Communication method and device

Abstract

The invention discloses a communication method and device, and the method comprises the steps: receiving a first request message which comprises model demand information; and sending a first response message, the first response message comprising information of a first model or information of a second model, the first model being used for training to obtain a model satisfying the model demand information, and the second model being a model satisfying the model demand information obtained by training the first model. According to the communication method and device provided by the embodiment of the invention, the new model capable of meeting the model demand information can be obtained by continuously training the existing model, and compared with the new model capable of meeting the model demand information obtained by training from zero, on one hand, the model can be more conveniently obtained; the time and / or power consumption of model training and model requesting can be reduced, and the efficiency of model training and model requesting can be improved; and on the other hand, the utilization rate of the existing model can be improved, so that the existing model is utilized more reasonably.

Description

Technical Field

[0001] This application relates to the field of communications, specifically to a communication method and apparatus. Background Technology

[0002] When a user requests a model from the model repository, if the model repository has an AI / ML model that meets the model selection criteria stored locally or on other devices, the model repository can send that AI / ML model to the user. If the model repository does not have an AI / ML model that meets the model selection criteria stored locally or on other devices, a device with model training capabilities can train a new model. This new model is trained from scratch and meets the model selection criteria.

[0003] When there are no AI / ML models that meet the model selection criteria in the model repository or other stored models, training a new model from scratch requires a long training time and is inefficient. Summary of the Invention

[0004] This application provides a communication method and apparatus that can reduce model request time and improve model request efficiency.

[0005] In a first aspect, a communication method is provided, applied to a first device, comprising: receiving a first request message, the first request message including model requirement information; and sending a first response message, the first response message including information of a first model or information of a second model, wherein the first model is used to train a model that satisfies the model requirement information, and the second model is a model that satisfies the model requirement information obtained by training the first model.

[0006] Based on the solution provided in the embodiments of this application, information about a first model that can meet the model requirements after continued training is sent to the requesting party through a first response message, or information about a second model that has been trained and can meet the model requirements is sent. This allows for the acquisition of a new model that can meet the model requirements by continuing to train an existing model. Compared to acquiring a new model that can meet the model requirements by training from scratch, this approach reduces the time and / or power consumption of model training and model request, thereby improving the efficiency of model training and model request. Furthermore, it enhances the utilization rate of existing models, making them more rationally used.

[0007] In some possible implementations, receiving the first request message includes: receiving a first request message from the second device, wherein the first request message further includes at least one of the following: whether the second device has the ability to train a model; and whether the second device has the ability to continue training a model.

[0008] It is understood that whether the second device has the ability to train a model and / or whether the second device has the ability to continue training a model can be indicated by carrying the first request message or by other messages besides the first request message. This application embodiment does not impose any limitations on this.

[0009] Based on the solution provided in the embodiments of this application, a first request message is sent to the first device to indicate whether the second device has the ability to train / continue training the model. This enables the first device to determine whether it can send information about a first model that needs to be trained to meet the model requirements to the second device based on whether the second device has the ability to train / continue training the model. This makes the response to the model request more reasonable and avoids sending information about a first model that needs to be trained to meet the model requirements to the second device when the second device does not have the ability to train / continue training the model.

[0010] In some possible implementations, before sending the first response message, the method further includes: determining a first model among at least one models based on at least one model corresponding to the first device, parameter information corresponding to the at least one model, and model requirement information, wherein the parameter information is used to continue training the corresponding at least one model.

[0011] For example, at least one model corresponding to the first device includes at least one model stored by the first device; or, at least one model corresponding to the first device includes at least one model stored by another device, and the first device can obtain at least one model stored by the other device.

[0012] For example, the first device can determine a first model from at least one model based on at least one corresponding model, parameter information corresponding to at least one model, model requirement information, and whether the second device has the ability to train and / or continue training the model. For instance, the first device can determine a model that needs and can be further trained as the first model from at least one corresponding model based on model requirement information and the second device's ability to train and / or continue training the model. Whether a model in at least one corresponding model needs further training can be determined by whether the model in at least one corresponding model meets the model requirement information, and whether a model in at least one corresponding model can be further trained can be determined by the parameter information corresponding to the model in at least one corresponding model. Alternatively, the first device can determine a model that needs and can be further trained as the first model based on model requirement information and the second device's lack of ability to train and / or continue training the model.

[0013] For example, after determining the first model, the first device can determine at least one member to train the first model, and determine the model parameters of the trained model based on the model parameters obtained after the at least one member continues to train the model (for example, aggregating the multiple model parameters obtained after multiple members continue to train the model to obtain aggregated model parameters), until the trained model parameters meet the model requirement information, at which point training can be stopped.

[0014] Based on the solution provided in the embodiments of this application, by determining the first model in the corresponding at least one model according to the parameter information and model requirement information of at least one model, it is possible to determine the first model that can meet the model requirement information after continued training for the party requesting the model, or to determine the second model that can meet the model requirement information after continued training of the first model.

[0015] In some possible implementations, before sending the first response message, the method further includes: determining a first model among at least one models based on at least one model corresponding to the first device, first indication information corresponding to the at least one model, and model requirement information, wherein the first indication information is used to indicate whether the first model can continue to be trained.

[0016] For example, at least one model corresponding to the first device includes at least one model stored by the first device; or, at least one model corresponding to the first device includes at least one model stored by another device, and the first device can obtain at least one model stored by the other device.

[0017] For example, the first device can determine a first model from at least one model based on at least one corresponding model, first indication information corresponding to at least one model, model requirement information, and whether the second device has the ability to train and / or continue training the model. For instance, the first device can determine a model that needs and can be further trained as the first model from at least one corresponding model based on model requirement information and the second device's ability to train and / or continue training the model. Whether a model in at least one corresponding model needs to be further trained can be determined by whether the model in at least one corresponding model meets the model requirement information, and whether a model in at least one corresponding model can be further trained can be determined by the first indication information corresponding to the model in at least one corresponding model. Alternatively, the first device can determine a model that needs and can be further trained as the first model from at least one corresponding model based on model requirement information and the second device's lack of ability to train and / or continue training the model.

[0018] For example, after determining the first model, the first device can determine at least one member to train the first model, and determine the model parameters of the trained model based on the model parameters obtained after the at least one member continues to train the model (for example, aggregating the multiple model parameters obtained after multiple members continue to train the model to obtain aggregated model parameters), until the trained model parameters meet the model requirement information, at which point training can be stopped.

[0019] Based on the solution provided in the embodiments of this application, by determining the first model in the corresponding at least one model according to the first indication information and model requirement information corresponding to at least one model, it is possible to determine the first model that can meet the model requirement information after continued training for the party requesting the model, or to determine the second model that can meet the model requirement information after continued training of the first model.

[0020] In some possible implementations, the method also includes: continuing to train the first model to obtain the second model, and the first response message including information about the second model.

[0021] For example, after the first device determines the first model, if the second device does not have the ability to train and / or continue training the model, the first device can continue training the first model to obtain the second model, and the first response information includes information about the second model. For instance, the first device can continue training the model based on the parameter information corresponding to the model that serves as the first model to obtain the second model.

[0022] Based on the solution provided in the embodiments of this application, when the second device does not have the ability to train / continue training the model, the first device can continue to train the first model to obtain the second model. This can avoid the situation where, when the second device does not have the ability to train / continue training the model, the model determined by the first device according to the first request information does not meet the model requirement information and cannot be further trained by the second device to meet the model requirement information.

[0023] In some possible implementations, when the first response information includes information about the first model, the first response message also includes second instruction information, which includes at least one of the following: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained.

[0024] The second device can determine to continue training the first model based on at least one of the second instruction information.

[0025] Based on the solution provided in the embodiments of this application, when the second device has the ability to train / continue training the model, the first device can directly instruct the second device to continue training the first model through the information in the first response message, thereby reducing the probability of the second device making an incorrect judgment on whether to continue training the first model.

[0026] In some possible implementations, before sending the first response message, the method further includes: determining whether to continue training the first model based on whether the second device has the ability to train the model and / or whether the second device has the ability to continue training the model.

[0027] Based on the solution provided in the embodiments of this application, by having the first device determine whether to continue training the first model based on whether the second device has the ability to train / continue training the model, the flexibility of the solution can be improved, making it more reasonable for the first device to continue training the first model.

[0028] In some possible implementations, the first response message may also include at least one of the following: whether the first model can continue to be trained; or whether the first model needs to be continued to be trained.

[0029] Based on the solution provided in the embodiments of this application, the first device can directly indicate whether the first model can continue to be trained, or whether the first model needs to continue to be trained, through the first response message, thereby reducing the probability of the second device making an incorrect judgment on whether the first model needs to continue training.

[0030] In some possible implementations, the model requirement information includes the VAL service identifier and at least one of the following: analysis accuracy; model profile, which includes at least one of the following: model domain, list of permitted vendors, ML model interoperability information, ML model stage, accuracy; analysis identifier; model requirements.

[0031] In a second aspect, a communication method is provided, applied to a second device, comprising: sending a first request message, the first request message including model requirement information; receiving a first response message, the first response message including information of a first model or information of a second model, wherein the first model is used to train a model that satisfies the model requirement information, and the second model is a model that satisfies the model requirement information obtained by training the first model.

[0032] Based on the solution provided in the embodiments of this application, a first model that can meet the model requirements after continued training is sent to the requesting party through a first response message, or a second model that has been trained and can meet the model requirements is sent. It is possible to obtain a new model that can meet the model requirements by continuing to train the existing model. Compared with obtaining a new model that can meet the model requirements by training from scratch, on the one hand, it can reduce the time and / or power consumption of model training and model request, and improve the efficiency of model training and model request; on the other hand, it can improve the utilization rate of the existing model, so that the existing model is used more rationally.

[0033] In some possible implementations, the first request message may also include at least one of the following: whether the second device has the ability to train a model; and whether the second device has the ability to continue training a model.

[0034] In some possible implementations, receiving the first response message includes: receiving a first response message sent by the first device, wherein the first model is determined based on at least one model corresponding to the first device, and parameter information and model requirement information corresponding to the at least one model are determined in the at least one model, and the parameter information is used to continue training the at least one model.

[0035] In some possible implementations, receiving the first response message includes: receiving a first response message sent by the first device, wherein the first model is determined based on at least one model corresponding to the first device, and the first indication information and model requirement information corresponding to the at least one model are determined in the at least one model, and the first indication information is used to continue training the corresponding at least one model.

[0036] In some possible implementations, when the first response information includes information about the first model, the first response message also includes second indication information corresponding to the first model. The second indication information includes at least one of the following: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained. In some possible implementations, the first response message also includes at least one of the following: whether the first model can be continued to be trained; or whether the first model needs to be continued to be trained.

[0037] In some possible implementations, the method also includes: continuing to train the first model based on the second instruction information.

[0038] In some possible implementations, the method further includes: determining whether to continue training the first model based on the information of the first model, the parameter information of the first model, and the model requirement information; and / or, continuing to train the first model based on the information of the first model, the parameter information of the first model, and the model requirement information.

[0039] In some possible implementations, model requirements include a VAL service identifier and at least one of the following: analytical accuracy; a model profile, which includes at least one of the following: the model's domain, a list of permitted vendors, ML model interoperability information, ML model stage, accuracy; analytical identifier; and model requirements.

[0040] Thirdly, a communication method is provided for a receiving end of model storage, comprising: receiving a second request message, the second request message being used to request storage of a first model, the second request message including information of the first model, the second request message also including parameter information of the first model and / or first indication information, the parameter information being used to continue training the first model, the first indication information being used to indicate whether the first model can be continued to be trained; and sending a second response message, the second response message indicating whether the storage of the first model was successful or failed.

[0041] Based on the solution provided in this application embodiment, by sending not only the model to be stored but also the parameter information or first indication information of the model to be stored when requesting model storage, it is possible to send a first model that can meet the model requirements after further training, or a second model that has been trained and can meet the model requirements, to the requesting party through the parameter information or first indication information of the stored model when none of the stored models meet the model requirements. Compared with training from scratch to obtain a new model that can meet the model requirements and sending the trained model to the requesting party, on the one hand, it can reduce the time and / or power consumption of model training and model request, and improve the efficiency of model training and model request; on the other hand, it can improve the utilization rate of existing models, so that existing models are used more rationally.

[0042] In some possible implementations, the parameter information includes a first use case environment, which is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data used to continue training the first model; a first training member requirement, which indicates the requirement for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0043] For example, the parameter information may be carried in the second request message.

[0044] Based on the solution provided in the embodiments of this application, by sending the requirements or data for continuing to train the model that needs to be stored when requesting model storage, it is possible to continue training the model according to the parameter information sent when the model is stored.

[0045] In some possible implementations, the method further includes: sending a third instruction message, the third instruction message including second training data and / or second training members, wherein the second training data and / or second training members are determined by the receiver of the model storage (e.g., an enabling server) based on the first model, the second training data is used to determine the first training data requirements, the second training data is used to train to obtain the first model, and the second training members are used to determine the first training member requirements, the second training members are used to train to obtain the first model.

[0046] Based on the solution provided in the embodiments of this application, the second training data and / or the second training members of the model are determined by the model that needs to be stored. This enables the determination of the requirements or data for continuing to train the model that needs to be stored when requesting model storage, even if no requirements or data for continuing to train the model that needs to be stored are sent. This enables the model to continue training.

[0047] In some possible implementations, the method also includes: determining whether the first model can continue to be trained based on parameter information and / or first indication information.

[0048] In some possible implementations, the first training data is required to include at least one of the following: the data type corresponding to the first model; the data label corresponding to the first model; the data domain corresponding to the first model; the analysis identifier corresponding to the first model; the dataset and data producer requirements corresponding to the first model; the valid time period corresponding to the data corresponding to the first model; and the location corresponding to the data corresponding to the first model.

[0049] In some possible implementations, the first training member is required to include at least one of the following: the member's location; the time when the member is available; the dataset corresponding to the member, which includes the original data and / or the processed data; the data features corresponding to the member; the AI / ML capabilities corresponding to the member, which include at least one of the following capabilities: client capabilities, enabling server capabilities, aggregation capabilities; and the application service identifier corresponding to the member.

[0050] In some possible implementations, the receiver of the model store is a model repository; or, the receiver of the model store is an enabling server.

[0051] In some possible implementations, the second request message may also include the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, the ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0052] Fourthly, a communication method is provided for use at the sending end of model storage, comprising: sending a second request message, the second request message being used to request storage of a first model, the second request message including information of the first model, the second request message also including parameter information of the first model and / or first indication information, the parameter information being used to continue training the first model, and the first indication information being used to indicate whether the first model can be continued to be trained; and receiving a second response message, the second response message indicating whether the storage of the first model was successful or failed.

[0053] Based on the solution provided in this application embodiment, by sending not only the model to be stored but also the parameter information or first indication information of the model to be stored when requesting model storage, it is possible to send a first model that can meet the model requirements after further training, or a second model that has been trained and can meet the model requirements, to the requesting party through the parameter information or first indication information of the stored model when none of the stored models meet the model requirements. Compared with training from scratch to obtain a new model that can meet the model requirements and sending the trained model to the requesting party, on the one hand, it can reduce the time and / or power consumption of model training and model request, and improve the efficiency of model training and model request; on the other hand, it can improve the utilization rate of existing models, so that existing models are used more rationally.

[0054] In some possible implementations, the parameter information includes a first use case environment, which is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data used to continue training the first model; a first training member requirement, which indicates the requirement for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0055] In some possible implementations, the method further includes: sending a third instruction message, the third instruction message including second training data and / or a second training member, wherein the second training data is used to determine the first training data requirements, the second training data is used to train to obtain a first model, and the second training member is used to determine the first training member requirements, the second training member is used to train to obtain the first model.

[0056] In some possible implementations, the first training data is required to include at least one of the following: the data type corresponding to the first model; the data label corresponding to the first model; the data domain corresponding to the first model; the analysis identifier corresponding to the first model; the dataset and data producer requirements corresponding to the first model; the valid time period corresponding to the data corresponding to the first model; and the location corresponding to the data corresponding to the first model.

[0057] In some possible implementations, the first training member is required to include at least one of the following: the member's location; the time when the member is available; the dataset corresponding to the member, which includes the original data and / or the processed data; the data features corresponding to the member; the AI / ML capabilities corresponding to the member, which include at least one of the following capabilities: client capabilities, enabling server capabilities, aggregation capabilities; and the application service identifier corresponding to the member.

[0058] In some possible implementations, the sending end of the model storage is an enabled client; or, the receiving end of the model storage is an enabled server.

[0059] In some possible implementations, the second request message also indicates the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0060] Fifthly, a communication device is provided. This device can be a device where the receiving end of the model request is located, such as a network device, or an enabling server deployed on the network device side, or a chip, circuit, or chip system configured in the network device, or it can be a device where a repository is located. This application does not limit the specific device to these embodiments. For ease of description, the following description will use this device as the first device and the device that jointly implements the model request with this device as the second device.

[0061] The device includes: a transceiver unit, which receives a first request message including model requirement information; and sends a first response message including information of a first model or information of a second model, wherein the first model is used to train a model that meets the model requirement information, and the second model is a model that meets the model requirement information obtained by training the first model.

[0062] In some possible implementations, the transceiver unit is specifically used to: receive a first request message from the second device, the first request message further including at least one of the following: whether the second device has the ability to train a model; whether the second device has the ability to continue training the model.

[0063] In some possible implementations, the apparatus further includes a processing unit, which is used to: determine a first model from at least one model based on at least one model corresponding to the first apparatus, parameter information corresponding to the at least one model, and model requirement information, wherein the parameter information is used to continue training the at least one model.

[0064] In some possible implementations, the apparatus further includes a processing unit, which is configured to: determine a first model among at least one models based on at least one model corresponding to the first apparatus, first indication information corresponding to the at least one model, and model requirement information, wherein the first indication information is used to indicate whether the first model can continue to be trained.

[0065] In some possible implementations, the processing unit is also used to: continue training the first model to obtain the second model, and the first response message includes information about the second model.

[0066] In some possible implementations, when the first response information includes information about the first model, the first response message also includes second instruction information, which includes at least one of the following: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained.

[0067] In some possible implementations, the processing unit is further configured to: determine whether to continue training the first model based on whether the second device has the ability to train the model and / or whether the second device has the ability to continue training the model.

[0068] In some possible implementations, the first response message may also include at least one of the following: whether the first model can continue to be trained; or whether the first model needs to be continued to be trained.

[0069] In some possible implementations, model requirements include a VAL service identifier and at least one of the following: analytical accuracy; a model profile, which includes at least one of the following: the model's domain, a list of permitted vendors, ML model interoperability information, ML model stage, accuracy; analytical identifier; and model requirements.

[0070] In some possible implementations, the processing unit includes a processor.

[0071] In some possible implementations, the transceiver unit includes a transceiver.

[0072] Sixthly, a communication device is provided. This device can be a device where the sending end of the model request is located, such as a terminal device / network device, or an enabling client deployed on the terminal device side, or a network device deploying an enabling server, or it can be a chip, circuit, or chip system configured in the terminal device / network device. The embodiments of this application do not limit this. For ease of description, the following description will use the device as a second device and the device that jointly implements the model request with the device as a first device.

[0073] The device includes a transceiver unit, which is configured to: send a first request message, the first request message including model requirement information; and receive a first response message, the first response message including information of a first model or information of a second model, wherein the first model is used to train a model that meets the model requirement information, and the second model is a model that meets the model requirement information obtained by training the first model.

[0074] In some possible implementations, the first request message may also include at least one of the following: whether the second device has the ability to train a model; and whether the second device has the ability to continue training a model.

[0075] In some possible implementations, the transceiver unit is specifically used to: receive a first response message sent by the first device, wherein the first model is determined based on at least one model corresponding to the first device, the parameter information corresponding to the at least one model and the model requirements are determined in the at least one model, and the parameter information is used to continue training the at least one model.

[0076] In some possible implementations, the transceiver unit is specifically used to: receive a first response message sent by the first device, wherein the first model is determined based on at least one model corresponding to the first device, and the first indication information and model requirements corresponding to the at least one model are determined in the at least one model, and the first indication information is used to continue training the corresponding at least one model.

[0077] In some possible implementations, when the first response information includes information about the first model, the first response message also includes second indication information corresponding to the first model. The second indication information includes at least one of the following: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained.

[0078] In some possible implementations, the second device has the ability to train the model and / or to continue training the model.

[0079] In some possible implementations, the first response message may also include at least one of the following: whether the first model can continue to be trained; or whether the first model needs to be continued to be trained.

[0080] In some possible implementations, the device further includes a processing unit for: continuing to train the first model according to the second instruction information.

[0081] In some possible implementations, the processing unit is used to: determine whether to continue training the first model based on the information of the first model, the parameter information of the first model, and the model requirement information; and / or, continue training the first model based on the information of the first model, the parameter information of the first model, and the model requirement information.

[0082] In some possible implementations, model requirements include a VAL service identifier and at least one of the following: analytical accuracy; a model profile, which includes at least one of the following: the model's domain, a list of permitted vendors, ML model interoperability information, ML model stage, accuracy; analytical identifier; and model requirements.

[0083] In some possible implementations, the processing unit includes a processor.

[0084] In some possible implementations, the transceiver unit includes a transceiver.

[0085] In a seventh aspect, an apparatus is provided, which may be a device where the receiving end of the model storage is located, such as a network device, or an enabling server deployed on the network device side, or a chip, circuit or chip system configured in the network device, or a device where the storage repository is located. The embodiments of this application do not limit this.

[0086] The device includes a transceiver unit, which is configured to: receive a second request message, the second request message being used to request storage of a first model, the second request message including information about the first model, the second request message also including parameter information of the first model and / or first indication information, the parameter information being used to continue training the first model, and the first indication information being used to indicate whether the first model can be continued to be trained; and send a second response message, the second response message indicating whether the storage of the first model was successful or failed.

[0087] In some possible implementations, the parameter information includes a first use case environment, which is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data used to continue training the first model; a first training member requirement, which indicates the requirement for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0088] In some possible implementations, the method further includes: sending a third instruction message, the third instruction message including second training data and / or second training members, wherein the second training data and / or second training members are determined by the receiver of the model storage (e.g., an enabling server) based on the first model, the second training data is used to determine the first training data requirements, the second training data is used to train to obtain the first model, and the second training members are used to determine the first training member requirements, the second training members are used to train to obtain the first model.

[0089] In some possible implementations, the apparatus further includes a processing unit for determining whether the first model can continue to be trained based on parameter information and / or first indication information.

[0090] In some possible implementations, the first training data is required to include at least one of the following: the data type corresponding to the first model; the data label corresponding to the first model; the data domain corresponding to the first model; the analysis identifier corresponding to the first model; the dataset and data producer requirements corresponding to the first model; the valid time period corresponding to the data corresponding to the first model; and the location corresponding to the data corresponding to the first model.

[0091] In some possible implementations, the first training member is required to include at least one of the following: the member's location; the time when the member is available; the dataset corresponding to the member, which includes the original data and / or the processed data; the data features corresponding to the member; the AI / ML capabilities corresponding to the member, which include at least one of the following capabilities: client capabilities, enabling server capabilities, aggregation capabilities; and the application service identifier corresponding to the member.

[0092] In some possible implementations, the receiving end of the model storage is a model repository; or, the receiving end of the model storage is an enabling server.

[0093] In some possible implementations, the second request message also indicates the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0094] In some possible implementations, the processing unit includes a processor.

[0095] In some possible implementations, the transceiver unit includes a transceiver.

[0096] Eighthly, an apparatus is provided, which may be a device where the sending end of the model storage is located, such as a terminal device / network device, or an enabling client deployed on the terminal device side, or a network device that deploys an enabling server, or a chip, circuit or chip system configured in the terminal device / terminal device. The embodiments of this application do not limit this.

[0097] The device includes a transceiver unit, which is configured to: send a second request message, the second request message including information about a first model, the second request message also including parameter information of the first model and / or first indication information, the parameter information being used to continue training the first model, and the first indication information being used to indicate whether the first model can be continued to be trained; and receive a second response message, the second response message indicating whether the first model was successfully or unsuccessfully stored.

[0098] In some possible implementations, the parameter information includes a first use case environment, which is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data used to continue training the first model; a first training member requirement, which indicates the requirement for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0099] In some possible implementations, the transceiver unit is further configured to: send a third instruction message, the third instruction message including second training data and / or second training members, wherein the second training data is used to determine the first training data requirements, the second training data is used to train to obtain the first model, and the second training members are used to determine the first training member requirements, the second training members are used to train to obtain the first model.

[0100] In some possible implementations, the first training data is required to include at least one of the following: the data type corresponding to the first model; the data label corresponding to the first model; the data domain corresponding to the first model; the analysis identifier corresponding to the first model; the dataset and data producer requirements corresponding to the first model; the valid time period corresponding to the data corresponding to the first model; and the location corresponding to the data corresponding to the first model.

[0101] In some possible implementations, the first training member is required to include at least one of the following: the member's location; the time when the member is available; the dataset corresponding to the member, which includes the original data and / or the processed data; the data features corresponding to the member; the AI / ML capabilities corresponding to the member, which include at least one of the following capabilities: client capabilities, enabling server capabilities, aggregation capabilities; and the application service identifier corresponding to the member.

[0102] In some possible implementations, the sending end of the model storage is an enabled client; or, the receiving end of the model storage is an enabled server.

[0103] In some possible implementations, the second request message also indicates the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0104] In some possible implementations, the processing unit includes a processor.

[0105] In some possible implementations, the transceiver unit includes a transceiver.

[0106] Ninth aspect, a communication device is provided, the device comprising: a processor for executing computer instructions to cause the device to perform the methods of the first aspect and any of its possible implementations to the methods of the fourth aspect and any of its possible implementations.

[0107] In some possible implementations, the device also includes a memory.

[0108] In some possible implementations, the device also includes a communication interface coupled to the processor, which is used for inputting and / or outputting information.

[0109] In a tenth aspect, a computer program product is provided, which, when executed by a communication device, implements the methods of the first aspect and any of its possible implementations to the fourth aspect and any of its possible implementations.

[0110] In an eleventh aspect, a computer-readable storage medium is provided, wherein a computer program or instructions are stored therein, which, when executed by a communication device, implement the methods of the first aspect and any of its possible implementations to the fourth aspect and any of its possible implementations.

[0111] As examples, these computer-readable storage devices include, but are not limited to, one or more of the following: read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), flash memory, electrically EPROM (EEPROM), and hard drive.

[0112] In some possible implementations, the aforementioned storage medium may specifically be a non-volatile storage medium.

[0113] In a twelfth aspect, a chip (or chip system) is provided, including at least one processor for running a computer program that causes a device having the chip mounted to perform the methods of the first aspect and any of its possible implementations to the fourth aspect and any of its possible implementations.

[0114] The chip may include an output circuit or interface for transmitting information or data, and an input circuit or interface for receiving information or data.

[0115] In a thirteenth aspect, a communication system is provided, comprising: an enabling server and an enabling client, wherein the enabling server is configured to execute the methods of the first aspect and any of its possible implementations, and the enabling client is configured to execute the methods of the second aspect and any of its possible implementations; or, the enabling server is configured to execute the methods of the third aspect and any of its possible implementations, and the enabling client is configured to execute the methods of the fourth aspect and any of its possible implementations.

[0116] Fourteenthly, a communication system is provided, comprising: a repository and an enabling client, the repository being used to execute the methods of the third aspect above and any of its possible implementations, and the enabling client being used to execute the methods of the fourth aspect above and any of its possible implementations.

[0117] The beneficial effects of the above aspects and any of their possible implementations can be referenced to the beneficial effects of the first aspect and any of its possible implementations, or the beneficial effects of the third aspect and any of its possible implementations. Further details will not be provided here. Attached Figure Description

[0118] Figure 1 This is a schematic diagram of a wireless communication system applicable to embodiments of this application.

[0119] Figure 2 This is a schematic diagram of a possible application framework in a communication system.

[0120] Figure 3 This is a schematic diagram of a possible application framework in a communication system.

[0121] Figure 4 This is a schematic diagram of a system architecture applicable to the communication method of this application.

[0122] Figure 5 This is a schematic diagram of an artificial intelligence / machine learning architecture applicable to embodiments of this application.

[0123] Figure 6 This is a schematic diagram of a model storage method applicable to an embodiment of this application.

[0124] Figure 7 This is a schematic diagram of a model request applicable to an embodiment of this application.

[0125] Figure 8 This is a schematic diagram of a communication method 800 applicable to an embodiment of this application.

[0126] Figure 9 This is a schematic diagram of a communication method 900 applicable to an embodiment of this application.

[0127] Figure 10 This is a schematic diagram of a communication method 1000 applicable to an embodiment of this application.

[0128] Figure 11 This is a schematic diagram of a communication method 2000 applicable to an embodiment of this application.

[0129] Figure 12 This is a schematic diagram of a communication method 3000 applicable to an embodiment of this application.

[0130] Figure 13 This is a schematic block diagram of a communication device provided in an embodiment of this application.

[0131] Figure 14 This is another schematic block diagram of the communication device provided in the embodiments of this application. Detailed Implementation

[0132] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0133] Before introducing the scheme of this application, the following points should be noted.

[0134] (1) The terminology used in the following embodiments is for the purpose of describing specific embodiments only and is not intended to be a limitation of this application. As used in the specification and appended claims of this application, "a plurality of" or "multiple" means two or more; the singular expressions "a," "an," "the," "the," "the," "the," and "this" are intended to also include expressions such as "one or more," unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, "at least one," "at least one," and "one or more" refer to one, two, or more. "And / or" describes the relationship between related objects, indicating that three relationships may exist, for example, A and / or B, which can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. In the textual description of this application, the character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, and c can represent: a, or b, or c, or a and b, or a and c, or b and c, or a, b, and c. Here, a, b, and c can be a single term or multiple terms.

[0135] (2) In the embodiments of this application, the ordinal numbers such as "first," "second," "#1," "#2," "#A," and "#B" are used to distinguish multiple objects and are not used to limit the size, content, order, timing, priority, or importance of the multiple objects. For example, the first instruction information and the second instruction information can be the same information or different information, and such names do not indicate that the content, size, application scenario, sending end / receiving end, priority, or importance of the two messages are different. In addition, the step numbers in the various embodiments described in this application are only for distinguishing different steps, and unless otherwise stated, the step numbers are not used to limit the order between steps.

[0136] (3) In this application, "send" and "receive" indicate the direction of signal transmission, and "transmission" can include at least one of sending and / or receiving. For example, "send information to XX" can be understood as the destination of the information being XX, which can include sending directly via the air interface or sending indirectly via the air interface from other units or modules. "Receive information from YY" can be understood as the source of the information being YY, which can include receiving directly from YY via the air interface or receiving indirectly from YY via the air interface from other units or modules. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface. In other words, sending and receiving can be performed between devices, such as between network devices and terminal devices, or within a device, such as sending or receiving between components, modules, chips, software modules, or hardware modules within a device via a bus, wiring, or interface.

[0137] (4) In this application, the terms “comprising” and “having” and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product or device; or, message #A includes B, which may be equivalent to message #A carrying the entire contents of B.

[0138] (5) In this application, "instruction" can include direct instruction and indirect instruction. When describing a certain instruction information A, it can include either direct instruction A or indirect instruction A. Unless otherwise stated, this does not necessarily mean that the instruction information carries A. Direct instruction information A means that information A is included; implicit instruction information A means that information A is indicated through the correspondence between information A and information B and the direct instruction information B. The correspondence between information A and information B can be predefined, pre-stored, pre-burned, or pre-configured.

[0139] (6) References to “one embodiment” or “some embodiments” as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, phrases such as “in some possible implementations” or “in other possible implementations” appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean “one or more, but not all, embodiments”, unless otherwise specifically emphasized. The terms “comprising,” “including,” “having,” and variations thereof mean “including, but not limited to,” unless otherwise specifically emphasized.

[0140] (7) In the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terms and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0141] (8) In this application, the words “exemplary,” “for example,” etc., are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as an “example” in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word “example” is intended to present a concept in a concrete manner. In the embodiments of this application, “of,” “corresponding, relevant,” and “corresponding” may sometimes be used interchangeably, and it should be noted that their intended meanings are consistent unless their distinction is emphasized.

[0142] (9) In this application, descriptions such as “when…” and “under the circumstances of…” all refer to the device making corresponding processing under certain objective circumstances. They are not time limits, nor do they require the device to make judgments during implementation, nor do they imply any other limitations.

[0143] The technical solutions of this application can be applied to various communication systems, such as: Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, 5G systems, or New Radio (NR) and future communication systems. The technical solutions provided in this application can also be applied to device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), and Internet of Things (IoT) communication systems.

[0144] Furthermore, the embodiments of this application are applicable to both homogeneous and heterogeneous network scenarios, and there are no restrictions on the transmission points. They can be applied to systems such as multi-point collaborative transmission between macro base stations, micro base stations, and macro base stations. The embodiments of this application are applicable to both low-frequency and high-frequency scenarios, including terahertz and optical communications.

[0145] In a communication system, a device can send signals to or receive signals from another device. These signals may include reference signals, information, signaling, or data. In this application, "device" can be replaced by an entity, network entity, communication equipment, communication module, node, or communication node.

[0146] Figure 1 This is a schematic diagram of a wireless communication system 100 applicable to embodiments of this application. For example... Figure 1 As shown, the wireless communication system includes a wireless access network 100. The wireless access network 100 can be a next-generation (e.g., future communication networks or higher) wireless access network, or a traditional (e.g., 5G, 4G, 3G, or 2G) wireless access network. One or more terminal devices (120a-120j, collectively referred to as 120) can be interconnected or connected to one or more network devices (110a, 110b, collectively referred to as 110) within the wireless access network 100. Network elements in the wireless communication system are connected via interfaces (e.g., NG, Xn) or over-the-air interfaces.

[0147] Figure 1 This is just an illustration; the wireless communication system may also include other devices, such as core network (CN) equipment, wireless relay equipment, and / or wireless backhaul equipment. Figure 1 It is not shown in the middle.

[0148] In practical applications, this wireless communication system can include multiple network devices and multiple terminal devices simultaneously, without limitation. A network device can serve one or more terminal devices simultaneously. A terminal device can also access one or more network devices simultaneously. The embodiments of this application do not limit the number of terminal devices and network devices included in the wireless communication system.

[0149] The communication system described above for use in the embodiments of this application is merely an example. The communication system applicable to the embodiments of this application is not limited to this. Any communication system capable of implementing the functions of the above-described devices is applicable to the embodiments of this application.

[0150] In this application, the terminal device can refer to 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. The terminal device can also be a cellular phone, cordless phone, session initiation protocol (SIP) phone, wireless local loop (WLL) station, personal digital assistant (PDA), handheld device with wireless communication capabilities, computing device, or other processing device connected to a wireless modem, vehicle-mounted device, wearable device, terminal device in a 5G network, or terminal device in a future public land mobile network (PLMN), etc. This application does not limit the scope of the terminal device to these specific types.

[0151] Wearable devices, also known as wearable smart devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices; they achieve powerful functions through software support, data interaction, and cloud interaction. Broadly defined, wearable smart devices include those with comprehensive functions, large sizes, and the ability to perform complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses. They also include devices focused on a specific application function that require the use of other devices, such as smart bracelets and smart jewelry for vital sign monitoring.

[0152] Furthermore, terminal devices can also be terminal devices in Internet of Things (IoT) systems. IoT is an important component of future information technology development, and its main technical feature is connecting objects to networks through communication technologies, thereby realizing an intelligent network that enables human-machine interconnection and machine-to-machine interconnection.

[0153] It should be understood that this application does not limit the specific form of the terminal device.

[0154] Network devices can be devices within a wireless network. For example, a network device can be a device deployed in a wireless network to provide wireless communication capabilities for terminal devices. For instance, a network device can be a radio access network (RAN) node that connects terminal devices to the wireless network. The RAN can be connected to the core network (e.g., the core network of Long Term Evolution (LTE) or the core network of 5G, etc.).

[0155] The network devices in this application embodiment can be access network devices, including but not limited to: various base stations, such as next-generation node B (gNodeB, gNB), evolved node B (eNB), or base station equipment in future evolved communication systems; they can also be enabling servers, wearable devices, vehicle-mounted devices, wireless relay nodes, wireless backhaul nodes, transmission points (TP), or transmission and reception points (TRP), etc.; they can also be one or a group of antenna panels (including multiple antenna panels) of a base station; or they can be network nodes constituting a base station, such as a bandwidth-based unit (BBU) or a distributed unit (DU), etc. The base station can be a macro base station, micro base station, pico base station, small cell, relay station, or balloon station, etc.

[0156] The network device in this application embodiment can also be a core network device, including but not limited to: access and mobility management function (AMF) network element, session management function (SMF) network element, user plane function (UPF) network element, policy control function (PCF) network element, or unified data management (UDM) network element, etc.

[0157] Application layer network elements refer to network devices in a computer network that are responsible for processing application layer protocols, including but not limited to: data collection application function (DCAF) network elements, provisioning application function (PAF) network elements, event consumer application function (ECAF) network elements, etc.

[0158] It is understood that all or part of the functions of the network device or terminal device in this application can also be implemented by software functions running on hardware, or by virtualization functions instantiated on a platform (such as a cloud platform).

[0159] In some deployments, the network devices mentioned in the embodiments of this application may be devices including centralized units (CU), DU, or devices including CU and DU, or devices with control plane CU nodes (central unit-control plane (CU-CP)) and user plane CU nodes (central unit-user plane (CU-UP)) and DU nodes. For example, the network devices may include gNB-CU-CP, gNB-CU-UP, and gNB-DU.

[0160] In some deployments, multiple RAN nodes collaborate to assist terminals in achieving wireless access, with different RAN nodes each implementing some of the base station's functions. For example, RAN nodes can be CUs, DUs, CU-CPs, CU-UPs, or radio units (RUs). CUs and DUs can be configured separately or included in the same network element, such as a BBU. RUs can be included in radio equipment or radio units, such as remote radio units (RRUs), active antenna systems (AAUs), or remote radio heads (RRHs).

[0161] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. Any of the units among CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0162] To support machine learning capabilities in wireless communication systems, artificial intelligence (AI) nodes may also be introduced.

[0163] Optionally, the communication system also includes at least one AI node.

[0164] Optionally, the AI ​​node may be deployed on one or more of the following: network devices, terminal devices, core network, or positioning devices; alternatively, the AI ​​node may be deployed independently, such as in a location other than any of the aforementioned devices. The AI ​​node may communicate with other devices in the communication system, which may be, for example, one or more of the following: network devices, terminal devices, core network elements, or sensing devices.

[0165] Optionally, the AI ​​node is used to perform AI-related operations. As an example, AI-related operations may include one or more of the following: model failure testing, model performance testing, model training testing, or data acquisition.

[0166] For example, a network device can forward AI model-related data reported by a terminal device to an AI node, which then performs AI-related operations. As another example, a network device or terminal device can forward AI model-related data to an AI node, which then performs AI-related operations. As yet another example, an AI node can send one or more of the outputs of AI-related operations, such as a trained neural network model, model evaluation, or test results, to a network device and / or a terminal device. For example, an AI node can directly send the outputs of AI-related operations to a network device and a terminal device. As yet another example, an AI node can send the outputs of AI-related operations to a terminal device through a network device. As yet another example, an AI node can send the outputs of AI-related operations to a network device through a terminal device.

[0167] It is understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, these nodes can be divided based on function, such as different AI nodes being responsible for different functions.

[0168] It can also be understood that AI nodes can be independent devices, or they can be integrated into the same device to achieve different functions. Alternatively, they can be network elements in hardware devices, software functions running on dedicated hardware, or virtualization functions instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of the aforementioned AI nodes.

[0169] For example, an AI node can be an AI network element or an AI module.

[0170] It should be understood that Figure 1 This is an example illustration of a communication system applicable to the embodiments of this application. It is a simplified diagram for ease of understanding only. The communication system described above may also include other network devices or other terminal devices. Figure 1 The diagram is not shown. The communication system used in the embodiments of this application is not limited to this.

[0171] It should also be understood that Figure 1 These are merely illustrative application scenarios for embodiments of this application, and this application does not limit the scenarios in which the method is applied. This application can be applied to communication between network devices, communication between network devices and terminal devices, communication between terminal devices, etc., and the embodiments of this application do not limit this.

[0172] Figure 2 This is a schematic diagram of a possible application framework in a communication system. For example... Figure 2 As shown, network elements in a communication system are connected via interfaces (e.g., next-generation (NG) interfaces, Xn interfaces) or air interfaces. These network element nodes, such as core network equipment, access network nodes or equipment (RAN nodes or equipment), terminals, or one or more devices in operation administration and maintenance (OAM), are equipped with one or more AI modules (for clarity, ...). Figure 2 (Only one is shown in the image). The access network node can be a single RAN node or can include multiple RAN nodes, such as CU and DU. The CU and / or DU can also be configured with one or more AI modules. Optionally, the CU can also be split into CU-CP and CU-UP. One or more AI models are configured in CU-CP and / or CU-UP.

[0173] The AI ​​module is used to implement corresponding AI functions. AI modules deployed in different network elements can be the same or different. Depending on the parameter configuration, the AI ​​module can implement different functions. The AI ​​module model can be configured based on one or more of the following parameters: structural parameters (e.g., at least one of the following: number of neural network layers, neural network width, inter-layer connections, neuron weights, neuron activation function, or bias in the activation function), input parameters (e.g., type and / or dimension of input parameters), or output parameters (e.g., type and / or dimension of output parameters). The bias in the activation function can also be referred to as the neural network bias.

[0174] An AI module can have one or more models. A model can infer an output that includes one or more parameters. The learning, training, or inference processes of different models can be deployed on different nodes or devices, or they can be deployed on the same node or device.

[0175] Figure 3 This is a schematic diagram of a possible application framework in a communication system. For example... Figure 3As shown, the communication system includes a RAN intelligent controller (RIC). For example, the RIC can be... Figure 2 The AI ​​module shown is used to implement AI-related functions. The RIC includes near-real-time RIC (near-RT RIC) and non-real-time RIC (non-RT RIC). Non-real-time RIC primarily processes non-real-time information, such as data that is not sensitive to latency, with latency in the order of seconds. Real-time RIC primarily processes near-real-time information, such as data that is relatively sensitive to latency, with latency in the order of tens of milliseconds.

[0176] The near real-time RIC is used for model training and inference. For example, it is used to train an AI model and then use that AI model for inference. The near real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data. Optionally, the near real-time RIC can deliver the inference results to the RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the near real-time RIC delivers the inference results to the DU, and the DU then sends the inference results to the RU.

[0177] The non-real-time RIC is also used for model training and inference. For example, it can be used to train an AI model and then use that model for inference. The non-real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data, and the inference results can be delivered to RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the non-real-time RIC delivers the inference results to the DU, and the DU then sends the inference results to the RU.

[0178] The near real-time RIC and non-real-time RIC can also be set up as separate network elements. Optionally, the near real-time RIC and non-real-time RIC can also be part of other devices. For example, the near real-time RIC can be set in the RAN node (e.g., in CU, DU), while the non-real-time RIC can be set in the OAM, cloud enabling server, core network device, or other network device.

[0179] Figure 4This is a schematic diagram of a system architecture applicable to the communication method of this application. In this system architecture, the service enabler architecture layer (SEAL) provides communication services for application clients and application enabler servers. As an example, SEAL may include a service enabler architecture layer data delivery (SEALDD) enabler server and a SEALDD client. The SEALDD client and application client can be part of a terminal device, running as software or system components on the terminal device, which can communicate with a radio access network (RAN). The SEALDD enabler server can be deployed as a standalone or integrated enabler server between the user plane function (UPF) and the application enabler server. In practical deployments, multiple SEALDD enabler servers can be deployed in a distributed manner, depending on the deployment of the UPF and the application enabler servers.

[0180] Figure 5 A schematic diagram of an artificial intelligence / machine learning (AI / ML) architecture 500 applicable to embodiments of this application is shown. In the vertical application layer (VAL), the VAL client communicates with the VAL enabling server via the VAL-UU interface, which supports both unicast and multicast transmission modes. In the service enabler architecture layer for verticals (SEAL), the AI / ML enabling client communicates with the AI / ML enabling server via the AI / ML-UU interface. The AI / ML enabling client can communicate with the VAL client via the AI / ML-C interface; the VAL enabling server can communicate with the AI / ML enabling server via the AIML-S interface. AI / ML enabling servers can communicate with 3GPP networks through network interfaces provided by 3GPP, including the N33 interface provided by the network exposure function (NEF) of the core network element and the service-oriented interface provided by OAM.

[0181] Vertical industry application layers can support the following SEAL services:

[0182] • Location management function

[0183] • Group management function

[0184] Configuration management function

[0185] Key management function

[0186] • Network resource management function

[0187] • Network slice capability management function

[0188] VAL enables the server to call one or more of the capabilities in SEAL above to enable the use of 3GPP network capabilities.

[0189] exist Figure 5 In the AI / ML architecture shown, AI / ML enabling clients can be deployed on the UE. These clients have AI / ML functions, including local model training and model inference. On the network side, AI / ML enabling servers can be deployed. These servers have AI / ML functions and can also select AI / ML members, assign federated learning tasks, and distribute the tasks to AI / ML clients.

[0190] Optionally, a central enabling server can also be deployed on the network side (for example, the central enabling server can be named AI / ML central server, or other names), which can be used to store trained AI / ML models.

[0191] The embodiments shown below are for ease of understanding and illustration only, and the method provided by the embodiments of this application is described in detail using the interaction between network devices and terminal devices as an example.

[0192] To facilitate understanding of the embodiments of this application, the terminology involved in the embodiments of this application will be briefly introduced below.

[0193] (1) AI / ML model

[0194] AI / ML models are algorithms or computer programs that implement AI functions. They represent the mapping relationship between the model's input and output; in other words, they are function models that map inputs of a certain dimension to outputs of a certain dimension. The parameters of these function models can be obtained through machine learning training. For example, f(x) = mx 2 +m' is a quadratic function model, which can be viewed as an AI / ML model. m' and m are the parameters of this AI / ML model, and m' and m can be obtained through machine learning training.

[0195] It is understood that AI / ML models can be implemented in hardware circuits, software, or a combination of both; there are no restrictions. Non-restrictive examples of software include: program code, program, subroutine, instruction, instruction set, code, code segment, software module, application program, or software application, etc.

[0196] The AI ​​module is a module with machine learning computing capabilities. In wireless communication systems, the AI ​​module can be located in operations, administration and maintenance (OAM), or in a gNB (e.g., in a separate architecture, it can be located in the CU / DU), in terminal equipment, or as a standalone network element entity, the RAN intelligent controller (RIC). The main function of the AI ​​module in a wireless communication system is to perform a series of AI calculations based on input data (e.g., network operation data provided by access network equipment, or network operation data monitored by OAM, such as network load, channel quality, etc., or user plane data transmission information provided by the core network, or data collected by communication equipment), including model building, training approximation, and reinforcement learning. Currently, the trained models provided by the AI ​​module have predictive capabilities for changes in the RAN side network and can be used for load prediction, terminal equipment path prediction, CSI prediction, optimal beam prediction, and positioning prediction. Furthermore, the AI ​​module can also perform policy reasoning from the perspectives of network energy saving and mobility optimization based on the predicted RAN network performance results of the trained models, to obtain reasonable and efficient energy-saving strategies and mobility optimization strategies. When the AI ​​model resides in the CU, and the CU's control plane and user plane are separated, the CP can be responsible for receiving the AI ​​model and subsequent AI inference and policy generation functions. When the CU-CP is further divided into CU-CP1 and CU-CP2, CU-CP1 can be responsible for receiving the AI ​​model and subsequent AI model inference functions, and generating specific interactive signaling, which is then sent by CU-CP2. When the AI ​​module is located in the OAM, its communication with the RAN-side gNB can reuse the current northbound interface. When the AI ​​module is located in the gNB or CU, the current F1, Xn, Uu, etc. interfaces can be reused; when the AI ​​module becomes an independent network entity, a new communication link needs to be established with the OAM and RAN sides, for example, based on a wired link or a wireless link.

[0197] (2) Model storage

[0198] Model storage refers to the act of users storing models in a model repository.

[0199] For example, such as Figure 6 As shown, Figure 6 This is a schematic diagram of a model storage method applicable to an embodiment of this application. A user of the model repository can send an AI / ML model information storage request #A to the model repository. The information included in the AI / ML model information storage request #A can be found in Tables 1 and 2.

[0200] AI / ML model information storage request #A may include the model and model-related information. For example, AI / ML model information storage request #A may include: VAL service identity (VAL service ID); AI / ML model information storage request #A may also include one or more of the following: analytics ID, AI / ML model, AI / ML model address, AI / ML model profile, etc. The AI / ML model profile may include one or more of the following: domain, list of allowed vendors, AI / ML model interoperability information, AI / ML model phase, AI / ML model storage and discovery requirements (e.g., AI / ML model storage and discovery requirements may include storage duration requirements, security / access requirements, etc.).

[0201] Table 1: ML model information storage request

[0202]

[0203] Table 2: ML model information

[0204]

[0205]

[0206] The model repository can send AI / ML model information storage responses to model repository users, indicating whether the model storage was successful or failed. The information included in the AI / ML model information storage response can be found in Table 3. Upon successful model storage, an AI / ML model ID can be configured for the successfully stored model and displayed to the model repository user.

[0207] Table 3: ML model information storage response

[0208]

[0209] (3) Model Request

[0210] A model request refers to a user of a model repository requesting a model from the model repository. For example, such as... Figure 7 As shown, Figure 7 This is a schematic diagram of a model request applicable to an embodiment of this application. A model repository user can send an AI / ML model information discovery request #A to the model repository. The information included in the AI / ML model information discovery request #A can be found in Table 4. The AI / ML model information discovery request #A may carry one or more of the following conditions for model selection: AI / ML modelID, domain, accuracy requirements, analysis ID, list of allowed vendors, AI / ML model interoperability information, and base model ID.

[0211] Table 4: ML model information discovery request

[0212]

[0213]

[0214] The model repository can send an AI / ML model information discovery response #A to its users. The information included in the AI / ML model information discovery response #A can be found in Table 5. The response may contain an AI / ML model that meets the model selection criteria, or the address of the AI / ML model (the address information storing the AI / ML model), etc.

[0215] Table 5: ML model information discovery response

[0216]

[0217]

[0218] When a user requests a model from the model repository, if the model repository or other devices have an AI / ML model that meets the model selection criteria, the model repository can send an AI / ML model information discovery response #A to the user, providing the user with the AI / ML model that meets the selection criteria. The user can then obtain the AI / ML model that meets the selection criteria through the AI / ML model information discovery response #A. If the model repository or other devices do not have an AI / ML model that meets the selection criteria, a device with model training capabilities can train a new model from scratch. This new model meets the model selection criteria, and the user can then obtain this model that meets the selection criteria.

[0219] When the model repository does not have an AI / ML model that meets the model selection criteria, training a new model from scratch requires a long training time and is inefficient.

[0220] This application provides a communication method, apparatus, and system that can reduce model request time and improve model request efficiency.

[0221] The communication method provided in this application embodiment can be applied between an enabling client (or the terminal device where the enabling client is located) and an enabling server. This communication method can also be applied between two different enabling servers. For ease of description, the following example illustrates possible implementations of the communication method 1000 provided in this application embodiment, using the application of this method between a second device (e.g., an enabling client) and a first device (e.g., an enabling server or repository).

[0222] like Figure 8 As shown, Figure 8 This is a schematic diagram of a communication method 800 applicable to an embodiment of this application. Method 800 may include:

[0223] S810: The first device receives a first request message from the second device, the first request message including model requirement information.

[0224] Specifically, the second device sends a first request message to the first device, and the first device receives the first request message sent by the second device. The first request message includes model requirement information and is used to obtain model information corresponding to the model requirement information requested by the first device.

[0225] For example, an ML model information discovery request could be one possible implementation of the first request message.

[0226] For example, the ML model information discovery response could be one possible implementation of the first response message.

[0227] The first model can be a model that satisfies some of the parameters in the model requirements information. For example, the model requirements information may include one or more of the following parameters: use case context, service ID, VAL server ID, input characteristics, output characteristics, accuracy, response time, domain, required accuracy performance, analytics ID, list of allowed vendors, required ML model interoperability, and ML model phase.

[0228] The first model is a model that satisfies one or more parameters in the model requirement information; or, the first model is a model that does not satisfy some or all of the parameters in the model requirement requested by the first device; or, the first model satisfies the following: after further training of the first model, the second model obtained by further training satisfies all the parameters in the model requirement information requested by the first device.

[0229] In some possible implementations, the first request message may also include at least one of the following: whether the second device has the ability to train a model; and whether the second device has the ability to continue training a model.

[0230] Specifically, when the second device sends a first request message requesting a model to the first device, the request message may carry whether the second device has the ability to train a model and / or whether the second device has the ability to continue training the model.

[0231] It is understood that whether the second device has the ability to train a model and / or whether the second device has the ability to continue training a model can be indicated through the first request message or through other messages besides the first request message. The aforementioned first request message, which includes whether the second device has the ability to train a model and / or whether the second device has the ability to continue training a model, is only an illustrative example. The embodiments of this application can also be applied to instruct the first device in other ways whether the second device has the ability to train a model and / or whether the second device has the ability to continue training a model.

[0232] When the second device has the capability to train a model and / or the capability to continue training a model, the second device can convey this capability through a first request message. Furthermore, when the first device determines a model based on the first request message, it can identify a model that needs further training by the second device to meet model requirement information. When the second device does not have the capability to train a model and / or the capability to continue training a model, the second device can convey this capability through a first request message. Furthermore, when the first device determines a model based on the first request message, it can identify a model that needs further training by a device other than the second device to meet model requirement information, or identify a model that already meets model requirement information. Alternatively, if the first request message does not specify whether the second device has the capability to train a model and / or the capability to continue training a model, it can be assumed that the second device has the capability to train a model and / or the capability to continue training a model (or it can be assumed that the second device does not have the capability to train a model and / or the capability to continue training a model; the specific default can be determined through prior agreement or negotiation between the first and second devices).

[0233] The following sections will describe these two different scenarios (whether the second device has the ability to train / continue training the model).

[0234] In some possible implementations, method 800 may also include:

[0235] S815: Determine whether to use the first model or the second model.

[0236] Specifically, the first device can determine the first model or the second model based on whether the second device has the ability to train the model and / or whether the second device has the ability to continue training the model.

[0237] In some possible implementations, S815 includes S815a.

[0238] S815a: The first device determines a first model from at least one model based on at least one model corresponding to the first device, parameter information corresponding to at least one model, and model requirement information, and the parameter information is used to continue training the corresponding at least one model.

[0239] For example, the first device determines the first model based on the model requirement information in the first request message sent by the second device and the parameter information of at least one model corresponding to the first device, wherein the first model satisfies one or more of the parameters in the model requirement information.

[0240] The parameter information may include a first use case environment, which is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirements for training data used to continue training the first model; a first training member requirement, which indicates the requirements for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0241] Model requirements information may include one or more of the following parameters: use case context, serviceID, VAL server ID, input characteristics, output characteristics, accuracy, response time, domain, required accuracy performance, analytics ID, list of allowed vendors, required MLmodel interoperability, and ML Model phase.

[0242] Optionally, in some other possible implementations, S815 includes S815b.

[0243] S815b: The first device determines a first model from at least one model based on at least one model corresponding to the first device, first indication information corresponding to at least one model, and model requirement information, wherein the first indication information is used to indicate whether the first model can continue to be trained.

[0244] Specifically, the first device can determine the first model based on the model requirement information, parameter information corresponding to at least one model corresponding to the first device, and first indication information corresponding to at least one model corresponding to the first device in the first request message sent by the second device.

[0245] For example, if there is a model in at least one model corresponding to the first device that does not meet the model requirement information, can be trained further, and can meet the model requirement information after training, the first device can identify that model as the first model.

[0246] For example, at least one model corresponding to the first device includes at least one model stored by the first device; or, at least one model corresponding to the first device includes at least one model stored by another device, and the first device can obtain at least one model stored by the other device.

[0247] In some possible implementations, in S815a and / or S815b, when the first device determines the first model, it can also determine the first model based on whether the second device has the ability to train the model or whether the second device has the ability to continue training the model.

[0248] For example, the first device can determine a first model from at least one model based on at least one corresponding model, first indication information corresponding to at least one model, model requirement information, and whether the second device has the ability to train and / or continue training the model. For instance, the first device can determine a model that needs and can be further trained as the first model from at least one corresponding model based on model requirement information and the second device's ability to train and / or continue training the model. Whether a model in at least one corresponding model needs further training can be determined by whether the model in at least one corresponding model meets the model requirement information, and whether a model in at least one corresponding model can be further trained can be determined by the first indication information corresponding to the model in at least one corresponding model. Alternatively, the first device can determine a model that needs and can be further trained as the first model from at least one corresponding model based on model requirement information and the second device's lack of the ability to train and / or continue training the model. Whether a model in at least one corresponding model needs further training can be determined by whether the model in at least one corresponding model meets the model requirement information, and whether a model in at least one corresponding model can be further trained can be determined by the first indication information corresponding to the model in at least one corresponding model.

[0249] In one possible scenario, the first device determines that the second device continues training the first model.

[0250] Specifically, if the second device has the ability to train a model or the ability to continue training a model, then the first device can determine that the second device will continue to train the first model.

[0251] Another possible scenario is that the first device determines that the second device will not continue training the first model.

[0252] Specifically, if the second device does not have the ability to train the model or does not have the ability to continue training the model, the first device determines that the second device will not continue training the first model.

[0253] For ease of description, the case where the second device continues to train the first model is referred to as case #1, and the case where the second device does not continue to train the first model is referred to as case #2. Case #1 and case #2 will be described separately.

[0254] Scenario #1:

[0255] For example, if the second device has the ability to train a model or the ability to continue training a model, then when the first device determines the model to respond to the first request message based on the model requirement information (e.g., the first device determines the model to respond to the first request message as either a first model or a second model based on the model requirement information), it can select the first model and send the information of the first model through the first response message. Furthermore, the second device can continue training the first model, and the model obtained by training the first model can meet the model requirement information.

[0256] It is understandable that scenario #1 also applies to situations where the first device has the ability to train or continue training the model, or where the first device does not have the ability to train or continue training the model, or where there is no first instruction information corresponding to the model. When there is no first instruction information corresponding to the model, the second device can determine whether the first model can be further trained based on the information of the first model and the model requirement information.

[0257] Scenario #2:

[0258] In some possible implementations, S815 includes S815a and S815c.

[0259] S815c: The first device continues to train the first model or instructs other members to continue training the first model based on the parameter information corresponding to the first model. The second model is the model obtained by continuing to train the first model.

[0260] For example, if the first device has the ability to train a model / continue training a model, or if other members have the ability to train a model / continue training a model, then when the first device determines the model that responds to the first request message based on the model requirement information (for example, the first device determines the model that responds to the first request message as a first model or a second model based on the model requirement information), it can select the first model. The first device can also continue training the first model or instruct other members to continue training the first model. By continuing to train the first model, a second model is obtained. The first device can also send the information of the second model through the first response message. The second model can meet the model requirement information.

[0261] In some possible implementations, S815 includes S815b and S815c.

[0262] For example, scenario #2 also applies to a first device determining a first model in at least one model and continuing to train the first model based on first indication information corresponding to at least one model corresponding to the first device.

[0263] For example, if the first instruction information indicates that the first model can continue to be trained, the first device can continue to train the first model according to the first instruction information.

[0264] It is understood that scenario #2 also applies to situations where the second device has the ability to train a model or continue training the model, or where the second device does not have the ability to train a model or continue training the model. This application does not impose such limitations on the embodiments.

[0265] S820: The first device sends a first response message to the second device. The first response message includes information about the first model or information about the second model. The first model is used to train a model that meets the model requirement information, and the second model is a model that meets the model requirement information obtained by training the first model.

[0266] Specifically, the first device sends a first response message to the second device, and the second device receives the first response message sent by the first device. The first response message includes information about the first model or information about the second model, and includes model information corresponding to the model requirement information requested by the first device.

[0267] For example, in scenario #1, the first response information includes information about the first model; in scenario #2, the first response information includes information about the second model.

[0268] When the first response information includes information about the first model, the second device needs to continue training the first model; when the first response information includes information about the second model, the second device does not need to continue training the second model.

[0269] For example, after determining the first model, the first device can determine at least one member (e.g., the first device can be a member for continuing to train the first model; other devices other than the first device can also be members for continuing to train the first model) to train the first model, and determine the model parameters of the trained model based on the model parameters obtained after the at least one member continues to train the model (e.g., aggregating the multiple model parameters obtained after multiple members continue to train the model to obtain aggregated model parameters) until the trained model parameters meet the model requirement information, at which point training can be stopped, and the model that meets the model requirement information after training is determined as the second model.

[0270] The first device can continue to train the first model based on the parameter information corresponding to the first model. The parameter information may include a first use case environment, which is the use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirements for training data used to continue training the first model; a first training member requirement, which indicates the requirements for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0271] For example, when the first response information includes information from the first model:

[0272] In some possible implementations, when the first response information includes information about the first model, the first response message also includes second instruction information, which includes at least one of the following: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained.

[0273] The second device can determine to continue training the first model based on at least one of the second instruction information.

[0274] In some other possible implementations, when the first response information includes information about the first model, the second device determines at least one of the following information based on the information about the first model and the model requirement information: the second device continues to train the first model; the first model can be continued to be trained; the first model needs to be continued to be trained.

[0275] For example, when the first response information includes information from the second model:

[0276] In some possible implementations, when the first response information includes information about the second model, the first response message also includes second indication information, which includes at least one of the following: the second device does not continue training the second model; the second model cannot be continued to be trained; the second model does not need to be continued to be trained.

[0277] In some other possible implementations, when the first response information includes information about the second model, the second device determines at least one of the following information based on the information about the second model and the model requirement information: the second device does not continue training the second model; the second model cannot be further trained; the second model does not need to be further trained.

[0278] For example, the first request message or model requirement information may include: Figures 11 to 13 The AI / ML model information discovery request #B or the aforementioned AI / ML model information discovery request #A. The first model may include... Figures 11 to 13 In model #D, the second model can include Figures 11 to 13 Model #E in the text.

[0279] In some possible implementations, the first response message may also include at least one of the following: whether the first model can continue to be trained; or whether the first model needs to be continued to be trained.

[0280] In some possible implementations, the model requirement information includes the VAL service identifier and at least one of the following: analysis accuracy; model profile, which includes at least one of the following: model domain, list of permitted vendors, ML model interoperability information, ML model stage, accuracy; analysis identifier; model requirements.

[0281] like Figure 9 As shown, Figure 9 This is a schematic diagram of a communication method 900 applicable to embodiments of this application. Method 900 can be applied between a receiving end and a sending end of a model storage system. Method 900 may include:

[0282] S910: The enabling server receives a second request message from the enabling client, or the repository receives a second request message sent by the enabling server. The second request message is used to request the storage of the first model. The second request message includes information about the first model. The second request message also includes parameter information and / or first indication information of the first model. The parameter information is used to continue training the first model, and the first indication information is used to indicate whether the first model can continue to be trained.

[0283] The ML model information storage request is one possible implementation of the second request message.

[0284] For example, when an enabling client or enabling server wants to request model storage from a repository, it can send a second request message to the repository through the enabling server. The second request message includes information about the first model that needs to be stored. Furthermore, the second request message also includes parameter information and / or first indication information corresponding to the first model. The repository can store the information about the first model and the parameter information and / or first indication information corresponding to the first model.

[0285] For example, information about the first model can be found in Table 2 (ML model information is one possible implementation of information about the first model).

[0286] For example, after the information of the first model and the parameter information and / or the first indication information corresponding to the first model are successfully stored, if other devices request a model from the repository / server, and the model determined by the repository / server according to the model requirements is the first model that does not meet the model requirements, the server or the member determined by the server to continue training the first model can continue training the first model according to the parameter information corresponding to the first model when it was stored; or, the server can determine whether the first model can be continued to be trained according to the first indication information corresponding to the first model when it was stored.

[0287] Among some possible implementations, the method also includes:

[0288] S930: The enabling server and / or repository send a second response message to the enabling client, indicating whether the first model storage was successful or failed.

[0289] The ML model information storage response is one possible implementation of the second response message.

[0290] In some possible implementations, the second request message includes parameter information, which includes a first use case environment. The first use case environment is a use case environment for continuing to train the first model. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data used to continue training the first model; a first training member requirement, which indicates the requirement for members used to continue training the first model; and first training data, which is used to continue training the first model.

[0291] For example, the parameter information may be carried in the second request message. The parameter information may include at least one of the following: requirements for training data to continue training the first model, requirements for members to continue training the first model, or first training data to continue training the first model. The server may determine the members to continue training the first model based on the first training member requirements; the server or the members determined by the server may continue training the first model based on the first training data; or, the server or the members determined by the server may determine the training data to continue training the first model based on the first training data requirements, and continue training the first model based on the training data to continue training the first model.

[0292] In some possible implementations, method 900 may also include:

[0293] Send a third instruction message, which includes second training data and / or second training members, wherein the second training data and / or second training members are determined by the enabling server based on the first model, the second training data is used to determine the first training data requirements, the second training data is used to train to obtain the first model, and the second training members are used to determine the first training member requirements, the second training members are used to train to obtain the first model.

[0294] For example, the enabling server can determine second training data for training to obtain the first model based on the first model, or the enabling server can determine second training members for training to obtain the first model based on the first model, and the second training data and / or the second training members can be used as the first training data and / or the first training members.

[0295] For example, the parameter information carried in the second request message, including the first training data requirement, the first training member requirement, or the first training data, may be the most up-to-date information compared to the second training data and / or the second training members. The enabling server can determine the members and / or data for continuing to train the first model based on the most up-to-date information; the enabling server can also determine the members and / or data for continuing to train the first model based on the second training data and / or the second training members.

[0296] Among some possible implementations, the method also includes:

[0297] S920: Enable the server and / or repository to determine whether the first model can continue training based on parameter information and / or first indication information.

[0298] For example, the enabling server and / or repository can determine whether the first model can continue training based on parameter information and / or first indication information. The enabling server and / or repository may also store indication information indicating whether the first model can continue training when storing the first model.

[0299] In some possible implementations, the first training data is required to include at least one of the following: the data type corresponding to the first model; the data label corresponding to the first model; the data domain corresponding to the first model; the analysis identifier corresponding to the first model; the dataset and data producer requirements corresponding to the first model; the valid time period corresponding to the data corresponding to the first model; and the location corresponding to the data corresponding to the first model.

[0300] In some possible implementations, the first training member is required to include at least one of the following: the member's location; the time when the member is available; the dataset corresponding to the member, which includes the original data and / or the processed data; the data features corresponding to the member; the AI / ML capabilities corresponding to the member, which include at least one of the following capabilities: client capabilities, enabling server capabilities, aggregation capabilities; and the application service identifier corresponding to the member.

[0301] In some possible implementations, the second request message may also include the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, the ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0302] For ease of description, the message sent by the enabling client to request a model from the repository based on the second request message sent by the model repository user will be referred to as the third request message. The third request message also includes the VAL service identifier of the first model and at least one of the following: the analysis accuracy of the first model; the model configuration file of the first model, which includes at least one of the following: the application domain of the first model, the list of allowed vendors, ML model interoperability information, the ML model stage, the accuracy of the first model; the analysis identifier of the first model; and the model requirements of the first model.

[0303] The above text combined Figure 8 and Figure 9 Methods 800 and 900 have been described in detail below, and the following text combines them with... Figures 10 to 12The following examples illustrate several possible scenarios in which the above-described methods 800 or 900 are applicable (for example, methods 800 and / or 900 can be applied to methods 1000, 2000, or 3000). In methods 1000, 2000, or 3000 provided in this application embodiment, the AI / ML model information discovery request #B may include the information in the above-described AI / ML model information discovery request #A, and may also include the use case environment of the AI / ML model and / or whether the requested AI / ML model can continue to be trained. The AI / ML model information discovery request #B may be a possible implementation of the above-described first request message; the AI / ML model information storage request #B may include the information in the above-described AI / ML model information storage request #A, and may also include the use case environment of the stored AI / ML model and / or whether the stored AI / ML model can continue to be trained. The AI / ML model information storage request #B may be the above-described second request. The AI / ML model information discovery response #B may include the information of the AI / ML model information discovery response #A, and may also include whether the responding AI / ML model satisfies the AI / ML model information discovery request #A, and / or whether the responding AI / ML model needs to be further trained. The AI / ML model information discovery response #B may be a possible implementation of the first response message. The model's use case environment may include the model's input parameters and / or output parameters, or the model's use case environment may be a possible implementation of the first use case environment. The AI / ML model information storage response may be a possible implementation of the second response message.

[0304] like Figure 10 As shown, Figure 10 This is a schematic diagram of a communication method 1000 applicable to embodiments of this application. The communication system may include multiple enabled clients, which may be distributed across different terminal devices or located on the same terminal device. When the AI / ML model user is located on a terminal device, the AI / ML model user can communicate with the enabling server through the enabled client; when the AI / ML model user is located on a network device (e.g., the AI / ML model user is located on an enabling server on the network side), the AI / ML model user can communicate directly with the enabling server. When the AI / ML model user communicates with the enabling server through the enabled client, the AI / ML model user needs to interact with the enabling server through its corresponding enabled client, for example... Figure 10 The AI / ML model user and its corresponding enabling client #2 are shown.

[0305] Figure 10 The method 1000 shown includes the following steps:

[0306] S1010a, AI / ML enabled client #1 sends a registration request message to AI / ML enabled server.

[0307] S1010b, the AI / ML enabling server sends a registration response message to AI / ML enabling client #1.

[0308] Specifically, AI / ML enabled clients can register by interacting with AI / ML enabled servers. This allows the AI / ML enabled server to maintain information about AI / ML members based on the content sent during client registration. AI / ML members can include devices capable of model training; for example, AI / ML members can include AI / ML enabled clients and / or AI / ML enabled servers. Because the AI / ML enabled server maintains this information, it can select AI / ML members when needed, based on this information. Furthermore, when it needs to determine the data required for further model training, the AI / ML enabled server can also determine the appropriate AI / ML members based on this information.

[0309] For example, a registration request message may indicate one or more of the following: service ID, analytics ID, app ID, and AI / ML enabling client #1 capabilities (e.g., computing power, storage capacity, etc.). A registration response message indicates whether registration was successful or failed.

[0310] Understandably, for ease of understanding, Figure 10 Only the registration of AI / ML enabled client #1 is shown in the example. Other AI / ML enabled clients in the communication system, such as AI / ML enabled client #2, can also be registered through steps similar to S1020a to S1020b. This application embodiment does not limit this.

[0311] Optionally, method 1000 may further include steps S1020a to S1020f. Steps S1020a to S1020f correspond to the AI / ML enabling client #1 obtaining training data that meets the requirements, training the model based on the obtained data, and sending the trained model to the AI / ML enabling server.

[0312] S1020a, AI / ML enabled client #1 sends a member discovery request to AI / ML enabled server.

[0313] Specifically, a member discovery request can indicate the requirements for AI / ML members. The AI / ML enabling server can select AI / ML members based on the member discovery request and the AI / ML member information maintained by the AI / ML enabling server.

[0314] For example, AI / ML enabled clients #A, #B, and #C each send registration request messages to the AI / ML enabled server and register successfully. The AI / ML enabled server maintains information about AI / ML enabled clients #A, #B, and #C. When AI / ML enabled client #1 sends a member discovery request to the AI / ML enabled server, the AI / ML enabled server can select, based on the member discovery request, the AI / ML enabled client that meets the requirements for AI / ML members in the member discovery request from among AI / ML enabled clients #A, #B, and #C.

[0315] S1020b, the AI / ML enabling server sends a member discovery response message to AI / ML enabling client #1.

[0316] Specifically, the AI / ML enabling server can send a member discovery response message that indicates a specified AI / ML member (e.g., the specified AI / ML member is the AI / ML enabling client #A that meets the requirements of the member discovery request for AI / ML members).

[0317] S1020c, AI / ML enabled client #1 retrieves data from a specified AI / ML member for training an AI / ML model.

[0318] For example, AI / ML enabled client #1 can obtain data for training AI / ML models through AI / ML enabled client #A.

[0319] S1020d, AI / ML enabled client #1 sends AI / ML model information storage request #B to AI / ML enabled server.

[0320] Specifically, after completing model training, the AI / ML enabled client #1 can send an AI / ML model information storage request #B to the AI / ML enabled server, transferring the trained model #A. The information included in the AI / ML model information storage request #B can be found in the ML model information storage request section.

[0321] For example, AI / ML enabling client #1 can be accessed through... Figure 6The method shown involves sending a trained model #A to the AI / ML enabling server and storing model #A there, thus achieving model storage. The AI / ML enabling client #1 can also send an AI / ML model information storage request #B to the AI / ML enabling server, which can then process the request. Figure 6 The method shown sends an AI / ML model information storage request #B to the repository to achieve model storage.

[0322] When the AI / ML enabled client #1 sends information about model #A to the AI / ML enabled server (see ML model information for details on model #A), it can also send the use case environment of model #A. The use case environment of model #A can include the input parameters and / or output parameters of model #A, and can indicate the applicable scenarios and problems that model #A solves.

[0323] In some possible implementations, when the AI / ML enabling client #1 sends information about model #A to the AI / ML enabling server, it can also send information indicating whether model #A can continue to be trained (for example, the AI / ML enabling client #1 can also send parameter information and / or first indication information of the first model to the AI / ML enabling server, the parameter information being used to continue training the first model, and the first indication information being used to indicate whether the first model can continue to be trained).

[0324] If the AI / ML enabling server determines that model #A can continue to be trained, it can also determine the data requirements #A for continuing to train model #A based on the parameter information sent by the AI / ML enabling client #1 (e.g., data type information corresponding to model #A, data label information corresponding to model #A, data domain information corresponding to model #A, analysis ID information corresponding to model #A, dataset and data producer requirements information corresponding to model #A, valid time period information corresponding to the data corresponding to model #A, location information corresponding to the data corresponding to model #A, etc.); or, it can also determine the requirements #A for AI / ML members used to continue training model #A (e.g., member location information, member available time information, member corresponding dataset information, member corresponding data feature information, member corresponding AI / ML capability identification, member corresponding application service identification, etc.; wherein, the member corresponding dataset may include raw data and / or processed data, and the AI / ML capability includes at least one of the following capabilities: client, server, aggregator); or, it can also determine the data used to continue training model #A. For example, the AI / ML enabling server can also determine information for continuing to train the model #A based on a first use case environment sent by the AI / ML enabling client #1. The first use case environment includes at least one of the following: a first training data requirement, which indicates the requirement for training data for continuing to train the model #A; a first training member requirement, which indicates the requirement for members for continuing to train the model #A; and first training data, which is used to continue training the model #A.

[0325] S1020e, AI / ML enabled servers deliver models to the repository.

[0326] After AI / ML enabling client #1 transmits the information of model #A to AI / ML enabling server through step S1020d, AI / ML enabling server can forward the information sent by AI / ML enabling client #1 in step S1020d to repository. AI / ML enabling client #1 can store the trained model #A information in repository through steps S1020d to S1020e.

[0327] In some other possible implementations, if step S1020d includes information indicating the data requirement #A and / or AI / ML member requirement #A for continuing training model #A, the AI / ML enabling server can determine the data requirement #B and / or AI / ML member requirement #B for continuing training the model based on one or more of the information included in the AI / ML model information storage request #B sent by the AI / ML enabling client #1 when sending information about model #A to the AI / ML enabling server. For example, the AI / ML enabling server can also determine second training data and / or second training members for continuing training model #A based on the information about model #A sent by the AI / ML enabling client #1. The second training data and / or second training members are determined by the enabling server based on the information about model #A. The second training data is used to determine the first training data requirement and is used for training to obtain model #A. The second training members are used to determine the first training member requirement and are used for training to obtain model #A.

[0328] The requirements for data #A and / or the requirements for AI / ML members #A can be the same as the requirements for data #B and / or the requirements for AI / ML members #B; the requirements for data #A and / or the requirements for AI / ML members #A can also be different from the requirements for data #B and / or the requirements for AI / ML members #B.

[0329] For example, the data requirement #A and / or AI / ML member requirement #A included in step S1020d can be requirements for continuing training of model #A. The data requirement #B and / or AI / ML member requirement #B can correspond to the requirements for training data and / or AI / ML members when model #A was last trained. The data requirement #A and / or AI / ML member requirement #A can be newer than the data requirement #B and / or AI / ML member requirement #B. By continuing to train model #A using the data requirement #A / #B and / or the AI / ML member requirement #A / #B, model #B is obtained, which can have better performance than model #A (e.g., higher analytical accuracy or lower response latency).

[0330] S1020f: The repository sends an AI / ML model information storage response, indicating whether the model storage was successful or failed.

[0331] Information included in the AI / ML model information storage response can be found in the ML model information storage response.

[0332] Optionally, method 1000 may also exclude steps S1020a to S1020d. When method 1000 excludes steps S1020a to S1020d, steps S1020e to S1020f correspond to enabling the AI / ML enabling server to train the model itself to obtain model #A and store the information of model #A in the repository.

[0333] Method 1000 may further include: enabling the server and / or repository to determine, based on parameter information and / or first indication information, whether model #A can continue to be trained.

[0334] Method 1000 may also include:

[0335] S1030a, Model repository users request models from AI / ML enabled client #2.

[0336] S1030b, AI / ML enabled client #2 requests a model from AI / ML enabled server.

[0337] Specifically, when users of the model repository are located on a terminal device, they can... Figure 7 The method shown sends an AI / ML model information discovery request #B to the AI / ML enabled client #2. The AI / ML enabled client #2 can forward the AI / ML model information discovery request #B to the AI / ML enabled server, thus fulfilling the model request. The information included in the model information discovery request #B can be found in the ML model information discovery request section.

[0338] When AI / ML enabling client #2 requests a model from AI / ML enabling server, the model information discovery request #B may also include at least one of the following: whether AI / ML enabling client #2 has the ability to train a model; whether AI / ML enabling client #2 has the ability to continue training a model.

[0339] S1030c, AI / ML enabled server for model determination.

[0340] For example, the AI / ML enabling server can select a model #C that meets the requirements from the locally stored AI / ML models based on the information in the AI / ML model information discovery request #B, or select a model #D that does not meet the requirements from the locally stored AI / ML models. However, model #D can meet the requirements for the model included in the AI / ML model information discovery request #B by continuing to train.

[0341] For example, the AI / ML enabling server can determine model #D based on the information in the AI / ML model information discovery request #B sent by the AI / ML enabling client #2, and the parameter information / first indication information of at least one model corresponding to the AI / ML enabling server, where model #D satisfies some (but not all) of the requirements of the AI / ML model information discovery request #B for the requested model.

[0342] For example, the requirements for the requested model in an AI / ML model information discovery request #B may include one or more of the following: analysis accuracy, response latency, inference accuracy, etc.

[0343] After selecting model #D, the AI / ML enabling server can determine whether to continue training model #D based on whether the AI / ML enabling client #2 has the ability to continue training model #D. If the AI / ML enabling server determines that model #D can be trained or needs to be trained, and the AI / ML enabling server or other members have the ability to continue training model #D (at this time, the AI / ML enabling client #2 may or may not have the ability to continue training model #D), model #E can be obtained according to the AI / ML model information discovery request #B in step S1030a. Model #E satisfies the requirements for the model included in the AI / ML model information discovery request #B in step S1030a.

[0344] For example, after the AI / ML enabling server selects model #D, it can discover the model selection conditions or use case environment of the model included in request #B based on the AI / ML model information, as well as the requirements for the model included in request #B, and continue training model #D to obtain model #E; or, after the AI / ML enabling server selects model #D, it can select a member that meets the requirements #A / #B of the AI / ML member corresponding to model #D, and receive model #E obtained by continuing to train model #D through the member that meets the requirements.

[0345] For example, after the AI / ML enabling server selects model #D, if the AI / ML enabling server determines that model #D can be further trained or needs to be further trained, and the AI / ML enabling client #2 has the ability to continue training model #D (at this time, the AI / ML enabling server or other members may or may not have the ability to continue training model #D), the AI / ML enabling server can discover response #A through AI / ML model information and send the information of model #D to the AI / ML enabling client #2.

[0346] Before the AI / ML enabling server or a qualified member can continue training model #D, it can obtain the data for continuing to train model #D from the member that meets the data requirements #A / #B, based on the data requirements #A / #B corresponding to model #D.

[0347] S1030d, the AI / ML enabling server sends an AI / ML model information discovery response #A to the AI / ML enabling client #2.

[0348] S1030e, AI / ML Enabled Client #2 forwards AI / ML model information discovery response #A to model repository users.

[0349] After the AI / ML enabling server determines model #C, model #D, or model #E, it can discover response #A through AI / ML model information. AI / ML model information discovery response #A includes information about model #C, model #D, or model #E.

[0350] When the AI / ML enabling server discovers response #A and sends information about model #D based on the AI / ML model information, it can also send a second instruction message. The second instruction message includes at least one of the following: AI / ML enabling client #2 continues training model #D; model #D can be continued to be trained; model #D needs to be continued to be trained.

[0351] In some possible implementations, the AI / ML enabling server sends an AI / ML model information discovery response #A to the model repository user.

[0352] For example, when a model repository user is located on a network device, they can directly send an AI / ML model information discovery request #B to the AI / ML enabling server, or directly receive an AI / ML model information discovery response #A sent by the AI / ML enabling server, without the need for forwarding by the AI / ML enabling client.

[0353] For information on AI / ML model information discovery response #A, please refer to the ML model information discovery response.

[0354] Optionally, if in step S1030c, there is no model #C or model #D that meets the requirements in the AI / ML models stored locally on the AI / ML enabling server, the following steps S1040a to S1040e can replace the above steps S1030c to S1030e.

[0355] S1040a, the AI / ML enabling server requests models from the repository.

[0356] Specifically, AI / ML enabling servers can be achieved through... Figure 7 The method shown sends an AI / ML model information discovery request #B to the repository to fulfill the model request.

[0357] S1040b, the repository determines the model.

[0358] For example, a repository can discover request #B based on AI / ML model information and select model #C from locally stored AI / ML models that meets the requirements, or select model #D, which can be further trained to meet the requirements for models included in AI / ML model information discovery request #B. For instance, the repository can select model #C or model #D based on the use case environment of the AI / ML model requested by the AI / ML enabler server, compared to the use case environments stored locally in the repository; the repository can also select model #C or model #D based on the performance of the AI / ML model requested by the AI / ML enabler server (e.g., model requirement information), compared to the performance of models stored locally in the repository.

[0359] After determining model #D, the repository can select members that meet the requirements #A / #B of the AI / ML members corresponding to model #D, and store the information of model #E obtained by continuing to train model #D through the members that meet the requirements.

[0360] For example, after the repository determines model #D, if the AI / ML enabling server meets the requirements #A / #B of the AI / ML member corresponding to model #D, the repository can send model #D to the AI / ML enabling server through AI / ML model information discovery response #A. The AI / ML enabling server can then continue training model #D to obtain model #E, and send information about model #E to AI / ML enabling client #2 or model repository user through AI / ML model information discovery response #A.

[0361] Before a member who meets the requirements can continue training model #D, they can obtain the data for continuing to train model #D from the member who meets the data requirements #A / #B, based on the data requirements #A / #B corresponding to model #D.

[0362] S1040c, The repository sends an AI / ML model information discovery response #A to the AI / ML enable server.

[0363] S1040d, the AI / ML enabling server forwards the AI / ML model information discovery response #A to the AI / ML enabling client #2.

[0364] S1040e, AI / ML Enabled Client #2 forwards AI / ML model information discovery response #A to AI model repository users.

[0365] The repository discovers response #A by sending information about model #C, model #D, or model #E through AI / ML model information discovery.

[0366] When the repository identifies model #D and a member that meets the requirements is an AI / ML enabling server, the AI / ML enabling server can continue training model #D to obtain model #E.

[0367] When the repository identifies model #D and the members that meet the requirements are other enabling servers or enabling clients, the repository can store information about model #E trained by other enabling servers or enabling clients.

[0368] After storing model #E in the repository, an AI / ML model information discovery response #A can be sent to the AI / ML enabling server. The AI / ML enabling server can forward the AI / ML model information discovery response #A to the AI / ML enabling client #2. The AI / ML enabling client #2 can then forward the AI / ML model information discovery response #A to the model repository user.

[0369] In some possible implementations, the AI / ML enabling server sends an AI / ML model information discovery response #A to the model repository user.

[0370] For example, in steps S1030a to S1040e above, if the AI / ML model user is located on a network device (e.g., the AI / ML model user is located on an enabling server on the network side), the AI / ML model user can communicate directly with the enabling server and send or receive the AI / ML model information discovery request #B in a manner similar to steps S1030a to S1040e above.

[0371] The above-mentioned training of model #D to obtain model #E can be performed by one device or by multiple devices, and the AI / ML enabling server can aggregate the multiple small models obtained by the multiple devices into model #E.

[0372] For example, after determining models #D1, #D2, and #D3 (which together form model #D), the AI / ML enabling server selects member #A and member #B for the first training iteration. Member #A trains model #D1 to obtain model #D1', and member #B trains model #D2 to obtain model #D2'. After members #A and #B complete the first training iteration, the AI / ML enabling server merges model #D1' and model #D2' into model #X1, and selects member #C for the second training iteration. Member #C trains model #D3 to obtain model #D3'. The AI / ML enabling server then merges model #X1 and model #D3' into model #X2. When model X2 meets the requirements of AI / ML model information discovery request #B, the continued training of model #D is completed, and model #X2 is adopted as model #E.

[0373] The above text combined Figure 10 This application provides a detailed description of a possible communication method 1000, in which the AI / ML enabling server can select members that meet the requirements #A / #B of the AI / ML members corresponding to model #D to continue training model #D. The following description, in conjunction with... Figure 11 This section describes in detail one possible implementation of AI / ML enabling servers to select members that meet certain requirements.

[0374] like Figure 11 As shown, Figure 11 This is a schematic diagram of a communication method 2000 applicable to an embodiment of this application.

[0375] Figure 11 The method 2000 shown includes the following steps:

[0376] In S2010a, AI / ML enabled client #1, AI / ML enabled client #2, and AI / ML enabled client #3 send registration request messages to the AI / ML enabled server.

[0377] In S2010b, the AI / ML enabling server sends registration response messages to AI / ML enabling clients #1, #2, and #3.

[0378] In S2020a, AI / ML enabled client #1 sends a member discovery request to AI / ML enabled server.

[0379] In S2020b, the AI / ML enabling server sends a member discovery response message to AI / ML enabling client #1.

[0380] S2020c, AI / ML enabled client #1 retrieves data from a specified AI / ML member for training an AI / ML model.

[0381] In S2020d, AI / ML enabled client #1 sends an AI / ML model information storage request #B to AI / ML enabled server.

[0382] In S2020e, AI / ML enables servers to pass models to repositories.

[0383] S2020f: The repository sends an AI / ML model information storage response, indicating whether the model storage was successful or failed.

[0384] Optionally, method 2000 may not include steps S2020a to S2020d. When method 2000 does not include steps S2020a to S2020d, steps S2020e to S2020f correspond to the AI / ML enabling server training the model itself to obtain model #A and storing the information of model #A in the repository.

[0385] Method 2000 may also include:

[0386] S2030a, Model repository users request models from AI / ML enabled client #2.

[0387] The specific implementation of steps S2010a to S2020f is similar to that of steps S1010a to S1020f, and can be referred to for further details. Figure 10 The textual description is omitted here from the embodiments of this application.

[0388] S2030b, AI / ML enabled client #2 requests a model from AI / ML enabled server.

[0389] Specifically, when users of the model repository are located on a terminal device, they can... Figure 7 The method shown involves sending an AI / ML model information discovery request #B to the AI / ML enabled client #2. The AI / ML enabled client #2 can then send an AI / ML model information discovery request #C to the AI / ML enabled server to fulfill the model request. The AI / ML model information discovery request #C can include the information included in the AI / ML model information discovery request #B, and can also include whether the AI / ML enabled client #2 has the capability to continue training the model.

[0390] For example, the AI / ML enabling server can select members that meet the requirements using a method similar to step S1030c described above. Specifically, if AI / ML enabling client #2 has the ability to continue training the model, the AI / ML enabling server can determine model #D, and AI / ML enabling client #2 can continue training model #D; if AI / ML enabling client #2 does not have the ability to continue training the model, the AI / ML enabling server can train model #D itself, or select other members to train model #D.

[0391] For details on the specific implementation of AI / ML enabling servers or other members training model #D, please refer to the relevant textual description of method 1000. This application embodiment will not elaborate further on this.

[0392] S2030c, AI / ML enabled servers request models from the repository.

[0393] If the AI / ML enabling server does not have a model #C that meets the requirements, or if a model #D that can meet the requirements after further training, the AI / ML enabling server can... Figure 7 The method shown sends an AI / ML model information discovery request #D to the repository to request a model. The AI / ML model information discovery request #D can include information from the AI / ML model information discovery request #B, and may also include whether the requested model includes models that need further training.

[0394] If the AI / ML enabling server and / or AI / ML enabling client #2 does not have the ability to continue training the model, the AI / ML enabling server can discover request #D through AI / ML model information and request a model #C that meets the requirements from the repository; if the AI / ML enabling server and / or AI / ML enabling client #2 has the ability to continue training the model, the AI / ML enabling server can discover request #D through AI / ML model information and request a model #C that meets the requirements and / or a model #D that meets the requirements after further training.

[0395] S2030d, the repository determines the model.

[0396] For example, the repository can discover request #D based on AI / ML model information, select a model #C that meets the requirements from the locally stored AI / ML models, or select a model #D that can meet the requirements after further training.

[0397] S2030e, the repository sends an AI / ML model information discovery response #B to the AI / ML enablement server.

[0398] Once the repository determines model #C or model #D, it can use AI / ML model information discovery response #B to send information about model #C or model #D.

[0399] Optionally, the AI / ML model information discovery response #B may also include one or more of the information that can represent the performance of the model, such as the analytical accuracy, response latency, and inference accuracy of model #C or model #D.

[0400] Optionally, the repository can also send an AI / ML model information discovery response #B to the AI / ML enable server to indicate whether the model included in the AI / ML model information discovery response #B satisfies the AI / ML model information discovery request #A / #B; or, send an AI / ML enable server to indicate whether the model included in the AI / ML model information discovery response #B needs to be further trained.

[0401] For example, the repository can use the AI / ML model information discovery response #B to send a message to the AI / ML enabling server that model #C satisfies the AI / ML model information discovery request #A / #B; or, the repository can use the AI / ML model information discovery response #B to send a message to the AI / ML enabling server that model #D does not satisfy the AI / ML model information discovery request #A / #B; or, the repository can use the AI / ML model information discovery response #B to send a message to the AI / ML enabling server that model #D needs to be further trained.

[0402] Method 2000 may also include:

[0403] S2040a, the AI / ML enabling server determines whether to continue training the model.

[0404] Specifically, the AI / ML enabling server can use the information in response #B based on the AI / ML model information to determine whether the information of the model sent by the repository meets the requirements or whether it needs to be further trained. If the information of the model sent by the repository does not meet the requirements, then the model needs to be further trained.

[0405] The AI / ML enabling server can continue training the model #D sent by the repository through an implementation similar to that in step S1030c above. If the AI / ML enabling server does not have the ability to continue training the model, it can also determine an enabling client or other enabling server for continuing to train the model #D sent by the repository, through an implementation similar to that in step S1030c above. When the AI / ML enabling server instructs the AI / ML enabling client or other enabling server to continue training the model #D, it can also send a request #A / #B for the data required to continue training the model #D; or, when the AI / ML enabling server instructs the AI / ML enabling client or other enabling server to continue training the model #D, it can also send the data required to continue training the model #D. Other enabling servers can train the model #D themselves according to the instructions of the AI / ML enabling server, and other enabling servers can also determine an enabling client for continuing to train the model #D indicated by the repository, according to the instructions of the AI / ML enabling server.

[0406] S2040b, the AI / ML enabled server, acquires data for continuing model training.

[0407] If the data requirement is met, and the member corresponding to #A / #B is AI / ML enabled client #3, which is used to continue training model #D, the data for continuing to train model #D can be obtained through AI / ML enabled client #3. For example, the AI / ML enabled server can train model #D based on the obtained data, and the AI / ML enabled server can also send the obtained data to the enabled client for continuing to train model #D or other enabled servers.

[0408] If the AI / ML enabling server does not have the ability to continue training model #D, the AI / ML enabling server can instruct the model repository user or AI / ML enabling client #3 to obtain model #E from other enabling servers that can continue training model #D through the ID or address information of other enabling servers; the AI / ML enabling server can also obtain the information of model #E from other enabling servers after training model #E on other enabling servers, and send the information of model #E to the model repository user or AI / ML enabling client #2 through AI / ML model information discovery response #A.

[0409] Method 2000 may also include:

[0410] S2050a, AI / ML enabled server continues training model #D.

[0411] S2050b, AI / ML Enable Server sends AI / ML Model Information Discovery Response #A.

[0412] S2050c, AI / ML Enabled Client #2 forwards AI / ML model information discovery response #A.

[0413] The AI / ML enabling server can continue training model #D based on the data obtained in step S2040b, and send the information of model #E obtained from training model #D to AI / ML enabling client #2 through AI / ML model information discovery response #A; AI / ML enabling client #2 can forward AI / ML model information discovery response #A to model repository user; or, AI / ML enabling server can directly send AI / ML model information discovery response #A to model repository user.

[0414] In steps S2030a to S2050c above, if the AI / ML model user is located on a network device (for example, the AI / ML model user is located on an enabling server on the network side), the AI / ML model user can communicate directly with the enabling server and send or receive the AI / ML model information discovery request #C or AI / ML model information discovery response #A / #B in a manner similar to steps S2030a to S2050c above.

[0415] The above text combined Figure 11 This application provides a detailed description of a possible communication method 2000, in which the repository determines and indicates to the AI / ML enabling server whether the model needs to continue training. The following section combines... Figure 12 This section describes in detail a possible implementation of an AI / ML enabling server that can independently determine whether a model sent from a repository needs to continue training.

[0416] like Figure 12 As shown, Figure 12 This is a schematic diagram of a communication method 3000 applicable to an embodiment of this application.

[0417] Figure 12 The method 3000 shown includes the following steps:

[0418] S3010a, AI / ML enabled client #1, AI / ML enabled client #2, and AI / ML enabled client #3 send registration request messages to the AI / ML enabled server.

[0419] In S3010b, the AI / ML enabling server sends registration response messages to AI / ML enabling clients #1, #2, and #3.

[0420] S3020a, AI / ML enabled client #1 sends a member discovery request to AI / ML enabled server.

[0421] S3020b, the AI / ML enabled server sends a member discovery response message to the AI / ML enabled client #1.

[0422] S3020c, AI / ML enabled client #1 retrieves data from a specified AI / ML member for training an AI / ML model.

[0423] S3020d, AI / ML enabled client #1 sends AI / ML model information storage request #B to AI / ML enabled server.

[0424] S3020e, AI / ML enabled servers deliver models to repositories.

[0425] S3020f: The repository sends an AI / ML model information storage response, indicating whether the model storage was successful or failed.

[0426] Optionally, method 3000 may also exclude steps S3020a to S3020d. When method 3000 excludes steps S3020a to S3020d, steps S3020e to S3020f correspond to the AI / ML enabling server training the model itself to obtain model #A and storing the information of model #A in the repository.

[0427] Method 3000 may also include:

[0428] S3030a, Model repository users request models from AI / ML enabled client #2.

[0429] S3030b, AI / ML enabled client #2 requests a model from AI / ML enabled server.

[0430] S3030c, AI / ML enabled server requests models from repository.

[0431] The specific implementation of steps S3010a to S3030c is similar to that of steps S2010a to S2030c, and can be referred to for details. Figure 11 The textual description is omitted here from the embodiments of this application.

[0432] S3030d, the repository sends an AI / ML model information discovery response #B to the AI / ML enable server.

[0433] Method 3000 may also include:

[0434] The S3030e AI / ML enabled server determines whether to continue training the model.

[0435] Specifically, the AI / ML enabling server can compare the model sent in the AI / ML model information discovery response #B with the AI / ML model information discovery request #B when the model repository user requests the model, and determine whether the model sent in the AI / ML model information discovery response #B meets the requirements of the model repository user, or whether the model sent in the AI / ML model information discovery response #B needs to be further trained.

[0436] The following section uses the example of the AI / ML enabling client #2 continuing to train the AI / ML model information discovery response #B to introduce a possible implementation of enabling the client to continue training the model #D.

[0437] Method 3000 may also include:

[0438] S3040a, the AI / ML enabling server sends an AI / ML model information discovery response #B to the AI / ML enabling client #2.

[0439] AI / ML enabling servers can discover response #B through AI / ML model information, send information about model #D, and also send the performance of model #D (e.g., one or more of the information that can represent the performance of the model, such as the analytical accuracy, response latency, and inference accuracy of model #D), or send information that model #D needs to be further trained.

[0440] S3040b, AI / ML Enabled Client #2 determines whether to continue training the model.

[0441] AI / ML Enabled Client #2 can detect response #B based on the AI / ML model information sent by the AI / ML Enabled Server and determine whether it is necessary to continue training the model.

[0442] For example, AI / ML enabled client #2 can determine whether model #D needs to continue training based on AI / ML model information discovery response #B sent by AI / ML enabled server; AI / ML enabled client #2 can also compare the model information sent by AI / ML model information discovery response #B with the AI / ML model information discovery request #B when the model repository user requests the model to determine whether the model sent by AI / ML model information discovery response #B meets the requirements of the model repository user, or whether the model sent by AI / ML model information discovery response #B needs to be further trained.

[0443] S3040c, AI / ML enabled client #2 obtains information about AI / ML members.

[0444] AI / ML Enabled Client #2 can determine the AI / ML member corresponding to the required data by using the information of AI / ML members maintained by the AI / ML Enabled Server, in a manner similar to steps S1020a to S1020b.

[0445] S3040d, AI / ML enabled client #2 obtains data for continuing model training.

[0446] For example, if AI / ML enabled client #2 determines, based on the information of AI / ML members maintained by the AI / ML enabled server, that the AI / ML member corresponding to the required data is AI / ML enabled client #3, AI / ML enabled client #2 can obtain the data from AI / ML enabled client #3 for continuing to train model #D.

[0447] S3040e, AI / ML enabled client #2 sends AI / ML model information discovery response #A.

[0448] AI / ML enabling client #2 can train model #D and obtain model #E based on data obtained from AI / ML enabling client #3. AI / ML enabling client #2 can then use AI / ML model information discovery response #A to send information about model #E to users of the model repository.

[0449] It is understood that, except for the specific instructions above, the specific implementation of each step in methods 2000 and 3000 can be referred to the textual description of the specific implementation of each step in method 1000; the above-mentioned acquisition of data for continuing training model #D through AI / ML enabled client #3 is only an illustrative example. AI / ML enabled server or AI / ML enabled client #2 can also acquire data for continuing model training through other enabled clients / servers, or multiple enabled clients / servers. For example, after determining model #D1, model #D2, and model #D3, AI / ML enabled server can acquire data for continuing training model #D1, model #D2, and model #D3 from multiple corresponding enabled clients according to the data requirements #A / #B corresponding to model #D1, model #D2, and model #D3. This application embodiment will not elaborate further on this.

[0450] The AI / ML enabling client involved in this application embodiment can be a newly added client or a software development kit (SDK) on the UE. The AI / ML enabling client can have AI / ML capabilities, or it can be an enhancement of an existing APP to have AI / ML capabilities. The AI / ML enabling server on the network side can be a newly added enabling server. The AI / ML enabling server can have AI / ML capabilities, or it can be an enhancement of an existing network element, such as an enhancement of energy aware scheduling (EAS) to have AI / ML capabilities.

[0451] It should be understood that methods 1000, 2000, and 3000 described above can also be applied to a VAL enabling server requesting a model from an AI / ML enabling server, or a VAL model repository user requesting a model from an AI / ML enabling server / enabling client. The specific implementation is similar to the steps described above. This application's embodiments will not elaborate further on this.

[0452] Regarding the above Figures 8 to 12 In the embodiments described, it should be noted that:

[0453] (1) The numbering of each step described in the embodiments is only an example and does not constitute a limitation of this application. Some steps may be added or deleted in the embodiments of this application according to actual needs.

[0454] (2) The above Figures 8 to 12 The embodiments can be implemented independently or in combination, for example... Figure 8 The illustrated embodiments and Figure 9 The illustrated embodiments can be combined with each other. Figure 8 The illustrated embodiments can be combined with Figures 10 to 12 The illustrated embodiments are combined with each other, etc.

[0455] Figure 13 and Figure 14 This is a schematic block diagram of a communication device provided in an embodiment of this application. These communication devices can be used to implement the functions of the first device, the second device, the enabling client (or the terminal device where the enabling client is located), or the enabling server (or the network device where the enabling server is located) in the above method embodiments, and thus can also achieve the beneficial effects of the above method embodiments.

[0456] Figure 13 This is a schematic block diagram of the communication device 4000 provided in an embodiment of this application. Figure 13As shown, the communication device 4000 may include a transceiver unit (or communication unit) 4020. Optionally, the communication device 4000 may further include a processing unit 4010. The communication device 4000 can be used to implement the above-mentioned... Figures 8 to 12 The methods shown in the embodiments include the functions of the first device, the second device, the enabling client (or the terminal device where the enabling client is located), or the enabling server (or the network device where the enabling server is located).

[0457] When the communication device 4000 is used to achieve Figure 8 When the first device in the method embodiment shown functions, the transceiver unit 4020 can be used to: receive a first request message or send a first response message, etc., and the processing unit 4010 can be used to: determine a first model, continue training the first model, determine whether to continue training the first model, etc.

[0458] When the communication device 4000 is used to achieve Figure 8 When the second device in the method embodiment shown functions, the transceiver unit 4020 can be used to: send a first request message or receive a first response message, etc., and the processing unit 4010 can be used to: continue training the first model according to the second instruction information / first response message, etc.

[0459] When the communication device 4000 is used to achieve Figure 9 When the repository function is implemented in the method embodiment shown, the transceiver unit 4020 can be used to: receive a third request message or send a second response message, etc., and the processing unit 4010 can be used to: determine whether the first model can be further trained based on parameter information and / or first indication information, etc.

[0460] The communication device 4000 may also include a storage unit ( Figure 13 (Not shown in the image), the storage unit can be used to store the first model indicated by the third request message, storage parameter information and / or first indication information, etc., according to the third request message.

[0461] When the communication device 4000 is used to achieve Figure 9 In the method embodiment shown, when enabling the server's function, the transceiver unit 4020 can be used to: receive a second request message or send a third request message, etc., and the processing unit 4010 can be used to: determine whether the first model can be further trained based on parameter information and / or first indication information, etc.

[0462] When the communication device 4000 is used to achieve Figure 9 When enabling the client's functionality in the method embodiment shown, the transceiver unit 4020 can be used to: send a second request message or receive a second response message, etc.

[0463] For a more detailed description of the aforementioned processing unit 4010, transceiver unit 4020, or storage unit, please refer to [reference needed]. Figures 8 to 12 The relevant descriptions in the method embodiments shown.

[0464] The apparatus 4000 of each of the above schemes includes the first apparatus, the second apparatus, the network device, the repository, the terminal device, and the enabling client (e.g., ) that implement the above methods. Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The functions of the corresponding steps performed by the enabling server (as shown) can be implemented in hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions; for example, the transceiver unit can be replaced by a transceiver (e.g., the sending unit in the transceiver unit can be replaced by a transmitter, and the receiving unit in the transceiver unit can be replaced by a receiver), and other units, such as processing units, can be replaced by processors, which respectively execute the transceiver operations and related processing operations in each method embodiment.

[0465] Furthermore, the aforementioned transceiver unit can also be a transceiver circuit (e.g., it may include a receiving circuit and a transmitting circuit), and the processing unit can be a processing circuit. The processing circuit can be one or more processors, or all or part of the circuitry within one or more processors used for control or processing functions. In embodiments of this application, Figure 13 The device mentioned can be the first device, the second device, a network device, a repository, a terminal device, an enabling client, or an enabling server as described in the foregoing embodiments. It can also be a chip or a chip system, such as a system on a chip (SoC). The transceiver unit can be an input / output circuit or a communication interface; the processing unit is a processor, microprocessor, or integrated circuit integrated on the chip. No limitations are imposed here.

[0466] Figure 14 This is a schematic block diagram of a communication device 5000 provided in an embodiment of this application. The device 5000 includes a processing circuit. The device may also include a communication circuit. The processing circuit and the communication circuit communicate with each other via an internal connection path. The processing circuit executes instructions to control the communication circuit to send and / or receive signals.

[0467] Taking a processing circuit that includes one or more processors and a communication circuit that is a transceiver as an example, such as Figure 14As shown, the communication device 5000 may include a transceiver 5020, and optionally, the communication device 5000 may also include a processor 5010. The processor 5010 and the transceiver 5020 are coupled to each other. It is understood that the transceiver 5020 may be a transceiver or an input / output interface. Optionally, the communication device 5000 may also include a memory 5030 for storing instructions executed by the processor 5010, or storing input data required by the processor 5010 to execute instructions, or storing data generated after the processor 5010 executes instructions. Sometimes, the transceiver 5020 may also be understood as part of the processor 5010, in which case the communication device 5000 includes the processor 5010.

[0468] In one possible implementation, device 5000 is used to implement the first device, second device, network device, repository, terminal device, and enabling client (e.g., in the above method embodiments). Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The various processes and steps corresponding to the enabling server are shown below.

[0469] It is understood that device 5000 can specifically be the first device, the second device, the network device, the repository, the terminal device, or the enabling client (e.g., in the above embodiments). Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The enabling server shown can also be a chip or chip system. Correspondingly, the communication circuit can be the interface circuit of the chip, or an input / output circuit, which is not limited here. Specifically, the device 5000 can be used to perform the above method embodiments with the first device, the second device, the network device, the repository, the terminal device, and the enabling client (e.g., Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The steps and / or processes corresponding to the enabling server shown.

[0470] When the communication device 5000 is used to achieve Figures 8 to 12 In the method shown, the processor 5010 is used to implement the functions of the processing unit 4010, and the transceiver 5020 is used to implement the functions of the transceiver unit 4020.

[0471] It is understood that, in order to achieve the functions described in the above embodiments, the first device, the second device, the network device, the repository, the terminal device, and the enabling client (e.g., Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The enabling server shown includes the corresponding hardware structure and / or software modules for performing various functions. Those skilled in the art will readily recognize that, based on the units and method steps described in conjunction with the embodiments disclosed in this application, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application scenario and design constraints of the technical solution.

[0472] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), image processors, artificial intelligence processors, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0473] The method steps in the embodiments of this application can be implemented in hardware or in software instructions executable by a processor. The software instructions can consist of corresponding software modules, which can be stored in random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. The storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC. Furthermore, the ASIC can reside in a first device, a second device, a network device, a repository, a terminal device, or an enabled client (e.g., [example device]). Figure 9 The enabled client shown Figures 10 to 12The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 The processor and storage medium may also exist as discrete components in the first device, the second device, the network device, the repository, the terminal device, and the enabling client (e.g., the enabling server shown). Figure 9 The enabled client shown Figures 10 to 12 The AI / ML enabled client #1, AI / ML enabled client #2, or AI / ML enabled client #3 shown, or the enabling server (e.g., Figures 9 to 12 In the enabled server shown.

[0474] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are performed entirely or partially. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, or other programmable device. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, enabling server, or data center to another website, computer, enabling server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as an enabling server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; it can also be an optical medium, such as a digital video optical disc; or it can be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both types of storage media.

[0475] In the above embodiments, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced by each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0476] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0477] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0478] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0479] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0480] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0481] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, an enabling server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0482] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A communication method, characterized in that, Applied to the first device, comprising: Receive a first request message, which includes model requirement information; Send a first response message, the first response message including information of a first model or information of a second model, wherein the first model is used to train a model that meets the model requirement information, and the second model is a model that meets the model requirement information obtained by training the first model.

2. The method according to claim 1, characterized in that, Receiving the first request message includes: Receive the first request message from the second device, wherein the first request message further includes at least one of the following: Does the second device have the ability to train a model? Does the second device have the ability to continue training the model? 3. The method according to claim 1 or 2, characterized in that, Before sending the first response message, the method further includes: Based on at least one model corresponding to the first device, the parameter information corresponding to the at least one model, and the model requirement information, the first model is determined from the at least one model, and the parameter information is used to continue training the corresponding at least one model.

4. The method according to claim 1 or 2, characterized in that, Before sending the first response message, the method further includes: Based on at least one model corresponding to the first device, first indication information corresponding to the at least one model, and model requirement information, the first model is determined among the at least one model, and the first indication information is used to indicate whether the first model can continue to be trained.

5. The method according to claim 3 or 4, characterized in that, The method further includes: The first model is trained to obtain the second model, and the first response message includes information about the second model.

6. The method according to any one of claims 1 to 4, characterized in that, When the first response message includes information about the first model, the first response message also includes second indication information, the second indication information including at least one of the following: The second device continues to train the first model; The first model can be further trained; The first model needs to be trained.

7. The method according to claim 3 or 4, characterized in that, Before sending the first response message, the method further includes: Whether to continue training the first model is determined based on whether the second device has the ability to train the model and / or whether the second device has the ability to continue training the model.

8. A communication method, characterized in that, Applied to a second device, comprising: Send a first request message, which includes model requirement information; A first response message is received, the first response message including information of a first model or information of a second model, wherein the first model is used to train a model that meets the model requirement information, and the second model is a model that meets the model requirement information obtained by training the first model.

9. The method according to claim 8, characterized in that, The first request message also includes at least one of the following: Does the second device have the ability to train a model? Does the second device have the ability to continue training the model? 10. The method according to claim 8 or 9, characterized in that, The first response message also includes second indication information corresponding to the first model, the second indication information including at least one of the following: The second device continues to train the first model; The first model can be further trained; The first model needs to be trained.

11. The method according to claim 10, characterized in that, The second device has the ability to train a model and / or the ability to continue training the model.

12. The method according to claim 10 or 11, characterized in that, The method further includes: Continue training the first model based on the second instruction information.

13. The method according to any one of claims 8 to 11, characterized in that, The method further includes: Based on the information of the first model, the parameter information of the first model, and the model requirement information, determine whether to continue training the first model; and / or, Based on the information from the first model, the parameter information of the first model, and the model requirement information, the first model continues to be trained.

14. A communication method, characterized in that, The receiving end used for model storage includes: A second request message is received, which is used to request storage of the first model. The second request message includes information about the first model, and also includes parameter information and / or first indication information of the first model. The parameter information is used to continue training the first model, and the first indication information is used to indicate whether the first model can continue to be trained. Send a second response message, which indicates whether the first model storage was successful or failed.

15. A communication method, characterized in that, The sending end used for model storage includes: Send a second request message, which is used to request storage of the first model. The second request message includes information about the first model, and also includes parameter information and / or first indication information of the first model. The parameter information is used to continue training the first model, and the first indication information is used to indicate whether the first model can be continued to be trained. Receive a second response message, which indicates whether the first model storage was successful or failed.

16. A communication device, characterized in that, include: A processor for executing computer instructions stored in memory to cause the apparatus to perform: the method of any one of claims 1 to 7, or the method of any one of claims 8 to 13, or the method of claim 14, or the method of claim 15.

17. The apparatus according to claim 16, characterized in that, The device is a chip or chip system.

18. A computer program product, characterized in that, When the computer program in the computer program product is executed by a communication device, it implements the method as described in any one of claims 1 to 7, or the method as described in any one of claims 8 to 13, or the method as described in claim 14, or the method as described in claim 15.

19. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions that, when executed by a communication device, implement the method as described in any one of claims 1 to 7, or the method as described in any one of claims 8 to 13, or the method as described in claim 14, or the method as described in claim 15.

Images