In one implementation, as shown in FIG3 , the method further includes:

[0095] S310: The first device determines model training information based on the model training configuration information;

[0096] S320: The first device performs model training based on the model training information to obtain one or more second models.

[0097] In another embodiment, step 310 and step 320 may be combined into one step, which is not limited in this application.

[0098] In one embodiment, the second model is an instance of the first model, that is, the first model is the expected model before model training, and the second model is the target model obtained after actual training according to the model training configuration information associated with the expected model.

[0099] In an embodiment of the present application, the first device can actually train and obtain a second model based on the model training configuration information obtained in the above steps. The second model can also be referred to as a target model. The target model is the model actually obtained after model training. The target model can be understood as an instance of the desired model, and the parameters, performance requirements, etc. of the target model can be a subset of the desired model. For example, the first model (desired model) is a positioning model, and the input parameters involved in the corresponding model training configuration information include P1, P2, and P3, and the output parameters include location information. The expected model size requirement is less than or equal to 500MB, and the expected model accuracy is a positioning error of less than or equal to 10 meters. Based on the P1, P2, and P3 parameters, a positioning model with a size of 300MB and a positioning accuracy of less than or equal to 5 meters is obtained; based on the P1 and P3 parameters, a positioning model with a size of 200MB and a positioning accuracy of less than or equal to 8 meters is obtained. The two positioning models trained in this example can be understood as two instances of the first model, that is, two second models (target models).

[0100] By setting different model training configuration information, you can flexibly train to obtain different target models.

[0101] In one embodiment, as shown in FIG3 , the method further includes: S330 , the first device generates a model training report. The model training report may include relevant information of the trained second model, such as the target model.

[0102] In one embodiment, the model training report includes one or more of the following information: model description information associated with the second model, the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, the contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, the contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, the prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and the prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

[0103] For example, the training report information of the target model may include one or more of the following information: model description information associated with the target model, the parameter type range in the expected model input parameter type range that is not selected as the model input parameter associated with the target model, the contribution rate or contribution rate level information corresponding to each parameter type contained in the parameter type range in the expected model input parameter type range that is not selected as the model input parameter associated with the target model, the contribution rate or contribution rate level information corresponding to each parameter type contained in the expected model input parameter type range, the parameter type range in the expected model output parameter type range that is not selected as the model output parameter associated with the target model, the prediction accuracy or prediction accuracy level information corresponding to each parameter type contained in the parameter type range in the expected model output parameter type range that is not selected as the model output parameter associated with the target model, and the prediction accuracy or prediction accuracy level information corresponding to each parameter type contained in the expected model output parameter type range.

[0104] In one embodiment, the model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective usage scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

[0105] In embodiments of the present application, the model description information associated with the model may not be completely the same type of information as the model training configuration information. For example, the model description information associated with the model may include some or all of the information in the model training configuration information, and may also include additional information not included in the model description information associated with the model, such as location, generalization characteristics, compilation format, storage format, computational complexity, etc.

[0106] Example 2-1, model description information associated with the target model: the model description information includes one or more of the following information: functional characteristics associated with the target model, input parameter type range associated with the target model, input parameter format requirement information associated with the target model, input parameter preprocessing rules associated with the target model, output parameter type range associated with the target model, output parameter format requirement information associated with the target model, output parameter preprocessing rules associated with the target model, application scenario information associated with the target model, deployment location information of the target model, capability requirement information for using the target model, performance monitoring indicator information of the target model, location information of the target model, effective usage scope information of the target model (for example, applicable configuration information associated with the target model), generalization feature information of the target model, version information of the target model, precision level information of the target model, compilation format information of the target model data, storage format information of the target model data, computational complexity information of the target model, model complexity information of the target model, and model size information of the target model.

[0107] The following are examples of model description information:

[0108] Information 1: Functional characteristics of model association

[0109] The functional characteristics associated with the model are used to indicate the functions associated with the corresponding model. For example, model 1 is used for the Channel State Information (CSI) compression feedback function, model 2 is used for the mobility enhancement function, model 3 is used for the beam prediction function, model 4 is used for the positioning function, etc.

[0110] Information 2: Model-associated input parameter type range

[0111] The input parameter type range associated with a model indicates the type of input parameters required by the corresponding model. For example, Model 3, used for beam prediction, requires input parameters such as the index of the actual measured beam and the signal measurement results for each beam.

[0112] Information 3: Model-related input parameter format requirements

[0113] The input parameter format requirement information associated with the model is used to indicate the format of the input parameters required by the corresponding model. For example: Model 3 is used for the beam prediction function. Assuming that the network device can schedule resources in 8 beam directions, but only sends measurement reference signals in 4 of the beam directions, the terminal device can only obtain measurement results of 4 beams in each measurement process. Scenario 1: The beam prediction model only predicts the signal measurement results of other beams that have not been measured at the current moment. If the beam prediction model used in scenario 1 only needs to input the measurement results of 4 beams, the measurement results of the 4 beams obtained by the terminal device in each measurement can just be used as the input of the beam prediction model of scenario 1 (the output can obtain the measurement results of the 8 beams associated with this measurement, that is: 4 true values, 4 predicted values, or the output is the measurement results of the predicted 4 beams). Scenario 2: The beam prediction model is used to predict the measurement results of 8 beams in the future. In this scenario, the input of the beam prediction model may be the set of measurement results of 4 beams obtained in the past N measurement processes, and the output of the beam prediction model is the set of measurement results of 8 beams in the future. For example, the input of the beam prediction model is the set of three sets of measurement results of 4 beams obtained at time T-3, T-2, and T-1, and the output of the beam prediction model is the set of two sets of measurement results of 8 beams at time T and time T+1.

[0114] In summary, although beam prediction is performed using the same measurement results of four beams (the model input parameter types are the same), different application scenarios require different input data formats. Therefore, the input parameter format requirements associated with the AI / ML model are also one of the key information for describing a model.

[0115] Information 4: Model-associated input parameter preprocessing rules

[0116] The input parameter preprocessing rules associated with the model are used to indicate how to reprocess the input data set of the corresponding model. The operations involved in the reprocessing process may include at least one of data dimensionality reduction, data dimensionality increase, redundancy elimination, and redistribution.

[0117] Data dimensionality reduction: If the data dimension is higher than the dimension required by the model input, the data needs to be reduced in dimension through preprocessing rules before entering the model;

[0118] Data dimensionality increase: If the data dimension is lower than the dimension required by the model input, the data needs to be increased in dimension through preprocessing rules before entering the model;

[0119] Redundancy elimination: If the distribution or information content of multiple sets of data is similar, redundancy needs to be eliminated through preprocessing rules before entering the data into the model. This is because redundant data often does not bring more performance improvements but instead increases ineffective operations.

[0120] Redistribution: In some cases, it is difficult to obtain good model generalization by training data with similar feature information together. Redistributing data with different feature information is conducive to obtaining a model with better generalization.

[0121] For example, in scenario 1 in Information 3, the terminal device can measure 4 of the 8 beams in cell A, 5 of the 8 beams in cell B, and 3 of the 8 beams in cell C. If the same beam prediction model (requiring only the measurement results of 4 beams) is used for beam prediction in cells A, B, and C, the measurement results of cell A can be used directly, but the measurement results obtained in cell B require data dimensionality reduction. Specifically, 4 measurement results are selected from the 5 available beam measurement results as the input parameters required by the model. The measurement results obtained in cell C require data dimensionality increase. Specifically, the measurement results of 3 available beams are converted to 4 to obtain the required model input parameters.

[0122] In another implementation, at least two of information 2, information 3, and information 4 may be defined in combination, and this application does not impose any limitation on this.

[0123] Information 5: Model-associated output parameter type range

[0124] The output parameter type range associated with a model indicates the type of output parameters required by the corresponding model. For example, if model 3 is used for beam prediction, the corresponding output parameter types include the index information of the actual unmeasured beam and the predicted signal measurement results of each actual unmeasured beam.

[0125] Information 6: Model-related output parameter format requirements

[0126] The model-associated output parameter format requirement information is used to indicate the format of the output parameters required by the corresponding model. For example, if Model 3 is used for beam prediction, the output parameter can be the beam prediction result for one future moment, or the output parameter can be a set of beam prediction results for multiple future moments. For example, if the model input is a set of three sets of measurement results for four beams obtained at times T-3, T-2, and T-1, the model output can be the measurement results for eight beams at time T, or two sets of measurement results for eight beams at times T and T+1.

[0127] In summary, although the prediction of future beam measurement results is performed (the model output data type is the same), the required output data formats of Model 3 in different application scenarios are different. Therefore, the output parameter format requirements associated with the model are also one of the key information for describing a model.

[0128] Information 7: Model-associated output parameter preprocessing rules

[0129] The output parameter preprocessing rules associated with a model indicate how to reprocess the output data set of the corresponding model. This reprocessing may involve at least one of data dimensionality reduction, data dimensionality increase, redundant elimination, and redistribution. For specific processing methods, see the instructions in Information 4.

[0130] In another implementation, at least two of information 5, information 6, and information 7 may be defined in combination, which is not limited in this application.

[0131] Information 8: Application scenario information associated with the model

[0132] The application scenario information associated with the model refers to different usage scenarios under the same functional characteristics. For example, for the model used for positioning, the application scenario can be further subdivided into a line of sight (LoS) scenario, a non-line of sight (NLoS) scenario, or a mixed scenario (including both LoS and NLoS). For another example, for the model used for mobility control, the application scenario can be further subdivided into a ground communication scenario, a non-ground communication scenario (including satellite communication, drone communication, hot air balloon communication and other sub-scenarios), and each scenario can be further divided into high-speed mobile scenarios, medium-speed mobile scenarios, low-speed mobile scenarios and other sub-scenarios, and can also be divided into high-frequency mobility scenarios, low-frequency mobility scenarios and other sub-scenarios. It is also possible to consider a combination of two scenarios, such as a ground communication high-frequency mobility scenario.

[0133] Optionally, information 8 may have a value indicating that the application scenario of the model is all scenarios (i.e., the model is applicable to any scenario corresponding to its associated functional characteristics). For example, for a model used for positioning, assuming that information 8 contains two bits, the value '00' can be used to indicate any scenario, '01' to indicate LoS scenario, '10' to indicate NLoS scenario, and '11' to indicate mixed scenario.

[0134] Information 9: Model deployment location information

[0135] The deployment location information of the model includes any of the following meanings: deployed in core network equipment, deployed in application servers, deployed in OAM, deployed in access network equipment, deployed in terminal equipment, or no deployment location restriction.

[0136] Information 10: Capability requirements for using the model

[0137] The capability requirement information for using the model is used to indicate several capability requirements required for using the model. For example, the capability requirement information for using the model includes one or more of the following: the minimum floating-point computing capability required when the model is used, the minimum running memory required when the model is used, the memory size level occupied by the model algorithm data before compilation, the memory size level occupied by the model algorithm data after compilation, the minimum input data rate (or bit rate) required when the model is used, the maximum input data rate (or bit rate) required when the model is used, the minimum output data rate (or bit rate) required when the model is used, and the maximum output data rate (or bit rate) required when the model is used.

[0138] Information 11: Model performance monitoring indicator information

[0139] The model's performance monitoring indicator information is used to monitor several indicators of the model. For example, the model's performance monitoring indicator information may include one or more of the following: average energy consumption per inference process, average floating-point calculation amount per inference process, average latency per inference process, system throughput, system bit error rate, and average data packet transmission latency.

[0140] Information 12: Model location information

[0141] The location information of the model is used to indicate the information related to the provider of the model data, specifically including: the country region information to which the model data belongs and / or the service provider information to which the model data belongs.

[0142] The service provider may be any of the following: OTT (Over The Top, providing application services to users through the Internet) manufacturer information, network operator information (for example, represented by a PLMN identifier), and public website information.

[0143] Information 13: Model's effective scope of use (e.g., applicable scenarios, configuration, site information, etc.)

[0144] The valid usage scope information of a model is used to indicate the conditions under which the model can be effectively used. For example, the valid usage scope information of a model includes one or more of the following: geographical area information allowing valid use, country information allowing valid use, PLMN information allowing valid use, Tracking Area Code (TAC) information allowing valid use, RANAC (RAN Area Code) information allowing valid use, cell information allowing valid use, and configuration information allowing valid use.

[0145] Information 14: generalization characteristics of the model.

[0146] Generalization characteristics reflect the breadth of applicable scenarios for a model. Generally speaking, the higher the generalization characteristics of a model, the more application scenarios it is applicable to. Model generalization characteristics information can include at least one type of model application scenario information (see Information 8 for definition), indicating that the corresponding associated model can be used in the aforementioned application scenarios.

[0147] Information 15: Model version information

[0148] As different parameters in the same application scenario change over time, the algorithm data of the corresponding model may need to be fine-tuned or updated. The updated model is associated with the same application scenario and functions as the original model, but the model algorithm data details differ slightly. Therefore, model algorithm version numbers can be used to distinguish them, facilitating the control and management of similar model algorithms.

[0149] Information 16: Model accuracy level information

[0150] Even for models with the same functionality and application scenario, different design complexities in the model algorithm will result in different inference accuracy. To distinguish between similar models in terms of inference accuracy, the model algorithm's accuracy level can be used to describe the inference accuracy of a given model.

[0151] Information 17: Compilation format information of model data

[0152] Model algorithms are essentially programs written in computer languages. Program file types differ depending on the language platform. Some languages ​​are compatible with each other, while others are not. Therefore, the compilation format of model algorithm data is crucial for describing a model. This information allows model users to understand the required compilation environment. Otherwise, mistakenly using an incompatible language compilation environment will result in the inability to use the corresponding model algorithm.

[0153] Information 18: Storage format information of model data

[0154] This information is used to indicate the format in which the model data is stored, such as zip compression format.

[0155] Information 19: Model computational complexity information

[0156] This information is used to indicate the basic computing power required to use the model, for example, the number of floating-point numbers that need to be calculated per second.

[0157] Information 20: Model complexity information

[0158] This information is used to indicate the complexity of the model, for example: how many layers of neurons the model consists of.

[0159] Information 21: Model size information

[0160] This information is used to indicate the memory size occupied by the model data before or after compilation.

[0161] Example 2-2, the parameter type range in the expected model input parameter type range that is not selected as the model input parameter associated with the target model: For example: the expected model input parameters include four types of parameters: a, b, c, and d. The model input parameters associated with the trained target model only include parameters a and b. Then parameters c and d belong to the parameter type range that is not selected as the model input parameters associated with the target model.

[0162] Example 2-3, contribution rate or contribution rate level information corresponding to each parameter type included in the expected model input parameter type range that is not selected as the model input parameter associated with the target model: For example: the expected model input parameters include four types of parameters: a, b, c, and d. The model input parameters associated with the trained target model only include parameters a and b. Then parameters c and d belong to the parameter type range that is not selected as the model input parameter associated with the target model. Among them, adding parameter c as a model input only obtains an additional 5% performance gain, while adding parameter d as a model input actually reduces the overall model performance by 10%. In this case, the contribution rate corresponding to parameter c is considered to be positive, or 5%, and the contribution rate corresponding to parameter d is considered to be negative, or -10%.

[0163] Example 2-4 shows the contribution rate or contribution rate level information for each parameter type included in the expected model input parameter type range: The expected model input parameters include four types of parameters: a, b, c, and d. The model input parameters associated with the trained target model only include parameters a and b, and parameters a and b are essential for target model prediction or inference. In this case, the contribution rates of parameters a and b are both 100%. At the same time, adding parameter c as a model input only yields an additional 5% performance gain, while adding parameter d as a model input actually reduces the overall model performance by 10%. In this case, the contribution rate of parameter c is considered positive, or 5%, and the contribution rate of parameter d is considered negative, or -10%.

[0164] Example 2-5, the parameter type range in the expected model output parameter type range that is not selected as the model output parameter associated with the target model: For example: the expected model output parameters include four types of parameters: m, n, p, and q. The model output parameters associated with the trained target model only include parameters m and n. Then parameters p and q belong to the parameter type range that is not selected as the model output parameters associated with the target model.

[0165] Example 2-6, prediction accuracy or prediction accuracy level information corresponding to each parameter type included in the parameter type range that is not selected as the model output parameter associated with the target model within the expected model output parameter type range: For example: the expected model output parameters include four types of parameters: m, n, p, and q. The model output parameters associated with the trained target model only include parameters m and n. Then, parameters p and q belong to the parameter type range that is not selected as the model output parameter associated with the target model. Using parameter p as the model output only yields a 30% prediction accuracy, while using parameter q as the model output only yields a 10% prediction accuracy. In this case, the prediction accuracy corresponding to parameter p is considered to be 30%, while the prediction accuracy corresponding to parameter q is considered to be 10%.

[0166] Example 2-7, prediction accuracy or prediction accuracy level information corresponding to each parameter type included in the expected model output parameter type range: For example: the expected model output parameters include four types of parameters: m, n, p, and q. The model output parameters associated with the trained target model only include parameters m and n. If the prediction accuracy of parameters m and n are 90% and 85%, respectively, the prediction accuracy corresponding to parameter m is considered to be 90%, and the prediction accuracy corresponding to parameter n is 85%. At the same time, using parameter p as the model output only yields a prediction accuracy of 30%, and using parameter q as the model output only yields a prediction accuracy of 10%. In this case, the prediction accuracy corresponding to parameter p is considered to be 30%, and the prediction accuracy corresponding to parameter q is considered to be 10%.

[0167] In one embodiment, the model training report and the second model are associated; or the model description information and the second model are associated. In an embodiment of the present application, the first device may associate the model training report information with the target model; and / or may associate the model description information with the target model. The generated model training report is conducive to backtracking the model training process and adjusting the model training strategy, for example, performing iterative model training to obtain a target model with better performance.

[0168] In one embodiment, as shown in FIG3 , the method further includes:

[0169] S340: The first device sends first information, where the first information includes information related to the second model.

[0170] In one embodiment, the first information includes one or more of the following information:

[0171] The model training report;

[0172] The index number associated with the model training report;

[0173] The index number associated with the second model;

[0174] The model describes the information;

[0175] The model describes the index number of the information association;

[0176] The second model.

[0177] In one embodiment, the first information is used to request one or more model identifiers of the second model and / or to exchange information related to the second model.

[0178] In one embodiment, as shown in FIG3 , the method further includes:

[0179] S350: The first device receives second information, where the second information is used to assign one or more model identifiers of the second model.

[0180] In one embodiment, the second information includes one or more of the following information:

[0181] Model identification information assigned to the second model;

[0182] The index number associated with the model training report;

[0183] The index number associated with the second model;

[0184] This model describes the index number of the information association.

[0185] In one embodiment, the first device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

[0186] As the provider of model training services, the first device can send the second model related information to the model training task initiator or model demander, such as the second device, third device, or fourth device, through the first message after one or more second model training is completed, thereby completing the model training task feedback. If it is considered that the first device obtains the assigned model identifier only when the complete model training and / or the model training task feedback also includes multiple rounds of information interaction with the model demander, then the first message introduces the index number associated with the second model related information to simplify the identification of multiple different second models. For example, if the first information receiver needs to give further instructions to the first information sender (for example, assigning a model identifier to one or more second models or further inquiring about model training details through the second message), it can replace the second model related information itself with the index identifier, and achieve accurate management of the relevant second models through the low-overhead index identifier. Optionally, after the model training is completed, it can also include the completion of the model testing task. The tester can be the first device or other devices in addition to the first device. For example, after the first device sends the first message, it receives a test message sent by the second device, third device, or fourth device. The test message contains some test parameters or use cases. The first device tests the second model based on the information contained in the test message and feeds back the test results to the test task initiator. The test task initiator determines whether to retrain the model (for example, resend the model training configuration information) or confirm that the trained model is valid based on the fed-back test results and / or local policy information, and then assigns a model identifier to the authenticated model, for example, by assigning a model identifier through the second message.

[0187] The recipient of the first information may be the second device or the fourth device, wherein the fourth device may be the same as or different from the third device.

[0188] In one embodiment, the method further includes: the first device sending capability information, where the capability information includes one or more of the following capability information:

[0189] Whether the first device supports the model training function;

[0190] A range of parameter types available to the first device;

[0191] The maximum number of models that can be trained simultaneously on the first device.

[0192] In the embodiment of the present application, the first device sending capability information may be a pre-step. For example, the first device first exchanges capability information with the second device and then executes S210.

[0193] Optionally, any information involved in the embodiments of the present application is carried by any of the following message types:

[0194] NR Positioning Protocol A (NRPPa) message, Long Term Evolution Positioning Protocol (LPP) message, Non-Access Stratum (NAS) message, Radio Resource Control (RRC) message, Media Access Control Control Element (MAC CE) message, Downlink Control Information (DCI) message, Uplink Control Information (UCI) message, Physical Uplink Control Channel (PUCCH) message, Physical Uplink Shared Channel (PUSCH) message, 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.

[0195] Optionally, any information involved in the embodiments of the present application is carried by any one of the following message types: unicast message, multicast message or broadcast message.

[0196] Unicast message (one-to-one): The source transmits unicast messages through a unicast channel. Only terminal devices or network devices that have been allocated the corresponding unicast resources can attempt to receive the unicast message. Unicast messages can also be called dedicated signaling.

[0197] Multicast message (one-to-many): The source transmits the multicast message through the multicast channel. Terminal devices or network devices that are members of the group and within the coverage area of ​​the multicast signal can attempt to receive the multicast message. When a terminal device or network device joins a group, it obtains multicast channel-related resources.

[0198] Broadcast message (one-to-any): The source transmits the broadcast message through the broadcast channel. Any terminal device or network device within the coverage of the broadcast signal can attempt to receive the broadcast message.

[0199] In an embodiment of the present application, the first device, the second device, the third device, and the fourth device may be terminal devices or network devices. The network device may be an access network device or a core network device or an AI / ML model related information management device or an operation administration and maintenance (OAM) device.

[0200] Exemplarily, the access network device can be any one of the following: next generation node B (or new generation access network node) (Generation Node B, gNB), centralized unit (CU), distributed unit (DU), centralized unit-control plane (CU-CP), centralized unit-user plane (CU-UP).

[0201] Exemplarily, the core network device can be any one of the following: Location Management Function (LMF) network element, Network Slice Selection Function (NSSF), Authentication Server Function (AUSF), Unified Data Management (UDM), Access and Mobility Management Function (AMF), Session Management Function (SMF), Policy Control Function (PCF), User Plane Function (UPF), Sensing Control Function (SF), Network Data Analytics Function (NWDAF) network element.

[0202] In an embodiment of the present application, the information configuration method proposed in the embodiment of the present application is applicable to a control method for model training, through which automated management of model training can be achieved. The specific method is as follows: model training configuration information is designed according to the actual model characteristics required, and the model training configuration information constrains multiple aspects of the model. The first device trains the model according to the model training configuration information and generates a model training report. The model training report reflects multiple aspects of the model training, analyzes the characteristics of each model input and / or model output while generating the target model and obtaining the model description information, and the generated model training report can be used as reference information for subsequent targeted optimization operations of model training, thereby obtaining the desired model in the process of iterative model training.

[0203] The information configuration method proposed in this embodiment does not require the precise type of input and / or output parameters actually required by the model before model training. Before model training, it is only necessary to give the model input and / or output parameter type range, and through the model training process, the input and / or output parameter types actually required by the model are selected from the given model input and / or output parameter type range. This model training method is flexible and avoids the problem of system design errors caused by artificially setting model inputs and / or model outputs. Because the artificially set model inputs and / or model outputs are mostly based on the experience of human engineers, when there are many parameters that affect model training, the experience of human engineers often deviates. Therefore, the problem of human engineer experience deviation can be avoided to a large extent by also handing over the selection of model input and / or model output parameters to the model training process.

[0204] Furthermore, by adopting the method of the embodiment of the present application, the types of models that can be trained are very flexible. By changing the content of the model training configuration information, a variety of models that meet the needs can be obtained, which is suitable for deployment in communication systems as a general implementation architecture for model training.

[0205] Furthermore, by adopting the method of the embodiment of the present application, the model training process is basically automated and does not require too much human design intervention, which is conducive to realizing the vision of AI-controlled self-optimizing network.

[0206] FIG4 is a schematic flow chart of an information configuration method according to an embodiment of the present application. The method 400 can optionally be applied to the system shown in FIG1 , but is not limited thereto. The method includes at least part of the following contents.

[0207] S410. The second device sends model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0208] In one embodiment, the model training configuration information includes one or more of the following information: the input parameter type range of the first model, the data source constraint information of the input parameters of the first model, the output parameter type range of the first model, the operating capability requirements of the first model, the output accuracy requirements of the first model, the complexity requirements of the first model, the size requirements of the first model, the algorithm structure requirements of the first model, the contribution rate of the input parameters of the first model or the contribution rate level threshold requirements, and the prediction accuracy rate of the output parameters of the first model or the prediction accuracy level threshold requirements.

[0209] In one embodiment, the input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

[0210] In one embodiment, the input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

[0211] In one embodiment, the data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

[0212] In one embodiment, the output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

[0213] In one embodiment, the output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

[0214] In one embodiment, there is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold;

[0215] Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model. All or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

[0216] The contribution rate level is associated with a contribution rate range, and each contribution rate level is associated with a contribution rate range.

[0217] In one embodiment, there is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold;

[0218] Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirements of the first model. All or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and the effective model output parameter prediction accuracy or prediction accuracy level threshold associated with different groups of output parameter types is different.

[0219] The prediction accuracy level is associated with a prediction accuracy range, and each prediction accuracy level is associated with a prediction accuracy range.

[0220] FIG5 is a schematic flow chart of an information configuration method according to another embodiment of the present application. The method 500 may include one or more features of the information configuration method of the above embodiment. In one implementation, as shown in FIG5 , the method further includes:

[0221] S510: The second device receives first information, where the first information includes information related to the second model.

[0222] In one embodiment, the second model is trained by the first device based on the model training configuration information, and the second model is an instance of the first model.

[0223] In one embodiment, the first information includes one or more of the following information:

[0224] a model training report associated with the second model;

[0225] The index number associated with the model training report;

[0226] The index number associated with the second model;

[0227] Model description information associated with the second model;

[0228] The model describes the index number of the information association;

[0229] The second model.

[0230] In one embodiment, the model training report is generated by the first device.

[0231] In one embodiment, the model training report includes one or more of the following information: model description information associated with the second model, parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

[0232] In one embodiment, the model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective usage scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

[0233] In one embodiment, the model training report and the second model are associated; or the model description information and the second model are associated.

[0234] In one embodiment, the first information is used to request one or more model identifiers of the second model and / or to exchange information related to the second model.

[0235] In one embodiment, as shown in FIG5 , the method further includes:

[0236] S520: The second device sends second information, where the second information is used to allocate one or more model identifiers of the second model.

[0237] In one embodiment, the second information includes one or more of the following information:

[0238] Model identification information assigned to the second model;

[0239] The index number associated with the model training report;

[0240] The index number associated with the second model;

[0241] This model describes the index number of the information association.

[0242] In one embodiment, the second device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

[0243] In one embodiment, the method further includes: the second device receiving capability information, where the capability information includes one or more of the following capability information:

[0244] Whether the first device supports the model training function;

[0245] A range of parameter types available to the first device;

[0246] The maximum number of models that can be trained simultaneously on the first device.

[0247] In one embodiment, the model training configuration information is sent by the second device to the first device, and the model training configuration information is generated by the second device.

[0248] In one embodiment, the model training configuration information is sent by the second device to the first device, the model training configuration information is received by the second device from a third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device receives the model training configuration information directly from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

[0249] In one embodiment, the second device is a terminal device or a network device.

[0250] For specific examples of the information configuration device in this embodiment executing the information configuration methods 400 and 500 , reference can be made to the relevant descriptions of the information configuration device in the above-mentioned information configuration methods 200 and 300 , which will not be repeated here for the sake of brevity.

[0251] FIG6 is a schematic block diagram of a first device 600 according to an embodiment of the present application. The first device 600 may include:

[0252] The acquisition unit 601 is used to acquire model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0253] In one embodiment, the model training configuration information includes one or more of the following information: the input parameter type range of the first model, the data source constraint information of the input parameters of the first model, the output parameter type range of the first model, the operating capability requirements of the first model, the output accuracy requirements of the first model, the complexity requirements of the first model, the size requirements of the first model, the algorithm structure requirements of the first model, the contribution rate of the input parameters of the first model or the contribution rate level threshold requirements, and the prediction accuracy rate of the output parameters of the first model or the prediction accuracy level threshold requirements.

[0254] In one embodiment, the input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

[0255] In one embodiment, the input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

[0256] In one embodiment, the data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

[0257] In one embodiment, the output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

[0258] In one embodiment, the output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

[0259] In one embodiment, there is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold;

[0260] Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model. All or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

[0261] In one embodiment, there is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold;

[0262] Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirements of the first model. All or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and the effective model output parameter prediction accuracy or prediction accuracy level threshold associated with different groups of output parameter types is different.

[0263] In one embodiment, as shown in FIG7 , the first device 700 further includes:

[0264] The first processing unit 701 is used to determine model training information based on the model training configuration information; perform model training based on the model training information to obtain one or more second models, where the second model is an instance of the first model.

[0265] In one embodiment, as shown in FIG7 , the first device further includes:

[0266] The second processing unit 702 is used to generate a model training report.

[0267] In one embodiment, the model training report includes one or more of the following information: model description information associated with the second model, the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, the contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, the contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, the prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and the prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

[0268] In one embodiment, the model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective usage scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

[0269] In one embodiment, the model training report and the second model are associated; or the model description information and the second model are associated.

[0270] In one embodiment, as shown in FIG7 , the first device further includes:

[0271] The first sending unit 703 is configured to send first information, where the first information includes information related to the second model.

[0272] In one embodiment, the first information includes one or more of the following information:

[0273] The model training report;

[0274] The index number associated with the model training report;

[0275] The index number associated with the second model;

[0276] The model describes the information;

[0277] The model describes the index number of the information association;

[0278] The second model.

[0279] In one embodiment, the first information is used to request one or more model identifiers of the second model and / or to exchange information related to the second model.

[0280] In one embodiment, as shown in FIG7 , the first device further includes:

[0281] The first receiving unit 704 is configured to receive second information, where the second information is used to assign one or more model identifiers of the second model.

[0282] In one embodiment, the second information includes one or more of the following information:

[0283] Model identification information assigned to the second model;

[0284] The index number associated with the model training report;

[0285] The index number associated with the second model;

[0286] This model describes the index number of the information association.

[0287] In one embodiment, the first device stores one or more of the following information:

[0288] The model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

[0289] In one embodiment, as shown in FIG7 , the first device further includes:

[0290] The second sending unit 705 is configured to send capability information, where the capability information includes one or more of the following capability information:

[0291] Whether the first device supports the model training function;

[0292] A range of parameter types available to the first device;

[0293] The maximum number of models that can be trained simultaneously on the first device.

[0294] In one embodiment, the model training configuration information is generated by the first device.

[0295] In one embodiment, the model training configuration information is received by the first device from the second device, and the model training configuration information is generated by the second device.

[0296] In one embodiment, the model training configuration information is received by the first device from the second device, the model training configuration information is received by the second device from the third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device receives the model training configuration information directly from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

[0297] In one embodiment, the first device is a terminal device or a network device.

[0298] The first devices 600 and 700 of the embodiments of the present application can implement the corresponding functions of the first devices in the aforementioned method embodiments. The processes, functions, implementation methods and beneficial effects corresponding to the various modules (sub-modules, units or components, etc.) in the first devices 600 and 700 can be found in the corresponding descriptions in the aforementioned method embodiments and will not be repeated here. It should be noted that the functions described in the various modules (sub-modules, units or components, etc.) in the first devices 600 and 700 of the application embodiments can be implemented by different modules (sub-modules, units or components, etc.) or by the same module (sub-module, unit or component, etc.).

[0299] FIG8 is a schematic block diagram of a second device 800 according to an embodiment of the present application. The second device 800 may include:

[0300] The first sending unit 801 is used to send model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

[0301] In one embodiment, the model training configuration information includes one or more of the following information:

[0302] The input parameter type range of the first model, the data source constraint information of the input parameters of the first model, the output parameter type range of the first model, the operating capability requirements of the first model, the output accuracy requirements of the first model, the complexity requirements of the first model, the size requirements of the first model, the algorithm structure requirements of the first model, the contribution rate of the input parameters of the first model or the contribution rate level threshold requirements, and the prediction accuracy rate or the prediction accuracy level threshold requirements of the output parameters of the first model.

[0303] In one embodiment, the input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

[0304] In one embodiment, the input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

[0305] In one embodiment, the data source constraint information of the input parameters of the first model includes one or more of the following information:

[0306] Source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

[0307] In one embodiment, the output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

[0308] In one embodiment, the output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

[0309] In one embodiment, there is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold;

[0310] Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model. All or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

[0311] In one embodiment, there is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold;

[0312] Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirements of the first model. All or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and the effective model output parameter prediction accuracy or prediction accuracy level threshold associated with different groups of output parameter types is different.

[0313] In one embodiment, as shown in FIG9 , the first device 900 further includes:

[0314] The first receiving unit 901 is configured to receive first information, where the first information includes information related to the second model.

[0315] In one embodiment, the second model is trained by the first device based on the model training configuration information, and the second model is an instance of the first model.

[0316] In one embodiment, the first information includes one or more of the following information:

[0317] a model training report associated with the second model;

[0318] The index number associated with the model training report;

[0319] The index number associated with the second model;

[0320] Model description information associated with the second model;

[0321] The model describes the index number of the information association;

[0322] The second model.

[0323] In one embodiment, the model training report is generated by the first device.

[0324] In one embodiment, the model training report includes one or more of the following information: model description information associated with the second model, parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

[0325] In one embodiment, the model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective usage scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

[0326] In one embodiment, the model training report and the second model are associated; or the model description information and the second model are associated.

[0327] In one embodiment, the first information is used to request one or more model identifiers of the second model and / or to exchange information related to the second model.

[0328] In one embodiment, as shown in FIG9 , the second device 900 further includes:

[0329] The second sending unit 902 is configured to send second information, where the second information is used to allocate one or more model identifiers of the second model.

[0330] In one embodiment, the second information includes one or more of the following information:

[0331] Model identification information assigned to the second model;

[0332] The index number associated with the model training report;

[0333] The index number associated with the second model;

[0334] This model describes the index number of the information association.

[0335] In one embodiment, the second device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

[0336] In one embodiment, as shown in FIG9 , the second device 900 further includes:

[0337] The second receiving unit 903 is configured to receive capability information, where the capability information includes one or more of the following capability information:

[0338] Whether the first device supports the model training function;

[0339] A range of parameter types available to the first device;

[0340] The maximum number of models that can be trained simultaneously on the first device.

[0341] In one embodiment, the model training configuration information is sent by the second device to the first device, and the model training configuration information is generated by the second device.

[0342] In one embodiment, the model training configuration information is sent by the second device to the first device, the model training configuration information is received by the second device from a third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device receives the model training configuration information directly from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

[0343] In one embodiment, the second device is a terminal device or a network device.

[0344] The second device 800, 900 of the embodiment of the present application can implement the corresponding functions of the second device in the aforementioned method embodiment. The processes, functions, implementation methods and beneficial effects corresponding to the various modules (sub-modules, units or components, etc.) in the second device 800, 900 can be found in the corresponding descriptions in the above-mentioned method embodiments, which will not be repeated here. It should be noted that the functions described in the various modules (sub-modules, units or components, etc.) in the second device 800, 900 of the embodiment of the application can be implemented by different modules (sub-modules, units or components, etc.) or by the same module (sub-module, unit or component, etc.).

[0345] Figure 10 is a schematic structural diagram of a communication device 1000 according to an embodiment of the present application. The communication device 1000 includes a processor 1010, which can call and execute a computer program from a memory to enable the communication device 1000 to implement the method in the embodiment of the present application.

[0346] In one embodiment, the communication device 1000 may further include a memory 1020. The processor 1010 may call and execute a computer program from the memory 1020 to enable the communication device 1000 to implement the method in the embodiment of the present application.

[0347] The memory 1020 may be a separate device independent of the processor 1010 , or may be integrated into the processor 1010 .

[0348] In one embodiment, the communication device 1000 may further include a transceiver 1030 , and the processor 1010 may control the transceiver 1030 to communicate with other devices. Specifically, the transceiver 1030 may send information or data to other devices, or receive information or data sent by other devices.

[0349] The transceiver 1030 may include a transmitter and a receiver. The transceiver 1030 may further include an antenna, and the number of antennas may be one or more.

[0350] In one embodiment, the communication device 1000 may be the first device of the embodiment of the present application, and the communication device 1000 may implement the corresponding processes implemented by the network device in each method of the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0351] In one embodiment, the communication device 1000 may be the second device of the embodiment of the present application, and the communication device 1000 may implement the corresponding processes implemented by the terminal device in each method of the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0352] 11 is a schematic structural diagram of a chip 1100 according to an embodiment of the present application. The chip 1100 includes a processor 1110, which can call and execute a computer program from a memory to implement the method according to the embodiment of the present application.

[0353] In one embodiment, the chip 1100 may further include a memory 1120. The processor 1110 may call and execute a computer program from the memory 1120 to implement the method executed by the terminal device or the network device in the embodiment of the present application.

[0354] The memory 1120 may be a separate device independent of the processor 1110 , or may be integrated into the processor 1110 .

[0355] In one embodiment, the chip 1100 may further include an input interface 1130. The processor 1110 may control the input interface 1130 to communicate with other devices or chips, and specifically, may obtain information or data sent by other devices or chips.

[0356] In one embodiment, the chip 1100 may further include an output interface 1140. The processor 1110 may control the output interface 1140 to communicate with other devices or chips, and specifically, may output information or data to other devices or chips.

[0357] In one embodiment, the chip can be applied to the first device in the embodiment of the present application, and the chip can implement the corresponding processes implemented by the network device in each method of the embodiment of the present application. For the sake of brevity, it will not be repeated here.

[0358] In one embodiment, the chip can be applied to the second device in the embodiment of the present application, and the chip can implement the corresponding processes implemented by the terminal device in each method of the embodiment of the present application. For the sake of brevity, it will not be repeated here.

[0359] The chips used in the first device and the second device may be the same chip or different chips.

[0360] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0361] The processor mentioned above may be a general-purpose processor, a digital signal processor (DSP), a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or other programmable logic devices, transistor logic devices, discrete hardware components, etc. The general-purpose processor mentioned above may be a microprocessor or any conventional processor, etc.

[0362] The memory mentioned above may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM).

[0363] It should be understood that the above-mentioned memories are exemplary but not restrictive. For example, the memories in the embodiments of the present application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM RAM (DR RAM), etc. In other words, the memories in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0364] Figure 12 is a schematic block diagram of a communication system 1200 according to an embodiment of the present application. The communication system 1200 includes a first device 1210 and a second device 1220. The first device 1210 obtains model training configuration information, which is used to indicate information related to the training of the first model. The second device 1220 sends the model training configuration information.

[0365] The first device 1210 can be used to implement the corresponding functions implemented by the first device in the above method, and the second device 1220 can be used to implement the corresponding functions implemented by the second device in the above method. For the sake of brevity, they are not described here in detail.

[0366] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When software is used for implementation, 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 in accordance with 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 devices. 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 a website, computer, server or data center by wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode to another website, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrations. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0367] It should be understood that in the 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.

[0368] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0369] 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 modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included within the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.< / position>

Claims

1. An information configuration method, comprising: The first device obtains model training configuration information, where the model training configuration information is used to indicate information related to training of the first model.

2. The method according to claim 1, wherein: The model training configuration information includes one or more of the following information: The input parameter type range of the first model, data source constraint information of the input parameters of the first model, the output parameter type range of the first model, the operating capability requirements of the first model, the output accuracy requirements of the first model, the complexity requirements of the first model, the size requirements of the first model, the algorithm structure requirements of the first model, the contribution rate of the input parameters of the first model or the contribution rate level threshold requirements, and the prediction accuracy rate or the prediction accuracy level threshold requirements of the output parameters of the first model.

3. The method according to claim 2, wherein: The input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

4. The method according to claim 2 or 3, wherein: The input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

5. The method according to any one of claims 2 to 4, wherein: The data source constraint information of the input parameters of the first model includes one or more of the following information: Source country information associated with the data of the input parameters of the first model, source geographical area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

6. The method according to any one of claims 2 to 5, wherein: The output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

7. The method according to any one of claims 2 to 6, wherein: The output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

8. The method according to any one of claims 2 to 7, wherein: There is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold; Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and all or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

9. The method according to any one of claims 2 to 8, wherein: There is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold; Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, all or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and different groups of output parameter types are associated with different valid model output parameter prediction accuracy or prediction accuracy level thresholds.

10. The method according to any one of claims 2 to 9, wherein: The method further comprises: The first device determines model training information based on the model training configuration information; The first device performs model training based on the model training information to obtain one or more second models, where the second models are instances of the first model.

11. The method according to any one of claims 1 to 10, wherein: The method further comprises: The first device generates a model training report.

12. The method according to claim 11, wherein: The model training report includes one or more of the following information: model description information associated with the second model, parameter type ranges in the input parameter type range of the first model that are not selected as model input parameters associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type included in the parameter type range of the input parameter type range of the first model that are not selected as model input parameters associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type included in the input parameter type range of the first model, parameter type ranges in the output parameter type range of the first model that are not selected as model output parameters associated with the second model, output parameters of the first model The prediction accuracy or prediction accuracy level information corresponding to each parameter type contained in the parameter type range that is not selected as the model output parameter associated with the second model, and the prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

13. The method according to claim 12, wherein: The model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective use scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

14. The method according to claim 12 or 13, wherein: The model training report and the second model are associated with each other; or the model description information and the second model are associated with each other.

15. The method according to any one of claims 12 to 14, wherein: The method further comprises: The first device sends first information, where the first information includes information related to the second model.

16. The method according to claim 15, wherein: The first information includes one or more of the following information: The model training report; The index number associated with the model training report; The index number associated with the second model; The model description information; The index number associated with the model description information; The second model.

17. The method according to claim 15 or 16, wherein: The first information is used to request one or more model identifiers of the second models and / or to exchange information related to the second model.

18. The method according to any one of claims 15 to 17, wherein: The method further comprises: The first device receives second information, where the second information is used to assign one or more model identifiers of the second model.

19. The method according to claim 18, wherein: The second information includes one or more of the following information: Model identification information assigned to the second model; The index number associated with the model training report; The index number associated with the second model; The index number associated with the model description information.

20. The method according to any one of claims 12 to 19, wherein: The first device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

21. The method according to any one of claims 1 to 20, wherein: The method further comprises: The first device sends capability information, where the capability information includes one or more of the following capability information: Whether the first device supports the model training function; A range of parameter types available to the first device; The maximum number of models that can be trained simultaneously on the first device.

22. The method according to any one of claims 1 to 21, wherein: The model training configuration information is generated by the first device.

23. The method according to any one of claims 1 to 21, wherein: The model training configuration information is received by the first device from the second device, and the model training configuration information is generated by the second device.

24. The method according to any one of claims 1 to 21, wherein: The model training configuration information is received by the first device from the second device, the model training configuration information is received by the second device from the third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device directly receives the model training configuration information from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

25. The method according to any one of claims 1 to 24, wherein: The first device is a terminal device or a network device.

26. An information configuration method, comprising: The second device sends model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

27. The method according to claim 26, wherein: The model training configuration information includes one or more of the following information: input parameter type range of the first model, data source constraint information of the input parameters of the first model, output parameter type range of the first model, operating capability requirements of the first model, output accuracy requirements of the first model, complexity requirements of the first model, size requirements of the first model, algorithm structure requirements of the first model, contribution rate of input parameters of the first model or contribution rate level threshold requirements, and prediction accuracy rate of output parameters of the first model or prediction accuracy level threshold requirements.

28. The method according to claim 27, wherein: The input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

29. The method according to claim 27 or 28, wherein: The input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

30. The method according to any one of claims 27 to 29, wherein: The data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

31. The method according to any one of claims 27 to 30, wherein: The output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

32. The method according to any one of claims 27 to 31, wherein: The output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

33. The method according to any one of claims 27 to 32, wherein: There is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold; Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and all or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

34. A method according to any one of claims 27 to 33, wherein: There is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold; Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, all or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and different groups of output parameter types are associated with different valid model output parameter prediction accuracy or prediction accuracy level thresholds.

35. The method according to any one of claims 26 to 34, wherein: The method further comprises: The second device receives first information, where the first information includes information related to the second model.

36. The method of claim 35, wherein: The second model is trained by the first device based on the model training configuration information, and the second model is an instance of the first model.

37. The method according to claim 35 or 36, wherein: The first information includes one or more of the following information: a model training report associated with the second model; The index number associated with the model training report; The index number associated with the second model; model description information associated with the second model; The index number associated with the model description information; The second model.

38. The method of claim 37, wherein: The model training report is generated by the first device.

39. The method according to claim 37 or 38, wherein: The model training report includes one or more of the following information: model description information associated with the second model, parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

40. The method of claim 39, wherein: The model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective use scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

41. The method according to any one of claims 37 to 40, wherein: The model training report and the second model are associated with each other; or the model description information and the second model are associated with each other.

42. The method according to any one of claims 35 to 40, wherein: The first information is used to request one or more model identifiers of the second models and / or to exchange information related to the second model.

43. The method according to any one of claims 35 to 42, wherein: The method further comprises: The second device sends second information, where the second information is used to allocate one or more model identifiers of the second models.

44. The method of claim 43, wherein: The second information includes one or more of the following information: Model identification information assigned to the second model; The index number associated with the model training report; The index number associated with the second model; The index number associated with the model description information.

45. The method according to any one of claims 35 to 44, wherein: The second device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

46. ​​The method according to any one of claims 26 to 45, wherein: The method further comprises: The second device receives capability information, where the capability information includes one or more of the following capability information: Whether the first device supports the model training function; A range of parameter types available to the first device; The maximum number of models that can be trained simultaneously on the first device.

47. The method according to any one of claims 26 to 46, wherein: The model training configuration information is sent by the second device to the first device, and the model training configuration information is generated by the second device.

48. The method according to any one of claims 26 to 46, wherein: The model training configuration information is sent by the second device to the first device, the model training configuration information is received by the second device from a third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device directly receives the model training configuration information from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

49. The method according to any one of claims 26 to 48, wherein: The second device is a terminal device or a network device.

50. A first device, comprising: An acquisition unit is used to acquire model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

51. The first device according to claim 50, wherein The model training configuration information includes one or more of the following information: input parameter type range of the first model, data source constraint information of the input parameters of the first model, output parameter type range of the first model, operating capability requirements of the first model, output accuracy requirements of the first model, complexity requirements of the first model, size requirements of the first model, algorithm structure requirements of the first model, contribution rate of input parameters of the first model or contribution rate level threshold requirements, and prediction accuracy rate of output parameters of the first model or prediction accuracy level threshold requirements.

52. The first device according to claim 51, wherein The input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

53. The first device according to claim 51 or 52, wherein: The input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

54. A first device according to any one of claims 51 to 53, wherein The data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

55. A first device according to any one of claims 51 to 54, wherein: The output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

56. A first device according to any one of claims 51 to 55, wherein: The output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

57. A first device according to any one of claims 51 to 56, wherein: There is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold; Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and all or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

58. A first device according to any one of claims 51 to 57, wherein: There is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold; Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, all or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and different groups of output parameter types are associated with different valid model output parameter prediction accuracy or prediction accuracy level thresholds.

59. A first device according to any one of claims 51 to 58, wherein The first device also includes: The first processing unit is used to determine model training information based on the model training configuration information; perform model training based on the model training information to obtain one or more second models, where the second models are instances of the first model.

60. The first device according to any one of claims 50 to 59, wherein: The device also includes: The second processing unit is used to generate a model training report.

61. The first device according to claim 60, wherein: The model training report includes one or more of the following information: model description information associated with the second model, parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

62. The first device according to claim 61, wherein The model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective use scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

63. The first device according to claim 61 or 62, wherein: The model training report and the second model are associated with each other; or the model description information and the second model are associated with each other.

64. A first device according to any one of claims 61 to 63, wherein The first device also includes: The first sending unit is used to send first information, where the first information includes information related to the second model.

65. The first device according to claim 64, wherein The first information includes one or more of the following information: The model training report; The index number associated with the model training report; The index number associated with the second model; The model description information; The index number associated with the model description information; The second model.

66. A first device according to claim 64 or 65, wherein The first information is used to request one or more model identifiers of the second models and / or to exchange information related to the second model.

67. A first device according to any one of claims 64 to 66, wherein: The first device also includes: The first receiving unit is used to receive second information, where the second information is used to allocate one or more model identifiers of the second model.

68. The first device according to claim 67, wherein The second information includes one or more of the following information: Model identification information assigned to the second model; The index number associated with the model training report; The index number associated with the second model; The index number associated with the model description information.

69. A first device according to any one of claims 61 to 68, wherein The first device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

70. The first device according to any one of claims 50 to 69, wherein: The first device also includes: The second sending unit is configured to send capability information, where the capability information includes one or more of the following capability information: Whether the first device supports the model training function; A range of parameter types available to the first device; The maximum number of models that can be trained simultaneously on the first device.

71. A first device according to any one of claims 50 to 70, wherein: The model training configuration information is generated by the first device.

72. The first device according to any one of claims 50 to 70, wherein: The model training configuration information is received by the first device from the second device, and the model training configuration information is generated by the second device.

73. A first device according to any one of claims 50 to 70, wherein: The model training configuration information is received by the first device from the second device, the model training configuration information is received by the second device from the third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device directly receives the model training configuration information from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

74. A first device according to any one of claims 50 to 73, wherein The first device is a terminal device or a network device.

75. A second device, comprising: The first sending unit is used to send model training configuration information, where the model training configuration information is used to indicate information related to the training of the first model.

76. The second device according to claim 75, wherein The model training configuration information includes one or more of the following information: input parameter type range of the first model, data source constraint information of the input parameters of the first model, output parameter type range of the first model, operating capability requirements of the first model, output accuracy requirements of the first model, complexity requirements of the first model, size requirements of the first model, algorithm structure requirements of the first model, contribution rate of input parameters of the first model or contribution rate level threshold requirements, and prediction accuracy rate of output parameters of the first model or prediction accuracy level threshold requirements.

77. The second device according to claim 76, wherein The input parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model input parameter type.

78. The second device according to claim 76 or 77, wherein The input parameter type range of the first model includes first-category parameters and second-category parameters, the first-category parameters include identification information and / or label information corresponding to at least one determined model input parameter type, and the second-category parameters include identification information and / or label information corresponding to at least one suggested model input parameter type.

79. A second device according to any one of claims 76 to 78, wherein The data source constraint information of the input parameters of the first model includes one or more of the following information: source country information associated with the data of the input parameters of the first model, source geographic area information associated with the data of the input parameters of the first model, source operator information associated with the data of the input parameters of the first model, source logical area information associated with the data of the input parameters of the first model, source scenario information associated with the data of the input parameters of the first model, and source configuration information associated with the data of the input parameters of the first model.

80. The second device according to any one of claims 76 to 79, wherein The output parameter type range of the first model includes identification information and / or label information corresponding to at least one suggested model output parameter type.

81. The second device according to any one of claims 76 to 80, wherein The output parameter type range of the first model includes third-category parameters and fourth-category parameters, the third-category parameters include identification information and / or label information corresponding to at least one determined model output parameter type, and the fourth-category parameters include identification information and / or label information corresponding to at least one suggested model output parameter type.

82. The second device according to any one of claims 76 to 81, wherein There is a one-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and each input parameter type included in the input parameter type range of the first model is associated with a first model input parameter contribution rate or contribution rate level threshold; Alternatively, there is a many-to-one mapping relationship between the input parameter types included in the input parameter type range of the first model and the input parameter contribution rate or contribution rate level threshold requirement of the first model, and all or a group of input parameter types included in the input parameter type range of the first model are associated with a first model input parameter contribution rate or contribution rate level threshold, and different groups of input parameter types are associated with different first model input parameter contribution rates or contribution rate level thresholds.

83. A second device according to any one of claims 76 to 82, wherein: There is a one-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, and each output parameter type included in the output parameter type range of the first model is associated with a first model output parameter prediction accuracy or prediction accuracy level threshold; Alternatively, there is a many-to-one mapping relationship between the output parameter types included in the output parameter type range of the first model and the output parameter prediction accuracy or prediction accuracy level threshold requirement of the first model, all or a group of output parameter types included in the output parameter type range of the first model are associated with a first model output parameter prediction accuracy or prediction accuracy level threshold, and different groups of output parameter types are associated with different valid model output parameter prediction accuracy or prediction accuracy level thresholds.

84. A second device according to any one of claims 75 to 83, wherein The second device also includes: The first receiving unit is used to receive first information, where the first information includes information related to the second model.

85. The second device according to claim 84, wherein The second model is trained by the first device based on the model training configuration information, and the second model is an instance of the first model.

86. The second device according to claim 84 or 85, wherein The first information includes one or more of the following information: a model training report associated with the second model; The index number associated with the model training report; The index number associated with the second model; model description information associated with the second model; The index number associated with the model description information; The second model.

87. The second device according to claim 86, wherein The model training report is generated by the first device.

88. The second device according to claim 86 or 87, wherein The model training report includes one or more of the following information: model description information associated with the second model, parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the parameter type range in the input parameter type range of the first model that is not selected as the model input parameter associated with the second model, contribution rate or contribution rate grade information corresponding to each parameter type contained in the input parameter type range of the first model, parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, prediction accuracy or prediction accuracy grade information corresponding to each parameter type contained in the parameter type range in the output parameter type range of the first model that is not selected as the model output parameter associated with the second model, and prediction accuracy or prediction accuracy information corresponding to each parameter type contained in the output parameter type range of the first model.

89. The second device according to claim 88, wherein The model description information associated with the second model includes one or more of the following information: functional characteristics associated with the second model, input parameter type range associated with the second model, input parameter format requirement information associated with the second model, input parameter preprocessing rules associated with the second model, output parameter type range associated with the second model, output parameter format requirement information associated with the second model, output parameter preprocessing rules associated with the second model, application scenario information associated with the second model, deployment location information of the second model, capability requirement information for using the second model, performance monitoring indicator information of the second model, location information of the second model, effective use scope information of the second model, generalization feature information of the second model, version information of the second model, accuracy level information of the second model, compilation format information of the second model data, storage format information of the second model data, computational complexity information of the second model, model complexity information of the second model, and model size information of the second model.

90. The second device according to any one of claims 86 to 89, wherein The model training report and the second model are associated with each other; or the model description information and the second model are associated with each other.

91. The second device according to any one of claims 84 to 89, wherein The first information is used to request one or more model identifiers of the second models and / or to exchange information related to the second model.

92. The second device according to any one of claims 84 to 91, wherein The second device also includes: The second sending unit is used to send second information, where the second information is used to allocate one or more model identifiers of the second model.

93. The second device according to claim 92, wherein: The second information includes one or more of the following information: Model identification information assigned to the second model; The index number associated with the model training report; The index number associated with the second model; The index number associated with the model description information.

94. The second device according to any one of claims 84 to 93, wherein The second device stores one or more of the following information: the model training report associated with the second model, the second model, the model identifier of the second model, and the model description information associated with the second model.

95. The second device according to any one of claims 75 to 94, wherein The second device also includes: The second receiving unit is configured to receive capability information, where the capability information includes one or more of the following capability information: Whether the first device supports the model training function; A range of parameter types available to the first device; The maximum number of models that can be trained simultaneously on the first device.

96. The second device according to any one of claims 75 to 95, wherein The model training configuration information is sent by the second device to the first device, and the model training configuration information is generated by the second device.

97. The second device according to any one of claims 75 to 95, wherein The model training configuration information is sent by the second device to the first device, the model training configuration information is received by the second device from a third device, and the model training configuration information is generated by the third device; wherein the third device and the second device are different devices, the second device directly receives the model training configuration information from the third device, or the second device receives the model training configuration information from the third device through one or more relay devices.

98. The second device according to any one of claims 75 to 97, wherein The second device is a terminal device or a network device.

99. A communication device comprising: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory so that the communication device executes the method as described in any one of claims 1 to 25 or any one of claims 26 to 49.

100. A chip, comprising: A processor, configured to call and run a computer program from a memory so that a device equipped with the chip executes a method as claimed in any one of claims 1 to 25 or any one of claims 26 to 49.

101. A computer-readable storage medium for storing a computer program, which, when executed by a device, causes the device to perform a method as claimed in any one of claims 1 to 25 or any one of claims 26 to 49.

102. A computer program product comprising computer program instructions, the computer program instructions causing a computer to perform the method of any one of claims 1 to 25 or any one of claims 26 to 49.

103. A computer program causing a computer to perform the method of any one of claims 1 to 25 or any one of claims 26 to 49.

104. A communication system comprising: A first device, configured to perform the method according to any one of claims 1 to 25; A second device, configured to execute the method according to any one of claims 26 to 49.

Citation Information

Patent Citations

  • Model training method, model prediction method, and model control system

    CN112799850A

  • Management and control model training method, device and system

    CN115146691A

  • Communication method and device

    CN115802370A

  • Communication method and device

    CN116055012A

Cited By

  • Substation equipment operation data monitoring method and system

    CN121348170A

  • Wireless communication method, terminal device and network device

    WO2026137336A1

  • Configuration request method, terminal device, and network device

    WO2026148548A1