Model training method and device and storage medium
By building a closed-loop model training process in the communication system and collaborating with multiple entities for AI model training, the problem that existing systems cannot provide AI model training services is solved, and the provision and quality assurance of intelligent endogenous services are achieved.
Patent Information
- Application Number
- CN202310652135.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-01
- Publication Date
- 2024-12-03
AI Technical Summary
Existing mobile communication systems cannot provide AI model training services and cannot meet users' needs for intelligent endogenous services.
By introducing interactions between multiple entities in the communication system, a closed-loop model training process is built, including requirements analysis, network adaptation, network collaboration and network execution stages, determining the target service quality indicator parameters and execution operation list, and collaborating with multiple entities Perform model training.
The communication system provides model training services to users, the network or third parties, meets the quality requirements of model training services, and improves the intelligent endogenous capabilities of the system.
Smart Images

Figure CN119071834A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of communication technology, and in particular to a model training method, device and storage medium. Background Art
[0002] The inherent intelligence of the sixth generation mobile communication system (6G) can refer to the fact that 6G networks natively support artificial intelligence (AI), take AI capabilities as the basic service of the network, realize AI as a Service (AIaaS), enable the network to self-learn and self-evolve, and then provide AI model training services, AI reasoning services, AI deployment and open services, etc. to users, networks or third parties.
[0003] However, the network architecture and network processes of existing mobile communication systems can only provide users with connection and data transmission services, but cannot provide intelligent intrinsic services such as AI model training services. Summary of the invention
[0004] The present application provides a model training method, device and storage medium to solve the technical problem that the communication system cannot provide intelligent endogenous services such as AI model training services.
[0005] In a first aspect, an embodiment of the present application provides a model training method, which is applied to a first entity, and the method includes:
[0006] Receiving a model training request message sent by the second entity;
[0007] According to the model training request message, determine the model training related information, the model training related information is used to determine the target service quality QoS indicator parameter list and the execution operation list, the execution operation list is used to indicate the execution operations of multiple entities in the model training;
[0008] Sending a first indication message to a third entity, where the first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list;
[0009] Receive a first response message sent by the third entity, the first response message including the model training result
[0010] A second response message is sent to the second entity, where the second response message includes the model training result.
[0011] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0012] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0013] In one implementation, the first indication message includes a target QoS indicator parameter list and an execution operation list; before sending the first indication message to the third entity, it also includes:
[0014] According to the model training related information, obtain the target QoS indicator parameter list;
[0015] obtaining network resource information from a fourth entity;
[0016] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0017] In one implementation, obtaining a target QoS indicator parameter list according to model training related information includes:
[0018] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0019] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0020] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0021] In one implementation, obtaining a target QoS indicator parameter list according to model training related information includes:
[0022] Sending model training related information to the fifth entity;
[0023] Receive the target QoS indicator parameter list sent by the fifth entity.
[0024] In one embodiment, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
[0025] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0026] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0027] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0028] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0029] The synergy indicator parameters include synergy type.
[0030] In one implementation, the first indication message further includes a service identifier of the model training, a plurality of indication parameters, and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage;
[0031] The first response message also includes a service identifier and a plurality of indication parameters;
[0032] The second response message also includes a service identifier and multiple indication parameters.
[0033] In one implementation, the first entity is deployed in a first access network device or a first core network device;
[0034] The second entity is the first terminal device or the second entity is deployed in the second terminal device, the second access network device or the second core network device;
[0035] The third entity is a third terminal device or the third entity is deployed on a fourth terminal device, a third access network device or a third core network device;
[0036] The fourth entity is deployed in a fourth access network device or a fourth core network device;
[0037] The fifth entity is deployed on a fifth access network device or a fifth core network device;
[0038] Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device;
[0039] The first terminal device, the third terminal device and the fifth terminal device are the same or different, the second terminal device, the fourth terminal device and the sixth terminal device are the same or different, the first access network device, the second access network device, the third access network device, the fourth access network device, the fifth access network device and the sixth access network device are the same or different, and the first core network device, the second core network device, the third core network device, the fourth core network device, the fifth core network device and the sixth core network device are the same or different.
[0040] In a second aspect, an embodiment of the present application provides a model training method, which is applied to a third entity, and the method includes:
[0041] Receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training;
[0042] Sending an operation execution request message to at least one sixth entity according to the target QoS indicator parameter list and the operation execution list, where the operation execution request message is used to instruct the at least one sixth entity to perform model training;
[0043] receiving an execution operation response message sent by at least one sixth entity;
[0044] A first response message is sent to the first entity, where the first response message includes a model training result.
[0045] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0046] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0047] The execution operation response message includes a service identifier of the model training, an identifier of the target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
[0048] In one implementation, the first indication message includes model training related information; before sending the execution operation request message to at least one sixth entity, the message further includes:
[0049] According to the model training related information, obtain the target QoS indicator parameter list;
[0050] obtaining network resource information from a fourth entity;
[0051] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0052] In one implementation, obtaining a target QoS indicator parameter list according to model training related information includes:
[0053] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0054] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0055] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0056] In one implementation, obtaining a target QoS indicator parameter list according to model training related information includes:
[0057] Sending model training related information to the fifth entity;
[0058] Receive the target QoS indicator parameter list sent by the fifth entity.
[0059] In one implementation, the target QoS indicator parameter list includes a target QoS indicator parameter of a third entity and a target QoS indicator parameter of at least one sixth entity.
[0060] In one embodiment, the first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes the execution operation of the third entity;
[0061] Sending an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list includes:
[0062] Acquire a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0063] obtaining network resource information from a fourth entity;
[0064] Determining an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity;
[0065] An execution operation request message is sent to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
[0066] In one implementation, obtaining a target QoS indicator parameter of at least one sixth entity includes:
[0067] Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0068] Sending at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of at least one sixth entity;
[0069] A target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that at least one candidate QoS indicator parameter is a target QoS indicator parameter of at least one sixth entity.
[0070] In one implementation, obtaining a target QoS indicator parameter of at least one sixth entity includes:
[0071] Sending a target QoS indicator parameter of the third entity and an identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity;
[0072] Receive at least one target QoS indicator parameter of a sixth entity sent by the fifth entity.
[0073] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0074] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0075] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0076] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0077] The synergy indicator parameters include synergy type.
[0078] In one embodiment, the method further comprises:
[0079] receiving feedback information from at least one sixth entity;
[0080] Sending a second instruction message to at least one sixth entity according to the feedback information, where the second instruction message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0081] A third response message sent by at least one sixth entity is received.
[0082] In one implementation, the first entity is deployed in a first access network device or a first core network device;
[0083] The third entity is a third terminal device or the third entity is deployed on a fourth terminal device, a third access network device or a third core network device;
[0084] The fifth entity is deployed on a fifth access network device or a fifth core network device;
[0085] Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device;
[0086] The third terminal device is the same as or different from the fifth terminal device, the fourth terminal device is the same as or different from the sixth terminal device, the first access network device, the third access network device, the fifth access network device and the sixth access network device are the same as or different, and the first core network device, the third core network device, the fifth core network device and the sixth core network device are the same as or different.
[0087] In a third aspect, an embodiment of the present application provides a model training method, which is applied to a sixth entity, and the method includes:
[0088] Receiving an execution operation request message sent by a third entity;
[0089] According to the execution operation request message, the model training is performed to obtain the model training result;
[0090] An execution operation response message is sent to the third entity, where the execution operation response message includes the model training result.
[0091] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0092] The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity, and multiple indication parameters.
[0093] In one embodiment, the method further comprises:
[0094] Sending feedback information to a third entity;
[0095] receiving a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update the model or to perform resource collaborative adjustment;
[0096] A third response message is sent to the third entity.
[0097] In one implementation, the third entity is a third terminal device or the third entity is deployed on a fourth terminal device, a third access network device, or a third core network device;
[0098] Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device;
[0099] The third terminal device is the same as or different from the fifth terminal device, the fourth terminal device is the same as or different from the sixth terminal device, the third access network device is the same as or different from the sixth access network device, and the third core network device is the same as or different from the sixth core network device.
[0100] In a fourth aspect, an embodiment of the present application provides a model training method, which is applied to a second entity, and the method includes:
[0101] Sending a model training request message to the first entity;
[0102] A second response message sent by the first entity is received, where the second response message includes a model training result.
[0103] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0104] In one embodiment, the second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
[0105] In one implementation, the first entity is deployed in a first access network device or a first core network device;
[0106] The second entity is the first terminal device or the second entity is deployed in the second terminal device, the second access network device or the second core network device;
[0107] The first access network device is the same as or different from the second access network device, and the first core network device is the same as or different from the second core network device.
[0108] In a fifth aspect, an embodiment of the present application provides a model training device, including:
[0109] A first receiving unit, configured to receive a model training request message sent by a second entity;
[0110] A first determining unit, configured to determine model training related information according to a model training request message, wherein the model training related information is used to determine a target quality of service QoS indicator parameter list and an execution operation list, wherein the execution operation list is used to indicate execution operations of multiple entities in the model training;
[0111] A first sending unit, configured to send a first indication message to a third entity, where the first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list;
[0112] The second receiving unit is used to receive a first response message sent by the third entity, where the first response message includes a model training result
[0113] The second sending unit is used to send a second response message to the second entity, and the second response message includes a model training result.
[0114] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0115] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0116] In one embodiment, the first indication message includes a target QoS indicator parameter list and an execution operation list; the device also includes:
[0117] A first acquisition unit is used to acquire a target QoS indicator parameter list according to model training related information;
[0118] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0119] The second determining unit is used to determine the execution operation list according to the target QoS indicator parameter list and the network resource information.
[0120] In one implementation, the first acquisition unit is specifically configured to:
[0121] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0122] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0123] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0124] In one implementation, the first acquisition unit is specifically configured to:
[0125] Sending model training related information to the fifth entity;
[0126] Receive the target QoS indicator parameter list sent by the fifth entity.
[0127] In one embodiment, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
[0128] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0129] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0130] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0131] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0132] The synergy indicator parameters include synergy type.
[0133] In one implementation, the first indication message further includes a service identifier of the model training, a plurality of indication parameters, and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage;
[0134] The first response message also includes a service identifier and a plurality of indication parameters;
[0135] The second response message also includes a service identifier and multiple indication parameters.
[0136] In a sixth aspect, an embodiment of the present application provides a model training device, including:
[0137] A first receiving unit is used to receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training;
[0138] A first sending unit, configured to send an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list, wherein the execution operation request message is used to instruct the at least one sixth entity to perform model training;
[0139] A second receiving unit, configured to receive an execution operation response message sent by at least one sixth entity;
[0140] The second sending unit is used to send a first response message to the first entity, where the first response message includes a model training result.
[0141] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0142] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0143] The execution operation response message includes a service identifier of the model training, an identifier of the target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
[0144] In one implementation, the first indication message includes model training related information, and the device further includes:
[0145] A first acquisition unit is used to acquire a target QoS indicator parameter list according to model training related information;
[0146] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0147] The determination unit is used to determine the execution operation list according to the target QoS indicator parameter list and network resource information.
[0148] In one implementation, the first acquisition unit is specifically configured to:
[0149] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0150] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0151] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0152] In one implementation, the first acquisition unit is specifically configured to:
[0153] Sending model training related information to the fifth entity;
[0154] Receive the target QoS indicator parameter list sent by the fifth entity.
[0155] In one implementation, the target QoS indicator parameter list includes a target QoS indicator parameter of a third entity and a target QoS indicator parameter of at least one sixth entity.
[0156] In one implementation, the first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes the execution operation of the third entity;
[0157] The first acquisition unit is further configured to acquire a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0158] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0159] The determining unit is further configured to determine an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity;
[0160] The first sending unit is further used to send an execution operation request message to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
[0161] In one implementation, the first acquisition unit is specifically configured to:
[0162] Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0163] Sending at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of at least one sixth entity;
[0164] A target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that at least one candidate QoS indicator parameter is a target QoS indicator parameter of at least one sixth entity.
[0165] In one implementation, the first acquisition unit is specifically configured to:
[0166] Sending a target QoS indicator parameter of the third entity and an identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity;
[0167] Receive at least one target QoS indicator parameter of a sixth entity sent by the fifth entity.
[0168] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0169] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0170] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0171] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0172] The synergy indicator parameters include synergy type.
[0173] In one embodiment, the device further comprises:
[0174] A third receiving unit, configured to receive feedback information from at least one sixth entity;
[0175] A third sending unit, configured to send a second indication message to at least one sixth entity according to the feedback information, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0176] The fourth receiving unit is used to receive a third response message sent by at least one sixth entity.
[0177] In a seventh aspect, an embodiment of the present application provides a model training device, comprising:
[0178] A first receiving unit, configured to receive an execution operation request message sent by a third entity;
[0179] A model training unit, used to perform model training according to the execution operation request message and obtain a model training result;
[0180] The first sending unit is used to send an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0181] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0182] The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity, and multiple indication parameters.
[0183] In one embodiment, the device further comprises:
[0184] A second sending unit, configured to send feedback information to a third entity;
[0185] A second receiving unit, used to receive a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0186] The third sending unit is used to send a third response message to the third entity.
[0187] In an eighth aspect, an embodiment of the present application provides a model training device, comprising:
[0188] A sending unit, configured to send a model training request message to the first entity;
[0189] A receiving unit is used to receive a second response message sent by the first entity, where the second response message includes a model training result.
[0190] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0191] In one embodiment, the second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
[0192] In a ninth aspect, an embodiment of the present application provides a model training device, including a memory, a transceiver, and a processor:
[0193] A memory is used to store a computer program; a transceiver is used to send and receive data under the control of a processor; and a processor is used to read the computer program in the memory and perform the following operations:
[0194] Receiving a model training request message sent by the second entity;
[0195] According to the model training request message, determine the model training related information, the model training related information is used to determine the target service quality QoS indicator parameter list and the execution operation list, the execution operation list is used to indicate the execution operations of multiple entities in the model training;
[0196] Sending a first indication message to a third entity, where the first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list;
[0197] Receive a first response message sent by the third entity, the first response message including the model training result
[0198] A second response message is sent to the second entity, where the second response message includes the model training result.
[0199] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0200] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0201] In one implementation, the first indication message includes a target QoS indicator parameter list and an execution operation list; the processor is further configured to perform the following operations:
[0202] According to the model training related information, obtain the target QoS indicator parameter list;
[0203] obtaining network resource information from a fourth entity;
[0204] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0205] In one implementation, the processor is specifically configured to perform the following operations:
[0206] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0207] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0208] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0209] In one implementation, the processor is specifically configured to perform the following operations:
[0210] Sending model training related information to the fifth entity;
[0211] Receive the target QoS indicator parameter list sent by the fifth entity.
[0212] In one embodiment, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
[0213] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0214] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0215] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0216] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0217] The synergy indicator parameters include synergy type.
[0218] In one implementation, the first indication message further includes a service identifier of the model training, a plurality of indication parameters, and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage;
[0219] The first response message also includes a service identifier and a plurality of indication parameters;
[0220] The second response message also includes a service identifier and multiple indication parameters.
[0221] In a tenth aspect, an embodiment of the present application provides a model training device, including a memory, a transceiver, and a processor:
[0222] A memory is used to store a computer program; a transceiver is used to send and receive data under the control of a processor; and a processor is used to read the computer program in the memory and perform the following operations:
[0223] Receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training;
[0224] Sending an operation execution request message to at least one sixth entity according to the target QoS indicator parameter list and the operation execution list, where the operation execution request message is used to instruct the at least one sixth entity to perform model training;
[0225] receiving an execution operation response message sent by at least one sixth entity;
[0226] A first response message is sent to the first entity, where the first response message includes a model training result.
[0227] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0228] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0229] The execution operation response message includes a service identifier of the model training, an identifier of the target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
[0230] In one implementation, the first indication message includes model training related information; the processor is further configured to perform the following operations:
[0231] According to the model training related information, obtain the target QoS indicator parameter list;
[0232] obtaining network resource information from a fourth entity;
[0233] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0234] In one implementation, the processor is specifically configured to perform the following operations:
[0235] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0236] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0237] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0238] In one implementation, the processor is specifically configured to perform the following operations:
[0239] Sending model training related information to the fifth entity;
[0240] Receive the target QoS indicator parameter list sent by the fifth entity.
[0241] In one implementation, the target QoS indicator parameter list includes a target QoS indicator parameter of a third entity and a target QoS indicator parameter of at least one sixth entity.
[0242] In one embodiment, the first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes an execution operation of the third entity; the processor is further used to perform the following operations:
[0243] Acquire a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0244] obtaining network resource information from a fourth entity;
[0245] Determining an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity;
[0246] An execution operation request message is sent to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
[0247] In one implementation, the processor is specifically configured to perform the following operations:
[0248] Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0249] Sending at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of at least one sixth entity;
[0250] A target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that at least one candidate QoS indicator parameter is a target QoS indicator parameter of at least one sixth entity.
[0251] In one implementation, the processor is specifically configured to perform the following operations:
[0252] Sending a target QoS indicator parameter of the third entity and an identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity;
[0253] Receive at least one target QoS indicator parameter of a sixth entity sent by the fifth entity.
[0254] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0255] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0256] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0257] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0258] The synergy indicator parameters include synergy type.
[0259] In one embodiment, the processor is further configured to perform the following operations:
[0260] receiving feedback information from at least one sixth entity;
[0261] Sending a second instruction message to at least one sixth entity according to the feedback information, where the second instruction message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0262] A third response message sent by at least one sixth entity is received.
[0263] In an eleventh aspect, an embodiment of the present application provides a model training device, including a memory, a transceiver, and a processor:
[0264] A memory is used to store a computer program; a transceiver is used to send and receive data under the control of a processor; and a processor is used to read the computer program in the memory and perform the following operations:
[0265] Receiving an execution operation request message sent by a third entity;
[0266] According to the execution operation request message, the model training is performed to obtain the model training result;
[0267] An execution operation response message is sent to the third entity, where the execution operation response message includes the model training result.
[0268] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0269] The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity, and multiple indication parameters.
[0270] In one embodiment, the processor is further configured to perform the following operations:
[0271] Sending feedback information to a third entity;
[0272] receiving a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update the model or to perform resource collaborative adjustment;
[0273] A third response message is sent to the third entity.
[0274] In a twelfth aspect, an embodiment of the present application provides a model training device, including a memory, a transceiver, and a processor:
[0275] A memory is used to store a computer program; a transceiver is used to send and receive data under the control of a processor; and a processor is used to read the computer program in the memory and perform the following operations:
[0276] Sending a model training request message to the first entity;
[0277] A second response message sent by the first entity is received, where the second response message includes a model training result.
[0278] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0279] In one embodiment, the second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
[0280] In a thirteenth aspect, an embodiment of the present application provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute any method of the first aspect.
[0281] In a fourteenth aspect, an embodiment of the present application provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute any method of the second aspect.
[0282] In the fifteenth aspect, an embodiment of the present application provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute any method of the third aspect.
[0283] In the sixteenth aspect, an embodiment of the present application provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute any method of the fourth aspect.
[0284] The present application provides a model training method, device and storage medium. The model training method constructs a closed-loop model training process through the interaction between various entities in a communication system, so that the communication system can provide model training services to users, networks or third parties on demand.
[0285] It should be understood that the contents described in the above summary of the invention are not intended to limit the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0286] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0287] Figure 1 A schematic diagram of an architecture of a communication system in the related art;
[0288] Figure 2 A schematic diagram of an architecture of a communication system provided in an embodiment of the present application;
[0289] Figure 3 A flowchart of a model training method provided in an embodiment of the present application;
[0290] Figure 4 A flowchart of another model training method provided in an embodiment of the present application;
[0291] Figure 5 A flowchart of another model training method provided in an embodiment of the present application;
[0292] Figure 6 A flowchart of another model training method provided in an embodiment of the present application;
[0293] Figure 7 A flowchart of an interaction method provided in an embodiment of the present application;
[0294] Figure 8 A schematic diagram of the structure of a model training device 800 provided in an embodiment of the present application;
[0295] Fig.9A A schematic diagram of the structure of a model training device 900 provided in an embodiment of the present application;
[0296] Fig. 9B Another structural schematic diagram of the model training device 900 provided in an embodiment of the present application;
[0297] Fig. 10A A schematic diagram of the structure of a model training device 1000 provided in an embodiment of the present application;
[0298] Fig. 10B Another structural schematic diagram of the model training device 1000 provided in an embodiment of the present application;
[0299] Fig.11A A schematic diagram of the structure of a model training device 1100 provided in an embodiment of the present application;
[0300] Fig. 11B Another structural schematic diagram of the model training device 1100 provided in an embodiment of the present application;
[0301] Fig.12 A schematic diagram of the structure of a model training device 1200 provided in an embodiment of the present application;
[0302] Fig.13 A schematic diagram of the structure of a model training device 1300 provided in an embodiment of the present application;
[0303] Fig.14 A schematic diagram of the structure of a model training device 1400 provided in an embodiment of the present application;
[0304] Fig.15 A schematic diagram of the structure of the model training device 1500 provided in an embodiment of the present application. DETAILED DESCRIPTION
[0305] In the embodiments of the present invention, the term "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.
[0306] In the embodiments of the present application, the term "plurality" refers to two or more than two, and other quantifiers are similar.
[0307] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0308] The embodiments of the present application provide a model training method and device for providing model training services to users on demand.
[0309] Among them, the method and the device are based on the same application concept. Since the method and the device solve the problem in a similar principle, the implementation of the device and the method can refer to each other, and the repeated parts will not be repeated.
[0310] The terminal device involved in the embodiment of the present application may be a device that provides voice and / or data connectivity to a user, a handheld device with a wireless connection function, or other processing devices connected to a wireless modem. In different systems, the name of the terminal device may also be different. For example, in the fifth generation (5th Generation, 5G) communication system, the terminal device may be called a user equipment (User Equipment, UE). The wireless terminal device can communicate with one or more core networks (Core Network, CN) via a radio access network (Radio Access Network, RAN). The wireless terminal device can be a mobile terminal device, such as a mobile phone (or "cellular" phone) and a computer with a mobile terminal device. For example, it can be a portable, pocket, handheld, computer-built-in or vehicle-mounted mobile device that exchanges language and / or data with a wireless access network. For example, personal communication service (Personal Communication Service, PCS) phone, cordless phone, Session Initiated Protocol (Session Initiated Protocol, SIP) phone, wireless local loop (Wireless Local Loop, WLL) station, personal digital assistant (Personal Digital Assistant, PDA) and other devices. The wireless terminal device may also be referred to as a system, a subscriber unit, a subscriber station, a mobile station, a mobile station, a remote station, an access point, a remote terminal device, an access terminal device, a user terminal device, a user agent, and a user device, but is not limited in the embodiments of the present application.
[0311] The access network device involved in the embodiments of the present application may be a base station, which may include multiple cells providing services for the terminal. Depending on the specific application scenario, the base station may also be called an access point, or may be a device in the access network that communicates with the wireless terminal device through one or more sectors on the air interface, or other names. The access network device may be used to interchange received air frames with Internet Protocol (IP) packets, and serve as a router between the wireless terminal device and the rest of the access network, wherein the rest of the access network may include an IP communication network. The access network device may also coordinate the attribute management of the air interface. For example, the access network device involved in the embodiments of the present application may be an access network device (Base Transceiver Station, BTS) in the Global System for Mobile communications (Global System for Mobile communications, GSM) or Code Division Multiple Access (Code Division Multiple Access, CDMA), or an access network device (NodeB) in Wide-band Code Division Multiple Access (WCDMA), or an evolutionary access network device (evolutional Node B, eNB or e-NodeB) in the long term evolution (long term evolution, LTE) system, a 5G base station (gNB) in the 5G network architecture (next generation system), or a home evolved Node B (Home evolved Node B, HeNB), a relay node, a home base station (femto), a pico base station (pico), etc., which is not limited in the embodiments of the present application. In some network structures, the access network device may include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit may also be arranged geographically separately.
[0312] Figure 1 FIG. 1 is a schematic diagram of an architecture of a communication system in the related technology. Figure 1 As shown, the architecture includes terminal equipment, access network equipment, core network equipment and data network (DN) part.
[0313] Among them, the core network equipment may also be referred to as a core network element or a core network functional entity. As an example, the core network functional entity may include any of the following: a Policy Control Function (PCF) entity, a Network Slice Selection Function (NSSF) entity, a Unified Data Management (UDM) entity, a Network Repository Function (NRF) entity, a Session Management Function (SMF) entity, an Access and Mobility Management Function (AMF) entity, and a User Plane Function (UPF) entity. As an example, the functions of the above functional entities are as follows:
[0314] The PCF entity is mainly used to guide the unified policy framework of network behavior and provide policy rule information for control plane entities (such as AMF, SMF, etc.).
[0315] The AMF entity is mainly used for access control, mobility management, registration and deregistration functions.
[0316] The SMF entity is mainly used for user plane network element selection, user plane network element redirection, IP address allocation for terminal devices, and session establishment, modification and release as well as Quality of Service (QoS) control.
[0317] The UPF entity is used to receive and forward user plane data. For example, the UPF entity can receive user plane data from a service server and send the user plane data to a terminal device through an access network device. The UPF entity can also receive user plane data from a terminal device through an access network device and forward it to a service server.
[0318] NSSF entity is mainly used for network slice selection.
[0319] The UDM entity is mainly used for the contract data management of terminal devices, including the storage and management of terminal device identification, access authorization of terminal devices, etc.
[0320] The data network part can be a data network that provides business services to users. Generally, the client is located in the terminal device and the server is located in the data network. The data network can be a private network, such as a local area network, or an external network that is not controlled by the operator, such as the Internet, or a proprietary network jointly deployed by operators.
[0321] The current network architecture of 5G mobile communication system can be Figure 1 As shown, similar to previous generations of mobile communication systems, it is designed with connection and data transmission as the center, and can only provide users with connection and data transmission services, but cannot provide intelligent intrinsic services such as model training services, let alone guarantee the QoS requirements of model training services.
[0322] Based on the technical problems in the prior art, an embodiment of the present application provides a model training method, which can construct a closed-loop model training process through the interaction between multiple entities in a communication system, so that the communication system can provide model training services to users, networks or third parties on demand.
[0323] In an embodiment of the present application, providing a model training service through a communication system may refer to completing the model training process through interaction between multiple entities in the communication system; the model training service may refer to the entire model training process.
[0324] Figure 2 A schematic diagram of the architecture of a communication system provided in an embodiment of the present application. Figure 2 As shown, the architecture includes a first entity, a second entity, a third entity, a fourth entity, a fifth entity and at least one sixth entity.
[0325] The second entity is mainly used to determine service requirements.
[0326] Exemplarily, the second entity may determine the service requirements by receiving a user's request; the second entity may also perceive changes in scenario requirements and determine the service requirements by analyzing the perceived data; the second entity may also analyze current data / third-party data and historical data through an AI algorithm to determine the service requirements.
[0327] The second entity may be referred to as a Requirement Analysing Function (RAF) entity.
[0328] The first entity may be used to determine an execution operation list of a target service according to a service requirement.
[0329] The first entity may be referred to as a Network Adaptability Function (NAF) entity.
[0330] The third entity may be used to coordinate the execution progress, functions and / or resources of at least one sixth entity, etc.; and may also be used to determine an execution operation list of a target service.
[0331] The third entity may be referred to as a Network Collaboration Function (NCF) entity.
[0332] The fourth entity may be used to register network resources and receive updates on the status of network function resources.
[0333] The fourth entity may be referred to as a Network-resource Registration and Collection Function (NRCF) entity.
[0334] The fifth entity may be used to generate QoS indicator parameters and store QoS indicator parameters generated by other network elements.
[0335] The fifth entity may be referred to as a Network Service Policy Function (NSPF) entity.
[0336] Each sixth entity can be used to implement a type of function, such as control plane network function, user plane network function, data plane network function, computing network function, etc. That is, the sixth entity can perform a single network service operation, such as connection establishment, data calculation, etc.; it can also perform collaborative operations to complete the provision of complex services, such as subnet establishment, AI joint training, etc. In other words, the execution operation of a service can be performed by a single sixth entity or by multiple sixth entities in collaboration.
[0337] The sixth entity may be referred to as a Network Execution Function (NEF) entity.
[0338] The first entity can be deployed in an access network or a core network; the second entity can be a first terminal device, or the second entity can be deployed in a second terminal device, an access network or a core network; the third entity can be a third terminal device, or the third entity can be deployed in a fourth terminal device, an access network or the core network; the fourth entity can be deployed in an access network or a core network; the fifth entity can be deployed in an access network or a core network; each sixth entity can be a fifth terminal device, or each sixth entity can be deployed in a sixth terminal device, an access network or a core network.
[0339] When the second entity is the first terminal device, the first terminal device has all the functions of the second entity, and all the functions of the second entity can be implemented by software.
[0340] The second entity is deployed in the second terminal device, which may mean that the second entity can be deployed in the second terminal device as a hardware device. The second terminal device has all the functions of the second entity, and all the functions of the second entity can be implemented by hardware or by a combination of software and hardware.
[0341] When the third entity is a third terminal device, the third terminal device has all functions of the third entity, and all functions of the third entity can be implemented through software.
[0342] The third entity being deployed in the fourth terminal device may mean that the third entity may be deployed in the fourth terminal device as a hardware device, the fourth terminal device has all the functions of the third entity, and all the functions of the third entity may be implemented through hardware or through a combination of software and hardware.
[0343] When the sixth entity is the fifth terminal device, the fifth terminal device has all the functions of the sixth entity, and all the functions of the sixth entity can be realized by software.
[0344] The sixth entity being deployed in the sixth terminal device may mean that the sixth entity may be deployed in the sixth terminal device as a hardware device, the sixth terminal device has all the functions of the sixth entity, and all the functions of the sixth entity may be implemented through hardware or through a combination of software and hardware.
[0345] In a possible implementation, the first entity, the second entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity may all be deployed in the access network, may all be deployed in the core network, or may be respectively deployed in at least two of the terminal device, the access network device and the core network device.
[0346] In other words, the first entity is deployed on the first access network device or the first core network device; the second entity is the first terminal device or the second entity is deployed on the second terminal device, the second access network device or the second core network device; the third entity is the third terminal device or the third entity is deployed on the fourth terminal device, the third access network device or the third core network device; the fourth entity is deployed on the fourth access network device or the fourth core network device; the fifth entity is deployed on the fifth access network device or the fifth core network device; each sixth entity is the fifth terminal device or each sixth entity is deployed on the sixth terminal device, the sixth access network device or the sixth core network device; wherein the first terminal device, the third terminal device and the fifth terminal device are the same or different, the second terminal device and the fourth terminal device are the same or different from the sixth terminal device, the first access network device, the second access network device, the third access network device, the fourth access network device, the fifth access network device and the sixth access network device are the same or different, and the first core network device, the second core network device, the third core network device, the fourth core network device, the fifth core network device and the sixth core network device are the same or different.
[0347] For example, the second entity may be a terminal device, and the first entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity may be respectively deployed in different devices of the core network.
[0348] As another example, the second entity may be deployed in a terminal device, and the first entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity may be respectively deployed in different devices of the core network.
[0349] As another example, the second entity may be deployed in an access network, and the first entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity may be respectively deployed in different devices of the core network.
[0350] As another example, the first entity, the second entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity may be respectively deployed in different devices of the core network.
[0351] As another example, the second entity may be deployed in a terminal device, the first entity and the fifth entity may be deployed in different devices of an access network, and the third entity, the fourth entity and at least one sixth entity may be respectively deployed in different devices of a core network.
[0352] As another example, the second entity may be a terminal device, the first entity and the sixth entity 1 may be deployed simultaneously in the same device of the access network, and the third entity, the fourth entity, the fifth entity and the sixth entity 2 may be deployed respectively in different devices of the core network.
[0353] As another example, the second entity can be deployed in the terminal device 1, the sixth entity 2 can be deployed in the terminal device 2, the first entity, the fourth entity, and the sixth entity 1 can be respectively deployed in different devices of the access network, and the third entity, the fifth entity, and the sixth entity 3 can be respectively deployed in different devices of the core network.
[0354] When the above-mentioned entity is deployed in an access network, it may refer to that the above-mentioned entity is deployed in a base station; when the above-mentioned entity is deployed in a core network, it may refer to that the above-mentioned entity is deployed on existing functional entities such as a network function (NF) entity and an operation administration and maintenance (OAM) entity, or the above-mentioned entity may be deployed as a separate device in the core network.
[0355] The architecture of the communication system of the embodiment of the present application includes not only the above-mentioned entities, but also other functional entities, such as PCF entity, AMF entity, SMF entity and / or UPF entity, which is not limited to the embodiment of the present application.
[0356] Based on the above communication system architecture, the following describes in detail how multiple entities in the architecture interact to achieve on-demand provision of model training services.
[0357] The model training in the embodiments of the present application may refer to AI model training.
[0358] Figure 3 A flow chart of a model training method provided in an embodiment of the present application. Figure 3 As shown, the method includes:
[0359] S301. The second entity sends a model training request message to the first entity.
[0360] In other words, the first entity receives the model training request message sent by the second entity.
[0361] The second entity may send a model training request message to the first entity when it is determined that the user, network or third party has a model training requirement.
[0362] Exemplarily, the second entity may determine the model training requirements by receiving a request sent by a user; it may also determine the model training requirements by analyzing perception data; it may also determine the model training requirements by analyzing current data and historical data through an AI algorithm, where the current data and historical data may be device data, network data, or third-party data.
[0363] In a possible implementation, the model training request message may include: an identifier of the second entity, multiple indication parameters of the model, a service type of the target service, and demand information of the target service.
[0364] The identifier of the second entity may be an identity identifier of the second entity.
[0365] Exemplarily, if the second entity is a terminal device, the identifier of the second entity can be a user permanent identifier (Subscription Permanent Identifier, SUPI), a user hidden identifier (Subscription concealed identifier, SUCI), an access and mobility management function terminal application layer signaling protocol identifier (Access and Mobility Management Function User Equipment NG Application Protocol Identifier, AMF UE NGAP ID) or a radio access network terminal application layer signaling protocol identifier (Radio Access Network User Equipment NG Application Protocol Identifier, RAN UE NGAP ID).
[0366] Multiple indication parameters may be used to indicate the current service phase and the next service phase.
[0367] For example, the multiple indication parameters may include a current service procedure indication parameter (Service Procedure Indicator, SPI) and a next-step service procedure indication parameter (Next-step Service Procedure Indicator, NSPI); wherein, the SPI may be used to indicate the current service stage, and the NSPI may be used to indicate the next service stage.
[0368] The current service stage may refer to the operation that the communication system should currently perform in the process of providing model training services, and the next service stage may refer to the operation that the communication system should perform next in the process of providing target services.
[0369] The communication system can go through the following stages in the entire process of providing model training services: Demand Analysis
[0370] (Require Analysing, RA) stage, network adaptation (Network Adaptability, NA) stage, network collaboration (Network Collaboration, NC) stage, network execution (Network Execution, NE) stage, and network execution response (Network Execution Response, NER) stage.
[0371] Among them, the RA stage is mainly to determine the service requirements; the NA stage is mainly to determine the execution operation list; the NC stage is mainly to coordinate the execution progress, functions and / or resources of at least one sixth entity, and determine at least one execution operation of each sixth entity; the NE stage is mainly to execute at least one execution operation; and the NER stage is mainly to respond to the execution operation.
[0372] The service phases indicated by SPI and NSPI may be the same or different. For example, SPI and NSPI may indicate different service phases through parameter values shown in the following table:
[0373]
[0374] The current service phase indicated by the multiple indication parameters in the model training request message may be the RA phase, and the next service phase indicated may be the NA phase.
[0375] If any message includes multiple indication parameters, the entity device that receives the corresponding message needs to first determine the service stage that the entity device currently needs to perform based on the multiple indication parameters, and then perform the operations that should be completed in the corresponding service stage.
[0376] The service type of the target service in the model training request message is model training.
[0377] In one possible implementation, the target service requirement information in the model training request message may include at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0378] The first model accuracy, the first model complexity, the first model inference overhead, and the first model generalization may be expected model data.
[0379] The source of training data can be the device itself, the network, or a third party.
[0380] The algorithm source can be the device itself, the network, or a third party.
[0381] If the second entity is a terminal device, the second entity needs to establish a connection with the network, determine the service requirements, and then send a model training request message to the first entity.
[0382] S302: The first entity determines model training related information according to the model training request message.
[0383] In one possible implementation, the model training related information may include at least one of the following: resource type, resource size, training algorithm, and training method.
[0384] Resources may include computing resources, storage resources, etc.
[0385] Exemplarily, after receiving the model training request message, the first entity can determine the computing resource type and computing resource size based on the training data volume, training duration, and first model complexity in the model training request message; can determine the storage resource size based on the training data volume; can determine the training algorithm based on the algorithm source, the first model inference overhead (optional) and the model generalization (optional); can determine the training method based on the first model complexity, the training data volume, the training data source, the training algorithm, the computing resource type (optional) and the computing resource size (optional).
[0386] Training methods can include centralized training, distributed training or joint training.
[0387] In one possible implementation, model training related information can be used to determine a target QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training.
[0388] In one possible implementation, the target QoS indicator parameter list may include the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list may include the target QoS indicator parameter of the third entity.
[0389] In a possible implementation manner, the execution operation list includes the execution operation of the third entity and the execution operation of at least one sixth entity, or the execution operation list includes the execution operation of the third entity.
[0390] In one possible implementation, the first entity may determine the target QoS indicator parameters of the third entity and the target QoS indicator parameters of at least one sixth entity based on information related to model training; or the target QoS indicator parameters of the third entity may be determined based on information related to model training, and then the third entity may determine the target QoS indicator parameters of at least one sixth entity; or the model training information may be sent to the third entity, and the third entity may determine the target QoS indicator parameters of the third entity and the target QoS indicator parameters of at least one sixth entity.
[0391] The target QoS indicator parameter list can be used to determine the execution operation list.
[0392] In a possible implementation, any of the above target QoS indicator parameters includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0393] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0394] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0395] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0396] The synergy indicator parameters include synergy type.
[0397] The service index parameter may refer to the index parameter that the final trained model should achieve; or may refer to the conditions that need to be met during the expected model reasoning process.
[0398] For example, the second model accuracy can be the model accuracy that the final trained model should achieve; the second model generalization can be the model generalization that the final trained model should achieve; the second model complexity can be the model complexity that the final trained model should achieve; the second model inference overhead can be the expected overhead numerical requirement in the model inference process.
[0399] The target QoS indicator parameters of different entities may be the same or different. When the model training does not require the collaboration of multiple sixth entities, the target QoS indicator parameters of the third entity may be the same as the target indicator parameters of the sixth entity; when the model training requires the collaboration of multiple sixth entities, the target QoS indicator parameters of the third entity may be different from the target indicator parameters of the sixth entity, wherein the target QoS indicator parameters of the third entity may refer to the target QoS indicator parameters of the entire model, and the target QoS indicator parameters of each sixth entity may refer to the target QoS indicator parameters of part of the model.
[0400] Exemplarily, if the training of model 1 requires the collaboration of two sixth entities, where the sixth entity 1 is trained to obtain part A of model 1, and the sixth entity 2 is trained to obtain part B of model 1, then the target QoS indicator parameters of the third entity may refer to the target QoS indicator parameters of model 1, the target QoS indicator parameters of the sixth entity 1 may refer to the target QoS indicator parameters of part A of model 1, and the target QoS indicator parameters of the sixth entity 2 may refer to the target QoS indicator parameters of part B of model 1.
[0401] Among them, the same target QoS indicator parameters of different entities may mean that the types and parameter values of specific parameters in the target QoS indicator parameters are exactly the same; different target QoS indicator parameters of different entities may mean that the types and / or parameter values of specific parameters in the target QoS indicator parameters are different.
[0402] S303. The first entity sends a first indication message to the third entity.
[0403] In other words, the third entity receives the first indication message sent by the first entity.
[0404] The first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list.
[0405] In a possible implementation, the first indication message may also include a service identifier of the model training, multiple indication parameters, and an identifier of the anchor network element, wherein the multiple indication parameters are used to indicate the current service stage and the next service stage.
[0406] The service identifier of model training may be a unique identifier assigned by the first entity to the target model training so as to distinguish it from other model training services or other intelligent endogenous services. The target model training may refer to the model training requested by the second entity.
[0407] The anchor network element may refer to a third entity or may refer to at least one sixth entity.
[0408] The first indication message may also include a Media Access Control (MAC) address of the anchor network element.
[0409] In one possible implementation, if the target QoS indicator parameter list only includes the target QoS indicator parameters of the third entity (i.e., does not include the target QoS indicator parameters of at least one sixth entity), and the execution operation list only includes the execution operations of the third entity (i.e., does not include the execution operations of at least one sixth entity), the first indication message may also include the identifier of at least one sixth entity.
[0410] The current service phase indicated by the multiple indication parameters in the first indication message may be the NA phase, and the next service phase indicated may be the NC phase.
[0411] When the model training service does not require multiple sixth entities to execute collaboratively (ie, the number of sixth entities is one), the multiple indication parameters in the first indication message may not indicate the NC stage.
[0412] S304: The third entity sends an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list.
[0413] In other words, each sixth entity receives the execution operation request message sent by the third entity.
[0414] The execution operation request message can be used to instruct at least one sixth entity to perform model training.
[0415] In one possible implementation, the execution operation request message may include at least one of the following: a service identifier for model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate the current service stage and the next service stage.
[0416] The target sixth entity is any one of the at least one sixth entity.
[0417] The identifier of the target sixth entity may be an execution operation identifier of the target sixth entity.
[0418] The current service phase indicated by the multiple indication parameters in the execution operation request message may be the NC phase, and the next service phase indicated may be the NE phase.
[0419] The third entity may send the execute operation request message to at least one sixth entity at the same time, or may send the execute operation request message to at least one sixth entity in sequence.
[0420] In one possible implementation, if the target QoS indicator parameter list includes the target QoS indicator parameters of at least one sixth entity, and the execution operation list includes the execution operation of at least one sixth entity, the target QoS indicator parameters of at least one sixth entity in the target QoS indicator parameter list and the execution operation in the execution operation list can be separated and sent to each corresponding sixth entity through an execution operation request message.
[0421] In one possible implementation, if the target QoS indicator parameter list only includes the target QoS indicator parameters of the third entity (i.e., does not include the target QoS indicator parameters of at least one sixth entity), and the execution operation list only includes the execution operations of the third entity (i.e., does not include the execution operations of at least one sixth entity), then the target QoS indicator parameters of at least one sixth entity and the execution operations of at least one sixth entity can be determined based on the target QoS indicator parameters of the third entity and the execution operations of the third entity, and then sent to each corresponding sixth entity via an execution operation request message.
[0422] Figure 3 Only the third entity is shown sending an execution operation request message to two sixth entities. When the model training service does not require multiple sixth entities to execute collaboratively (that is, the number of sixth entities is one), the third entity can send an execution operation request message to only one sixth entity. At this time, the multiple indication parameters in the execution operation request message may not indicate the NC stage.
[0423] S305. Each sixth entity performs model training according to the execution operation request message to obtain a model training result.
[0424] After receiving the execution operation request message, the sixth entity can first determine the service stage (NE stage) it should perform based on multiple indication parameters, and then perform model training based on the execution operation of the target sixth entity in the execution operation request message, the amount of training data, the source of training data (optional) and the target QoS indicator parameters of the target sixth entity to obtain the model training results.
[0425] If the model training is performed collaboratively by multiple sixth entities, the model training result of each sixth entity may refer to a partial result of the entire model training and may be used to determine the result of the entire model training.
[0426] If the model training does not require the collaboration of multiple sixth entities (ie, there is only one sixth entity), the model training result of the sixth entity may refer to the result of the entire model training.
[0427] S306. Each sixth entity sends an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0428] In other words, the third entity sends an execute operation response message to at least one sixth entity.
[0429] In a possible implementation, the execution operation response message may also include a service identifier for model training, an identifier of a target sixth entity, and multiple indication parameters.
[0430] Each sixth entity may send an execute operation response message.
[0431] The current service phase indicated by the multiple indication parameters in the execution operation response message may be the NE phase, and the next service phase indicated may be the NER phase.
[0432] S307. The third entity sends a first response message to the first entity, where the first response message includes a model training result.
[0433] In other words, the first entity receives the first response message sent by the third entity.
[0434] The model training result in the first response message is the result of the entire model training.
[0435] In a possible implementation, the first response message also includes a service identifier for model training and multiple indication parameters.
[0436] The current service phase indicated by the multiple indication parameters in the first response message may be the NER phase, and the next service phase indicated may be the NER phase.
[0437] S308. The first entity sends a second response message to the second entity, where the second response message includes a model training result.
[0438] In other words, the second entity receives the second response message sent by the first entity.
[0439] The model training result in the second response message is the result of the entire model training.
[0440] In a possible implementation, the second response message also includes a service identifier for model training and multiple indication parameters.
[0441] The current service phase indicated by the multiple indication parameters in the second response message may be the NER phase, and the next service phase indicated may be the NER phase.
[0442] Figure 3The illustrated embodiment constructs a closed-loop model training process through the interaction between the first entity, the second entity, the third entity and at least one sixth entity, so that the communication system can intelligently provide model training services to users, networks or third parties based on model training requirements during the communication process; at the same time, QoS indicator parameters suitable for the model training service are also determined during the model training process. Compared with previous QoS indicator parameters, the QoS indicator parameters determined in the embodiment of the present application can meet the QoS requirements of the model training service.
[0443] Figure 3 The embodiment shown illustrates the basic process of providing model training services by the communication system. Figure 3 Based on the embodiment shown below, combined with Figures 4 to 6 This section describes in detail how each parameter in the model training service is determined.
[0444] Figure 4 A flowchart of another model training method provided in an embodiment of the present application. Figure 4 As shown, the method includes:
[0445] S401. The second entity sends a model training request message to the first entity.
[0446] S402: The first entity determines model training related information according to the model training request message.
[0447] It should be noted that the execution process from S401 to S402 can refer to the execution process from S301 to S302, which will not be repeated here.
[0448] S403. The first entity obtains a target QoS indicator parameter list according to the model training related information, where the target QoS indicator parameter list includes the target QoS indicator parameters of the third entity and the target QoS indicator parameters of at least one sixth entity.
[0449] In a possible implementation, the first entity may obtain a target QoS indicator parameter list according to the model training related information in the following manner:
[0450] Determine a candidate QoS indicator parameter list based on relevant information of model training; send the candidate QoS indicator parameter list to the fifth entity, the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; receive the target QoS indicator parameter list or confirmation message sent by the fifth entity, the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0451] The candidate QoS indicator parameter list includes candidate QoS indicator parameters of the third entity and candidate QoS indicator parameters of at least one sixth entity.
[0452] The candidate QoS indicator parameter list may be the same as the target QoS indicator parameter list, or may be different from the target QoS indicator parameter list. The only difference is that the specific values of the parameters in the two lists are different.
[0453] After receiving the candidate QoS indicator parameter list sent by the first entity, the fifth entity can determine whether the candidate QoS indicator parameter list needs to be adjusted in combination with global data (network data such as bandwidth, computing resource usage, and storage resource usage, etc.); if not, a confirmation message can be sent to the first entity to indicate that the candidate QoS indicator parameter list is available, and the candidate QoS indicator parameter list can be saved as a target QoS indicator parameter list; if so, the candidate QoS indicator parameter list can be adjusted in combination with the global data to obtain the target QoS indicator parameter list, and the target QoS indicator parameter list can be sent to the first entity, and the target QoS indicator parameter list can be saved at the same time.
[0454] By verifying whether the candidate QoS indicator parameter list needs to be adjusted by the fifth entity, not only can the model training service be ensured to proceed smoothly, but also mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by verifying whether the candidate QoS indicator parameter list needs to be adjusted by the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0455] In a possible implementation, the first entity may obtain a target QoS indicator parameter list based on model training related information in the following manner: send model training related information to the fifth entity, and receive the target QoS indicator parameter list sent by the fifth entity.
[0456] After the fifth entity receives the information related to model training, it can first query whether there is a QoS indicator parameter list corresponding to the information related to model training in the QoS indicator parameter list stored in itself. If so, the QoS indicator parameter list corresponding to the information related to model training is determined as the target QoS indicator parameter list and sent to the first entity; if not, the target QoS indicator parameter list is determined based on the information related to model training and global data, and the target QoS indicator parameter list is sent to the first entity, and the target QoS indicator parameter list is saved at the same time.
[0457] When the fifth entity determines the QoS indicator parameter list corresponding to the model training related information as the target QoS indicator parameter list, it can combine the global data to determine whether the QoS indicator parameter list corresponding to the model training related information can be used as the target QoS indicator parameter list. If not, it can adjust the above QoS indicator parameter list to the target QoS indicator parameter list in combination with the global data.
[0458] By determining the target candidate QoS indicator parameter list through the fifth entity, not only can the model training service be ensured to proceed smoothly, but also mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by determining the target QoS indicator parameter list through the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0459] S404: The first entity obtains network resource information from the fourth entity.
[0460] The network resource information may be used to indicate load conditions of multiple third entities and multiple sixth entities in the network.
[0461] S405: The first entity determines an execution operation list according to the target QoS indicator parameter list and the network resource information, where the execution operation list includes execution operations of the third entity and at least one execution operation of the sixth entity.
[0462] The execution operation of each entity can be one or more; the execution operation may include task calculation, data processing, AI training and / or AI reasoning, etc.
[0463] The first entity may determine execution operations of multiple entities according to a target QoS indicator parameter list, determine a suitable third entity and at least one sixth entity according to network resource information, and then assign execution operations of the multiple entities to the third entity and at least one sixth entity.
[0464] S406: The first entity sends a first indication message to the third entity, where the first indication message includes a target QoS indicator parameter list and an execution operation list.
[0465] S407: The third entity sends an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list.
[0466] S408. Each sixth entity performs model training according to the execution operation request message to obtain a model training result.
[0467] S409. Each sixth entity sends an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0468] S410. The third entity sends a first response message to the first entity, where the first response message includes a model training result.
[0469] S411. The first entity sends a second response message to the second entity, where the second response message includes a model training result.
[0470] It should be noted that the execution process of S406 to S411 can refer to the execution process of S303 to S308, and will not be repeated here.
[0471] exist Figure 4 In the illustrated embodiment, a closed-loop model training process is constructed through the interaction between the first entity, the second entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity, so that the communication system can intelligently provide model training services to users, networks or third parties based on model training requirements during the communication process; at the same time, during the model training process, the first entity and the fifth entity also determine the QoS indicator parameters suitable for the model training service. Compared with previous QoS indicator parameters, the QoS indicator parameters determined in the embodiment of the present application can meet the QoS requirements of the model training service.
[0472] Figure 5 A flowchart of another model training method provided in an embodiment of the present application. Figure 5 As shown, the method includes:
[0473] S501. The second entity sends a model training request message to the first entity.
[0474] S502: The first entity determines model training related information according to the model training request message.
[0475] S503: The first entity sends a first indication message to the third entity, where the first indication information includes information related to model training.
[0476] It should be noted that the execution process of S501 to S503 can refer to the execution process of S301 to S303, which will not be repeated here.
[0477] S504: The third entity obtains a target QoS indicator parameter list according to the model training related information, where the target QoS indicator parameter list includes the target QoS indicator parameters of the third entity and the target QoS indicator parameters of at least one sixth entity.
[0478] In a possible implementation, the third entity may obtain the target QoS indicator parameter list according to the model training related information in the following manner:
[0479] Determine a candidate QoS indicator parameter list based on relevant information of model training; send the candidate QoS indicator parameter list to the fifth entity, the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; receive the target QoS indicator parameter list or confirmation message sent by the fifth entity, the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0480] After receiving the candidate QoS indicator parameter list sent by the third entity, the fifth entity can determine whether the candidate QoS indicator parameter list needs to be adjusted in combination with the global data. If not, it can send a confirmation message to the third entity to indicate that the candidate QoS indicator parameter list is available and save the candidate QoS indicator parameter list as the target QoS indicator parameter list; if so, it can adjust the candidate QoS indicator parameter list in combination with the global data to obtain the target QoS indicator parameter list, and send the target QoS indicator parameter list to the third entity, while saving the target QoS indicator parameter list.
[0481] By verifying whether the candidate QoS indicator parameter list needs to be adjusted by the fifth entity, not only can the model training service be ensured to proceed smoothly, but also mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by verifying whether the candidate QoS indicator parameter list needs to be adjusted by the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0482] In a possible implementation, the third entity may obtain the target QoS indicator parameter list based on the model training related information in the following manner: send the model training related information to the fifth entity, and receive the target QoS indicator parameter list sent by the fifth entity.
[0483] After the fifth entity receives the information related to model training, it can first query whether there is a QoS indicator parameter list corresponding to the information related to model training in the QoS indicator parameter list stored in itself. If so, the QoS indicator parameter list corresponding to the information related to model training is determined as the target QoS indicator parameter list and sent to the third entity; if not, the target QoS indicator parameter list is determined based on the information related to model training and global data, and the target QoS indicator parameter list is sent to the third entity, and the target QoS indicator parameter list is saved at the same time.
[0484] When the fifth entity determines the QoS indicator parameter list corresponding to the model training related information as the target QoS indicator parameter list, it can combine the global data to determine whether the QoS indicator parameter list corresponding to the model training related information can be used as the target QoS indicator parameter list. If not, it can adjust the above QoS indicator parameter list to the target QoS indicator parameter list in combination with the global data.
[0485] By determining the target candidate QoS indicator parameter list through the fifth entity, not only can the model training service be ensured to proceed smoothly, but also mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by determining the target QoS indicator parameter list through the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0486] S505: The third entity obtains network resource information from the fourth entity.
[0487] S506: The third entity determines an execution operation list according to the target QoS indicator parameter list and the network resource information. The execution operation list includes execution operations of the third entity and at least one execution operation of a sixth entity.
[0488] It should be noted that the execution process of S505 to S506 can refer to the execution process of S404 to S405, and will not be repeated here.
[0489] S507: The third entity sends an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list.
[0490] S508. Each sixth entity performs model training according to the execution operation request message to obtain a model training result.
[0491] S509. Each sixth entity sends an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0492] S510. The third entity sends a first response message to the first entity, where the first response message includes a model training result.
[0493] S511. The first entity sends a second response message to the second entity, where the second response message includes a model training result.
[0494] It should be noted that the execution process of S507 to S511 can refer to the execution process of S304 to S308, and will not be repeated here.
[0495] Figure 5 The illustrated embodiment constructs a closed-loop model training process through the interaction between the first entity, the second entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity, so that the communication system can intelligently provide model training services to users, networks or third parties based on model training requirements during the communication process; at the same time, during the model training process, the third entity and the fifth entity also determine the QoS indicator parameters suitable for the model training service. Compared with previous QoS indicator parameters, the QoS indicator parameters determined in the embodiment of the present application can meet the QoS requirements of the model training service.
[0496] Figure 6 A flowchart of another model training method provided in an embodiment of the present application. Figure 6 As shown, the method includes:
[0497] S601. The second entity sends a model training request message to the first entity.
[0498] S602: The first entity determines model training related information according to the model training request message.
[0499] S603: The first entity obtains a target indicator parameter list according to the model training related information, where the target QoS indicator parameter list includes the target QoS indicator parameters of the third entity.
[0500] S604: The first entity obtains network resource information from the fourth entity.
[0501] S605: The first entity determines an execution operation list according to the target QoS indicator parameter list and the network resource information, where the execution operation list includes execution operations of the third entity.
[0502] It should be noted that the execution process of S603 to S605 can refer to the execution process of S403 to S405, which will not be repeated here.
[0503] S606: The first entity sends a first indication message to the third entity, where the first indication message includes a target QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity.
[0504] S607: The third entity obtains a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and an identifier of at least one sixth entity.
[0505] In a possible implementation manner, the third entity may obtain the target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity in the following manner:
[0506] Determine at least one candidate QoS indicator parameter based on the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity; send at least one candidate QoS indicator parameter to the fifth entity, and the at least one candidate QoS indicator parameter is used to determine the target QoS indicator parameter of at least one sixth entity; receive the target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity, and the confirmation message is used to indicate that the at least one candidate QoS indicator parameter is the target QoS indicator parameter of at least one sixth entity.
[0507] The candidate QoS indicator parameter may refer to the candidate QoS indicator parameter of the sixth entity
[0508] After receiving at least one candidate QoS indicator parameter sent by the third entity, the fifth entity can determine whether the at least one candidate QoS indicator parameter needs to be adjusted in combination with the global data. If not, it can send a confirmation message to the third entity to indicate that the at least one candidate QoS indicator parameter is available, and save the at least one candidate QoS indicator parameter as the target QoS indicator parameter of at least one sixth entity. If so, it can adjust the at least one candidate QoS indicator parameter in combination with the global data to obtain the target QoS indicator parameter of at least one sixth entity, and send the at least one target QoS indicator parameter of the sixth entity to the third entity, while saving the target QoS indicator parameter of at least one sixth entity.
[0509] By verifying whether at least one candidate QoS indicator parameter needs to be adjusted by the fifth entity, not only can the model training service be ensured to proceed smoothly, but also mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by verifying whether at least one candidate QoS indicator parameter needs to be adjusted by the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0510] In one possible implementation, the third entity may obtain the target QoS indicator parameters of at least one sixth entity based on the target QoS indicator parameters of the third entity and the identifier of at least one sixth entity in the following manner: sending the target QoS indicator parameters of the third entity and the identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameters of the third entity are used to determine the target QoS indicator parameters of at least one sixth entity; and receiving the target QoS indicator parameters of at least one sixth entity sent by the fifth entity.
[0511] After the fifth entity receives the target QoS indicator parameters of the third entity and the identifier of at least one sixth entity, it can determine the target QoS indicator parameters of at least one sixth entity based on the target QoS indicator parameters of the third entity, the identifier of at least one sixth entity and global data, and send the target QoS indicator parameters of at least one sixth entity to the third entity, while saving the target QoS indicator parameters of at least one sixth entity.
[0512] By determining the target candidate QoS indicator parameters of at least one sixth entity through the fifth entity, not only can the model training service be ensured to proceed smoothly, but also the mutual interference between multiple intelligent endogenous services in the communication system can be avoided. That is, by determining the target QoS indicator parameters of at least one sixth entity through the fifth entity, it can be ensured that multiple intelligent endogenous services in the communication system can proceed smoothly.
[0513] S608: The third entity obtains network resource information from the fourth entity.
[0514] S609: The third entity determines an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity.
[0515] It should be noted that the execution process of S608 to S609 can refer to the execution process of S404 to S405, and will not be repeated here.
[0516] S610: The third entity sends an execution operation request message to each sixth entity.
[0517] S611. Each sixth entity performs model training according to the execution operation request message to obtain a model training result.
[0518] S612. Each sixth entity sends an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0519] S613. The third entity sends a first response message to the first entity, where the first response message includes a model training result.
[0520] S614. The first entity sends a second response message to the second entity, where the second response message includes a model training result.
[0521] It should be noted that the execution process of S611 to S614 can refer to the execution process of S305 to S308, which will not be repeated here.
[0522] exist Figure 6 In the illustrated embodiment, a closed-loop model training process is constructed through the interaction between the first entity, the second entity, the third entity, the fourth entity, the fifth entity and at least one sixth entity, so that the communication system can intelligently provide model training services to users, networks or third parties based on model training requirements during the communication process; at the same time, during the model training process, the first entity, the third entity and the fifth entity also determine the QoS indicator parameters suitable for the model training service. Compared with previous QoS indicator parameters, the QoS indicator parameters determined in the embodiment of the present application can meet the QoS requirements of the model training service.
[0523] Based on any of the above embodiments, Figure 7 Describe how the third entity updates the model of at least one sixth entity, and coordinates the execution progress, functions and / or resources of at least one sixth entity. Since the interaction process between the third entity and at least one sixth entity is the same, Figure 7 Only the interaction process between the third entity and one sixth entity is shown.
[0524] Figure 7 This is a flow chart of an interactive method provided in an embodiment of the present application. Figure 7As shown, the method includes:
[0525] S701. The sixth entity sends feedback information to the third entity.
[0526] The feedback information may include information such as the execution status and resources of the sixth entity.
[0527] The sixth entity may send feedback information to the third entity after receiving the execution operation request message; may also send feedback information to the third entity during the execution process; the sixth entity may also periodically send feedback information to the third entity.
[0528] S702: The third entity sends a second instruction message to the sixth entity based on the feedback information, where the second instruction message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment.
[0529] The third entity can receive feedback information sent by multiple sixth entities that execute different target services, coordinate resources of the multiple sixth entities based on the feedback information, and can indicate the update model of the multiple sixth entities based on the feedback information, and send a second indication message to the multiple sixth entities to instruct the multiple sixth entities to cooperate in coordination or update.
[0530] The third entity may also receive feedback information sent by multiple sixth entities executing the same target service, coordinate the execution progress, functions and resources of the multiple sixth entities based on the feedback information, and may instruct the update models of the multiple sixth entities based on the feedback information, and send a second indication message to the multiple sixth entities to instruct the multiple sixth entities to cooperate in collaboration or update.
[0531] S703. The sixth entity sends a third response message to the third entity.
[0532] The third response message may be used to indicate agreement to collaboration, completion of collaboration, or rejection of collaboration.
[0533] After receiving the second indication message, the sixth entity may send a third response message to the third entity, and then the sixth entity may perform model update or resource coordination adjustment according to the second indication message.
[0534] The sixth entity may also send a third response message to the third entity after updating the model or performing resource coordination adjustment according to the second indication message.
[0535] Figure 8 Schematic diagram of the structure of the model training device 800 provided in the embodiment of the present application. Figure 8 As shown, the device 800 includes: a memory 810, a transceiver 820, and a processor 830.
[0536] The memory 810 is used to store computer programs; the transceiver 820 is used to send and receive data under the control of the processor 830; the processor 830 is used to read the computer program stored in the memory 810 and perform the following operations:
[0537] Receiving a model training request message sent by the second entity;
[0538] According to the model training request message, determine the model training related information, the model training related information is used to determine the target service quality QoS indicator parameter list and the execution operation list, the execution operation list is used to indicate the execution operations of multiple entities in the model training;
[0539] Sending a first indication message to a third entity, where the first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list;
[0540] Receive a first response message sent by the third entity, the first response message including the model training result
[0541] A second response message is sent to the second entity, where the second response message includes the model training result.
[0542] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0543] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0544] In one implementation, the first indication message includes a target QoS indicator parameter list and an execution operation list; the processor 830 is further configured to perform the following operations:
[0545] According to the model training related information, obtain the target QoS indicator parameter list;
[0546] obtaining network resource information from a fourth entity;
[0547] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0548] In one implementation, the processor 830 is specifically configured to perform the following operations:
[0549] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0550] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0551] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0552] In one implementation, the processor 830 is specifically configured to perform the following operations:
[0553] Sending model training related information to the fifth entity;
[0554] Receive the target QoS indicator parameter list sent by the fifth entity.
[0555] In one embodiment, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
[0556] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0557] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0558] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0559] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0560] The synergy indicator parameters include synergy type.
[0561] In one implementation, the first indication message further includes a service identifier of the model training, a plurality of indication parameters, and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage;
[0562] The first response message also includes a service identifier and a plurality of indication parameters;
[0563] The second response message also includes a service identifier and multiple indication parameters.
[0564] Among them, Figure 8In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 830 and various circuits of memory represented by memory 810 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 820 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 830 is responsible for managing the bus architecture and general processing, and the memory 810 may store data used by the processor 830 when performing operations.
[0565] Optionally, the processor 830 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD), and the processor may also adopt a multi-core architecture.
[0566] It should be noted here that the above-mentioned model training device 800 provided in the present application can implement all the method steps implemented by the first entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0567] Fig.9A A structural diagram of a model training device 900 provided in an embodiment of the present application, such as Fig.9A As shown, the device 900 includes: a memory 910 , a transceiver 920 and a processor 930 .
[0568] The memory 910 is used to store computer programs; the transceiver 920 is used to send and receive data under the control of the processor 930; the processor 930 is used to read the computer program stored in the memory 910 and perform the following operations:
[0569] Receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training;
[0570] Sending an operation execution request message to at least one sixth entity according to the target QoS indicator parameter list and the operation execution list, where the operation execution request message is used to instruct the at least one sixth entity to perform model training;
[0571] receiving an execution operation response message sent by at least one sixth entity;
[0572] A first response message is sent to the first entity, where the first response message includes a model training result.
[0573] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0574] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0575] The execution operation response message includes a service identifier of the model training, an identifier of the target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
[0576] In one implementation, the first indication message includes information related to model training; the processor 930 is further configured to perform the following operations:
[0577] According to the model training related information, obtain the target QoS indicator parameter list;
[0578] obtaining network resource information from a fourth entity;
[0579] Determine the execution operation list based on the target QoS indicator parameter list and network resource information.
[0580] In one implementation, the processor 930 is specifically configured to perform the following operations:
[0581] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0582] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0583] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0584] In one implementation, the processor 930 is specifically configured to perform the following operations:
[0585] Sending model training related information to the fifth entity;
[0586] Receive the target QoS indicator parameter list sent by the fifth entity.
[0587] In one implementation, the target QoS indicator parameter list includes a target QoS indicator parameter of a third entity and a target QoS indicator parameter of at least one sixth entity.
[0588] In one implementation, the first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes an execution operation of the third entity; the processor 930 is further configured to perform the following operations:
[0589] Acquire a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0590] obtaining network resource information from a fourth entity;
[0591] Determining an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity;
[0592] An execution operation request message is sent to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
[0593] In one implementation, the processor 930 is specifically configured to perform the following operations:
[0594] Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0595] Sending at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of at least one sixth entity;
[0596] A target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that at least one candidate QoS indicator parameter is a target QoS indicator parameter of at least one sixth entity.
[0597] In one implementation, the processor 930 is specifically configured to perform the following operations:
[0598] Sending a target QoS indicator parameter of the third entity and an identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity;
[0599] Receive at least one target QoS indicator parameter of a sixth entity sent by the fifth entity.
[0600] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0601] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0602] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0603] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0604] The synergy indicator parameters include synergy type.
[0605] In one implementation, the processor 930 is further configured to perform the following operations:
[0606] receiving feedback information from at least one sixth entity;
[0607] Sending a second instruction message to at least one sixth entity according to the feedback information, where the second instruction message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0608] A third response message sent by at least one sixth entity is received.
[0609] Fig. 9B Another structural diagram of the model training device 900 provided in the embodiment of the present application. Fig. 9B As shown, when the model training device 900 is a terminal device, the device may also include a user interface 940. For different terminal devices, the user interface 940 may also be an interface that can be connected to required external or internal devices. The connected devices include but are not limited to a keypad, display, speaker, microphone, joystick, etc.
[0610] Among them, Fig.9A and Fig. 9BIn the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 930 and various circuits of memory represented by memory 910 are linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 920 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 930 is responsible for managing the bus architecture and general processing, and the memory 910 may store data used by the processor 930 when performing operations.
[0611] Optionally, the processor 930 may be a central processing unit CPU, ASIC, FPGA or CPLD, and the processor may also adopt a multi-core architecture.
[0612] The processor 930 is used to execute all method steps of the third entity of the embodiment of the present application according to the obtained executable instructions by calling the computer program stored in the memory 910. The processor 930 and the memory 910 can also be arranged physically separately.
[0613] It should be noted here that the above-mentioned model training device 900 provided in the present application can implement all the method steps implemented by the third entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0614] Fig. 10A A structural diagram of a model training device 1000 provided in an embodiment of the present application, such as Fig. 10A As shown, the device 1000 includes: a memory 1010 , a transceiver 1020 and a processor 1030 .
[0615] The memory 1010 is used to store computer programs; the transceiver 1020 is used to send and receive data under the control of the processor 1030; the processor 1030 is used to read the computer program stored in the memory 1010 and perform the following operations:
[0616] Receiving an execution operation request message sent by a third entity;
[0617] According to the execution operation request message, the model training is performed to obtain the model training result;
[0618] An execution operation response message is sent to the third entity, where the execution operation response message includes the model training result.
[0619] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0620] The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity, and multiple indication parameters.
[0621] In one implementation, the processor 1030 is further configured to perform the following operations:
[0622] Sending feedback information to a third entity;
[0623] receiving a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update the model or to perform resource collaborative adjustment;
[0624] A third response message is sent to the third entity.
[0625] Fig. 10B Another structural diagram of the model training device 1000 provided in the embodiment of the present application. Fig. 10B As shown, when the model training device 1000 is a terminal device, the device may also include a user interface 1040. For different terminal devices, the user interface 1040 may also be an interface that can be connected to required external or internal devices. The connected devices include but are not limited to a keypad, display, speaker, microphone, joystick, etc.
[0626] Among them, Fig. 10A and Fig. 10B In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 1030 and various circuits of memory represented by memory 1010 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 1020 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 1030 is responsible for managing the bus architecture and general processing, and the memory 1010 may store data used by the processor 1030 when performing operations.
[0627] Optionally, the processor 1030 may be a central processing unit CPU, ASIC, FPGA or CPLD, and the processor may also adopt a multi-core architecture.
[0628] The processor 1030 is used to execute all method steps of the sixth entity of the embodiment of the present application according to the obtained executable instructions by calling the computer program stored in the memory 1010. The processor 1030 and the memory 1010 can also be arranged physically separately.
[0629] It should be noted here that the above-mentioned model training device 1000 provided in the present application can implement all the method steps implemented by the sixth entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0630] Fig.11A A structural diagram of a model training device 1100 provided in an embodiment of the present application, such as Fig.11A As shown, the model training device 1100 includes: a memory 1110, a transceiver 1120 and a processor 1130.
[0631] The memory 1110 is used to store computer programs; the transceiver 1120 is used to send and receive data under the control of the processor 1130; the processor 1130 is used to read the computer program stored in the memory 1110 and perform the following operations:
[0632] Sending a model training request message to the first entity;
[0633] A second response message sent by the first entity is received, where the second response message includes a model training result.
[0634] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0635] In one embodiment, the second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
[0636] Fig. 11B Another structural diagram of the model training device 1100 provided in the embodiment of the present application. Fig. 11B As shown, when the model training device 1100 is a terminal device, the device may also include a user interface 1140. For different terminal devices, the user interface 1140 may also be an interface that can be connected to required external or internal devices. The connected devices include but are not limited to a keypad, display, speaker, microphone, joystick, etc.
[0637] Among them, Fig.11A and Fig. 11BIn the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 1130 and various circuits of memory represented by memory 1110 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 1120 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 1130 is responsible for managing the bus architecture and general processing, and the memory 1110 may store data used by the processor 1130 when performing operations.
[0638] Optionally, the processor 1130 may be a central processing unit CPU, ASIC, FPGA or CPLD, and the processor may also adopt a multi-core architecture.
[0639] The processor 1130 is used to execute all method steps of the second entity of the embodiment of the present application according to the obtained executable instructions by calling the computer program stored in the memory 1110. The processor 1130 and the memory 1110 can also be arranged physically separately.
[0640] It should be noted here that the above-mentioned model training device 1100 provided in the present application can implement all the method steps implemented by the second entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0641] Fig.12 Schematic diagram of the structure of the model training device 1200 provided in the embodiment of the present application. Fig.12 As shown, the device 1200 includes:
[0642] The first receiving unit 1210 is used to receive a model training request message sent by the second entity;
[0643] A first determining unit 1220 is used to determine model training related information according to the model training request message, where the model training related information is used to determine a target service quality QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in the model training;
[0644] The first sending unit 1230 is configured to send a first indication message to a third entity, where the first indication message includes information related to model training, or the first indication message includes a target QoS indicator parameter list and an execution operation list;
[0645] The second receiving unit 1240 is used to receive a first response message sent by the third entity, where the first response message includes a model training result
[0646] The second sending unit 1250 is used to send a second response message to the second entity, where the second response message includes a model training result.
[0647] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0648] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0649] In one implementation, the first indication message includes a target QoS indicator parameter list and an execution operation list; the apparatus 1200 further includes:
[0650] A first acquisition unit is used to acquire a target QoS indicator parameter list according to model training related information;
[0651] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0652] The second determining unit is used to determine the execution operation list according to the target QoS indicator parameter list and the network resource information.
[0653] In one implementation, the first acquisition unit is specifically configured to:
[0654] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0655] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0656] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0657] In one implementation, the first acquisition unit is specifically configured to:
[0658] Sending model training related information to the fifth entity;
[0659] Receive the target QoS indicator parameter list sent by the fifth entity.
[0660] In one embodiment, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
[0661] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0662] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0663] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0664] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0665] The synergy indicator parameters include synergy type.
[0666] In one implementation, the first indication message further includes a service identifier of the model training, a plurality of indication parameters, and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage;
[0667] The first response message also includes a service identifier and a plurality of indication parameters;
[0668] The second response message also includes a service identifier and multiple indication parameters.
[0669] It should be noted here that the above-mentioned device 1200 provided in the present application can implement all the method steps implemented by the first entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0670] Fig.13 Schematic diagram of the structure of the model training device 1300 provided in the embodiment of the present application. Fig.13 As shown, the device 1300 includes:
[0671] The first receiving unit 1310 is used to receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training;
[0672] A first sending unit 1320 is configured to send an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list, where the execution operation request message is used to instruct the at least one sixth entity to perform model training;
[0673] The second receiving unit 1330 is configured to receive an execution operation response message sent by at least one sixth entity;
[0674] The second sending unit 1340 is used to send a first response message to the first entity, where the first response message includes a model training result.
[0675] In one embodiment, the model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
[0676] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0677] The execution operation response message includes a service identifier of the model training, an identifier of the target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
[0678] In one implementation, the first indication message includes model training related information, and the apparatus 1300 further includes:
[0679] A first acquisition unit is used to acquire a target QoS indicator parameter list according to model training related information;
[0680] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0681] The determination unit is used to determine the execution operation list according to the target QoS indicator parameter list and network resource information.
[0682] In one implementation, the first acquisition unit is specifically configured to:
[0683] Determine a list of candidate QoS indicator parameters based on relevant information from model training;
[0684] Sending a candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine a target QoS indicator parameter list;
[0685] A target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
[0686] In one implementation, the first acquisition unit is specifically configured to:
[0687] Sending model training related information to the fifth entity;
[0688] Receive the target QoS indicator parameter list sent by the fifth entity.
[0689] In one implementation, the target QoS indicator parameter list includes a target QoS indicator parameter of a third entity and a target QoS indicator parameter of at least one sixth entity.
[0690] In one implementation, the first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes the execution operation of the third entity;
[0691] The first acquisition unit is further configured to acquire a target QoS indicator parameter of at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0692] A second acquiring unit, configured to acquire network resource information from a fourth entity;
[0693] The determining unit is further configured to determine an execution operation of each sixth entity according to a target QoS indicator parameter and network resource information of at least one sixth entity;
[0694] The first sending unit is further used to send an execution operation request message to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
[0695] In one implementation, the first acquisition unit is specifically configured to:
[0696] Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of at least one sixth entity;
[0697] Sending at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of at least one sixth entity;
[0698] A target QoS indicator parameter of at least one sixth entity or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that at least one candidate QoS indicator parameter is a target QoS indicator parameter of at least one sixth entity.
[0699] In one implementation, the first acquisition unit is specifically configured to:
[0700] Sending a target QoS indicator parameter of the third entity and an identifier of at least one sixth entity to the fifth entity, where the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity;
[0701] Receive at least one target QoS indicator parameter of a sixth entity sent by the fifth entity.
[0702] In one embodiment, the target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter;
[0703] The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead;
[0704] The resource indicator parameters include at least one of the following: resource type, resource size, resource priority, and resource usage duration;
[0705] The transmission indicator parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth;
[0706] The synergy indicator parameters include synergy type.
[0707] In one embodiment, the apparatus 1300 further includes:
[0708] A third receiving unit, configured to receive feedback information from at least one sixth entity;
[0709] A third sending unit, configured to send a second indication message to at least one sixth entity according to the feedback information, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0710] The fourth receiving unit is used to receive a third response message sent by at least one sixth entity.
[0711] It should be noted here that the above-mentioned device 1300 provided in the present application can implement all the method steps implemented by the third entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0712] Fig.14 Schematic diagram of the structure of the model training device 1400 provided in the embodiment of the present application. Fig.14 As shown, the device 1400 includes:
[0713] The first receiving unit 1410 is configured to receive an execution operation request message sent by a third entity;
[0714] The model training unit 1420 is used to perform model training according to the execution operation request message and obtain a model training result;
[0715] The first sending unit 1430 is used to send an execution operation response message to the third entity, where the execution operation response message includes a model training result.
[0716] In one embodiment, the execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage;
[0717] The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity, and multiple indication parameters.
[0718] In one embodiment, the apparatus 1400 further includes:
[0719] A second sending unit, configured to send feedback information to a third entity;
[0720] A second receiving unit, used to receive a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment;
[0721] The third sending unit is used to send a third response message to the third entity.
[0722] It should be noted here that the above-mentioned device 1400 provided in the present application can implement all the method steps implemented by the sixth entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0723] Fig.15 Schematic diagram of the structure of the model training device 1500 provided in the embodiment of the present application. Fig.15 As shown, the device comprises:
[0724] The sending unit 1510 is used to send a model training request message to the first entity;
[0725] The receiving unit 1520 is used to receive a second response message sent by the first entity, where the second response message includes a model training result.
[0726] In one embodiment, the model training request message includes at least one of the following: training time, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
[0727] In one embodiment, the second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
[0728] It should be noted here that the above-mentioned device 1500 provided in the present application can implement all the method steps implemented by the second entity in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.
[0729] It should be noted that the division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0730] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.
[0731] An embodiment of the present application further provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute all method steps of the first entity in the above method embodiment.
[0732] An embodiment of the present application further provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute all method steps of the third entity in the above method embodiment.
[0733] An embodiment of the present application further provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute all method steps of the sixth entity in the above method embodiment.
[0734] An embodiment of the present application further provides a non-transitory readable storage medium, which stores a computer program, and the computer program is used to enable a processor to execute all method steps of the second entity in the above method embodiment.
[0735] Non-transitory readable storage media can be any available media or data storage devices that can be accessed by a computer, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)), etc.
[0736] An embodiment of the present application also provides a computer program product, including a computer program, which implements the method described in any one of the above method embodiments when the computer program is executed by a processor.
[0737] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.
[0738] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0739] These processor executable instructions may also be stored in a processor readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0740] These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0741] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A model training method, characterized in that: Applied to a first entity, the method comprises: Receiving a model training request message sent by the second entity; Determine, according to the model training request message, model training related information, the model training related information is used to determine a target quality of service QoS indicator parameter list and an execution operation list, the execution operation list is used to indicate the execution operations of multiple entities in the model training; Sending a first indication message to a third entity, where the first indication message includes the model training related information, or the first indication message includes the target QoS indicator parameter list and the execution operation list; Receive a first response message sent by the third entity, where the first response message includes a model training result A second response message is sent to the second entity, where the second response message includes a model training result.
2. The method according to claim 1, characterized in that The model training request message includes at least one of the following: training duration, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
3. The method according to claim 1 or 2, characterized in that: The model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
4. The method according to claim 3, characterized in that The first indication message includes the target QoS indicator parameter list and the execution operation list; Before sending the first indication message to the third entity, the method further includes: Obtain a target QoS indicator parameter list based on the model training related information; obtaining network resource information from a fourth entity; The execution operation list is determined according to the target QoS indicator parameter list and the network resource information.
5. The method according to claim 4, characterized in that According to the model training related information, a target QoS indicator parameter list is obtained, including: Determine a list of candidate QoS indicator parameters based on the model training related information; Sending the candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; The target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
6. The method according to claim 4, characterized in that According to the model training related information, a target QoS indicator parameter list is obtained, including: Sending the model training related information to a fifth entity; Receive the target QoS indicator parameter list sent by the fifth entity.
7. The method according to claim 5 or 6, characterized in that: The target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
8. The method according to claim 7, characterized in that The target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter; The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead; The resource indicator parameter includes at least one of the following: resource type, resource size, resource priority, and resource usage duration; The transmission index parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth; The collaboration indicator parameter includes a collaboration type.
9. The method according to claim 1, characterized in that: The first indication message further includes a service identifier of the model training, a plurality of indication parameters and an identifier of the anchor network element, wherein the plurality of indication parameters are used to indicate a current service stage and a next service stage; The first response message also includes the service identifier and the multiple indication parameters; The second response message also includes the service identifier and the multiple indication parameters.
10. The method according to claim 8, characterized in that The first entity is deployed in a first access network device or a first core network device; The second entity is a first terminal device or the second entity is deployed in a second terminal device, a second access network device or a second core network device; The third entity is a third terminal device or the third entity is deployed in a fourth terminal device, a third access network device or a third core network device; The fourth entity is deployed in a fourth access network device or a fourth core network device; The fifth entity is deployed in a fifth access network device or a fifth core network device; Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device; The first terminal device, the third terminal device and the fifth terminal device are the same or different, the second terminal device, the fourth terminal device and the sixth terminal device are the same or different, the first access network device, the second access network device, the third access network device, the fourth access network device, the fifth access network device and the sixth access network device are the same or different, and the first core network device, the second core network device, the third core network device, the fourth core network device, the fifth core network device and the sixth core network device are the same or different.
11. A model training method, characterized in that: Applied to a third entity, the method comprises: Receiving a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training; Sending an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list, wherein the execution operation request message is used to instruct the at least one sixth entity to perform model training; receiving an execution operation response message sent by the at least one sixth entity; A first response message is sent to the first entity, where the first response message includes a model training result.
12. The method according to claim 11, characterized in that The model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
13. The method according to claim 11 or 12, characterized in that: The execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage; The execution operation response message includes a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
14. The method according to any one of claims 11 to 13, characterized in that: The first indication message includes information related to model training; Before sending the execution operation request message to at least one sixth entity, the method further includes: According to the model training related information, obtain the target QoS indicator parameter list; obtaining network resource information from a fourth entity; The execution operation list is determined according to the target QoS indicator parameter list and the network resource information.
15. The method according to claim 14, characterized in that According to the model training related information, the target QoS indicator parameter list is obtained, including: Determine a list of candidate QoS indicator parameters based on the model training related information; Sending the candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; The target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
16. The method according to claim 14, characterized in that According to the model training related information, the target QoS indicator parameter list is obtained, including: Sending the model training related information to a fifth entity; Receive the target QoS indicator parameter list sent by the fifth entity.
17. The method according to any one of claims 11 to 16, characterized in that: The target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of the at least one sixth entity.
18. The method according to any one of claims 11 to 13, characterized in that: The first indication message includes a target quality of service QoS indicator parameter list, an execution operation list and an identifier of the at least one sixth entity, the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity, and the execution operation list includes the execution operation of the third entity; Sending an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list includes: Acquire a target QoS indicator parameter of the at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of the at least one sixth entity; obtaining network resource information from a fourth entity; Determining an execution operation of each sixth entity according to a target QoS indicator parameter of the at least one sixth entity and the network resource information; The execution operation request message is sent to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
19. The method according to claim 18, characterized in that Obtaining a target QoS indicator parameter of the at least one sixth entity includes: Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of the at least one sixth entity; Sending the at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of the at least one sixth entity; Receive the target QoS indicator parameter of the at least one sixth entity or a confirmation message sent by the fifth entity, wherein the confirmation message is used to indicate that the at least one candidate QoS indicator parameter is the target QoS indicator parameter of the at least one sixth entity.
20. The method according to claim 18, characterized in that Obtaining a target QoS indicator parameter of the at least one sixth entity includes: Sending a target QoS indicator parameter of the third entity and an identifier of the at least one sixth entity to the fifth entity, wherein the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity; Receive a target QoS indicator parameter of the at least one sixth entity sent by the fifth entity.
21. The method according to claim 13, 17 or 18, characterized in that The target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter; The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead; The resource indicator parameter includes at least one of the following: resource type, resource size, resource priority, and resource usage duration; The transmission index parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth; The collaboration indicator parameter includes a collaboration type.
22. The method according to any one of claims 11 to 21, characterized in that: The method further comprises: receiving feedback information from the at least one sixth entity; Sending a second indication message to the at least one sixth entity according to the feedback information, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment; A third response message sent by the at least one sixth entity is received.
23. The method according to claim 15, 16, 19 or 20, characterized in that: The first entity is deployed in a first access network device or a first core network device; The third entity is a third terminal device or the third entity is deployed in a fourth terminal device, a third access network device or a third core network device; The fifth entity is deployed in a fifth access network device or a fifth core network device; Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device; The third terminal device is the same as or different from the fifth terminal device, the fourth terminal device is the same as or different from the sixth terminal device, the first access network device, the third access network device, the fifth access network device and the sixth access network device are the same as or different, and the first core network device, the third core network device, the fifth core network device and the sixth core network device are the same as or different.
24. A model training method, characterized in that: Applied to a sixth entity, the method comprises: Receiving an execution operation request message sent by a third entity; Performing model training according to the execution operation request message to obtain a model training result; Send an execution operation response message to the third entity, where the execution operation response message includes the model training result.
25. The method according to claim 24, characterized in that The execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage; The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity and multiple indication parameters.
26. The method according to any one of claims 23 to 25, characterized in that: The method further comprises: Sending feedback information to the third entity; receiving a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update a model or to perform collaborative resource adjustment; A third response message is sent to the third entity.
27. The method according to claim 24, characterized in that The third entity is a third terminal device or the third entity is deployed in a fourth terminal device, a third access network device or a third core network device; Each sixth entity is a fifth terminal device or each sixth entity is deployed in a sixth terminal device, a sixth access network device or a sixth core network device; The third terminal device is the same as or different from the fifth terminal device, the fourth terminal device is the same as or different from the sixth terminal device, the third access network device is the same as or different from the sixth access network device, and the third core network device is the same as or different from the sixth core network device.
28. A model training method, characterized in that: Applied to the second entity, the method comprises: Sending a model training request message to the first entity; Receive a second response message sent by the first entity, where the second response message includes a model training result.
29. The method according to claim 28, characterized in that The model training request message includes at least one of the following: training duration, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
30. The method according to claim 28 or 29, characterized in that The second response message also includes a service identifier for model training and multiple service process indication parameters, where the multiple indication parameters are used to indicate a current service stage and a next service stage.
31. The method according to claim 28, characterized in that The first entity is deployed in a first access network device or a first core network device; The second entity is a first terminal device or the second entity is deployed in a second terminal device, a second access network device or a second core network device; The first access network device is the same as or different from the second access network device, and the first core network device is the same as or different from the second core network device.
32. A model training device, characterized in that: include: A first receiving unit, configured to receive a model training request message sent by a second entity; A determination unit, configured to determine model training related information according to the model training request message, wherein the model training related information is used to determine a target quality of service QoS indicator parameter list and an execution operation list, wherein the execution operation list is used to indicate execution operations of multiple entities in the model training; A first sending unit, configured to send a first indication message to a third entity, where the first indication message includes the model training related information, or the first indication message includes the target QoS indicator parameter list and the execution operation list; The second receiving unit is configured to receive a first response message sent by the third entity, wherein the first response message includes a model training result The second sending unit is used to send a second response message to the second entity, where the second response message includes a model training result.
33. A model training device, characterized in that: include: A first receiving unit is used to receive a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training; A first sending unit, configured to send an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list, wherein the execution operation request message is used to instruct the at least one sixth entity to perform model training; A second receiving unit, configured to receive an execution operation response message sent by the at least one sixth entity; The second sending unit is used to send a first response message to the first entity, where the first response message includes a model training result.
34. A model training device, characterized in that: include: A receiving unit, configured to receive an execution operation request message sent by a third entity; A model training unit, used to perform model training according to the execution operation request message to obtain a model training result; A sending unit is used to send an execution operation response message to the third entity, and the execution operation response message includes the model training result.
35. A model training device, characterized in that: include: A sending unit, configured to send a model training request message to the first entity; A receiving unit is used to receive a second response message sent by the first entity, where the second response message includes a model training result.
36. A model training device, characterized in that: Including memory, transceiver, processor: The memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: Receiving a model training request message sent by the second entity; Determine, according to the model training request message, model training related information, the model training related information is used to determine a target quality of service QoS indicator parameter list and an execution operation list, the execution operation list is used to indicate the execution operations of multiple entities in the model training; Sending a first indication message to a third entity, where the first indication message includes the model training related information, or the first indication message includes the target QoS indicator parameter list and the execution operation list; Receive a first response message sent by the third entity, where the first response message includes a model training result A second response message is sent to the second entity, where the second response message includes a model training result.
37. The device according to claim 36, characterized in that The model training request message includes at least one of the following: training duration, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
38. The device according to claim 36 or 37, characterized in that The model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
39. The device according to claim 38, characterized in that The first indication message includes the target QoS indicator parameter list and the execution operation list; the processor is specifically configured to perform the following operations: Obtain a target QoS indicator parameter list based on the model training related information; obtaining network resource information from a fourth entity; The execution operation list is determined according to the target QoS indicator parameter list and the network resource information.
40. The device according to claim 39, characterized in that The processor is also used to perform the following operations: Determine a list of candidate QoS indicator parameters based on the model training related information; Sending the candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; The target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
41. The device according to claim 39, characterized in that The processor is also used to perform the following operations: Sending the model training related information to a fifth entity; Receive the target QoS indicator parameter list sent by the fifth entity.
42. The device according to claim 40 or 41, characterized in that The target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of at least one sixth entity, or the target QoS indicator parameter list includes the target QoS indicator parameter of the third entity.
43. The device according to claim 42, characterized in that The target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter; The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead; The resource indicator parameter includes at least one of the following: resource type, resource size, resource priority, and resource usage duration; The transmission index parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth; The collaboration indicator parameter includes a collaboration type.
44. A model training device, characterized in that: Including memory, transceiver, processor: The memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: Receiving a first indication message sent by a first entity, where the first indication message includes information related to model training, or the first indication message includes a target quality of service QoS indicator parameter list and an execution operation list, where the execution operation list is used to indicate execution operations of multiple entities in model training; Sending an execution operation request message to at least one sixth entity according to the target QoS indicator parameter list and the execution operation list, wherein the execution operation request message is used to instruct the at least one sixth entity to perform model training; receiving an execution operation response message sent by the at least one sixth entity; A first response message is sent to the first entity, where the first response message includes a model training result.
45. The device according to claim 44, characterized in that The model training related information includes at least one of the following: resource type, resource size, training algorithm, and training method.
46. The device according to claim 44 or 45, characterized in that The execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage; The execution operation response message includes a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, and a model training result of the target sixth entity.
47. The device according to any one of claims 44 to 46, characterized in that The first indication message includes model training related information; the processor is further configured to perform the following operations: According to the model training related information, obtain the target QoS indicator parameter list; obtaining network resource information from a fourth entity; The execution operation list is determined according to the target QoS indicator parameter list and the network resource information.
48. The device according to claim 47, characterized in that The processor is also used to perform the following operations: Determine a list of candidate QoS indicator parameters based on the model training related information; Sending the candidate QoS indicator parameter list to a fifth entity, where the candidate QoS indicator parameter list is used to determine the target QoS indicator parameter list; The target QoS indicator parameter list or a confirmation message sent by the fifth entity is received, where the confirmation message is used to indicate that the candidate QoS indicator parameter list is the target QoS indicator parameter list.
49. The device according to claim 47, characterized in that The processor is also used to perform the following operations: Sending the model training related information to a fifth entity; Receive the target QoS indicator parameter list sent by the fifth entity.
50. The device according to any one of claims 44 to 49, characterized in that The target QoS indicator parameter list includes the target QoS indicator parameter of the third entity and the target QoS indicator parameter of the at least sixth entity.
51. The device according to any one of claims 44 to 46, characterized in that The first indication message includes a target quality of service QoS indicator parameter list, an execution operation list, and an identifier of the at least one sixth entity, the target QoS indicator parameter list includes a target QoS indicator parameter of the third entity, and the execution operation list includes the execution operation of the third entity; the processor is specifically configured to perform the following operations: Acquire a target QoS indicator parameter of the at least one sixth entity according to the target QoS indicator parameter of the third entity and the identifier of the at least one sixth entity; obtaining network resource information from a fourth entity; Determining an execution operation of each sixth entity according to a target QoS indicator parameter of the at least one sixth entity and the network resource information; The execution operation request message is sent to each sixth entity, where the execution operation request message includes a target QoS indicator parameter and an execution operation of each sixth entity.
52. The device according to claim 51, characterized in that The processor is specifically configured to perform the following operations: Determining at least one candidate QoS indicator parameter according to the target QoS indicator parameter of the third entity and the identifier of the at least one sixth entity; Sending the at least one candidate QoS indicator parameter to the fifth entity, where the at least one candidate QoS indicator parameter is used to determine a target QoS indicator parameter of the at least one sixth entity; Receive the target indicator QoS parameter of the at least one sixth entity or a confirmation message sent by the fifth entity, wherein the confirmation message is used to indicate that the at least one candidate QoS indicator parameter is the target QoS indicator parameter of the at least one sixth entity.
53. The device according to claim 51, characterized in that The processor is specifically configured to perform the following operations: Sending a target QoS indicator parameter of the third entity and an identifier of the at least one sixth entity to the fifth entity, wherein the target QoS indicator parameter of the third entity is used to determine a target QoS indicator parameter of the at least one sixth entity; Receive a target QoS indicator parameter of the at least one sixth entity sent by the fifth entity.
54. The device according to claim 46, 50 or 51, characterized in that The target QoS indicator parameter includes at least one of the following: a service indicator parameter, a resource indicator parameter, a transmission indicator parameter, and a coordination indicator parameter; The service indicator parameter includes at least one of the following: second model accuracy, second model generalization, second model complexity, and second model reasoning overhead; The resource indicator parameter includes at least one of the following: resource type, resource size, resource priority, and resource usage duration; The transmission index parameters include at least one of the following: transmission reliability, transmission delay, and transmission bandwidth; The collaboration indicator parameter includes a collaboration type.
55. The device according to any one of claims 44 to 54, characterized in that The processor is also used to perform the following operations: receiving feedback information from the at least one sixth entity; Sending a second indication message to the at least one sixth entity according to the feedback information, where the second indication message is used to instruct the sixth entity to update the model or perform resource collaborative adjustment; A third response message sent by the at least one sixth entity is received.
56. A model training device, characterized in that: Including memory, transceiver, processor: The memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: Receiving an execution operation request message sent by a third entity; Performing model training according to the execution operation request message to obtain a model training result; Send an execution operation response message to the third entity, where the execution operation response message includes the model training result.
57. The device according to claim 56, characterized in that The execution operation request message includes at least one of the following: a service identifier of model training, an identifier of a target sixth entity, multiple indication parameters, an execution operation of the target sixth entity, a training data volume, a training data source, and a target service quality QoS indicator parameter of the target sixth entity, wherein the multiple indication parameters are used to indicate a current service stage and a next service stage; The execution operation response message also includes a service identifier of the model training, an identifier of the sixth entity and multiple indication parameters.
58. The device according to claim 56 or 57, characterized in that The processor is also used to perform the following operations: Sending feedback information to the third entity; receiving a second indication message sent by the third entity, where the second indication message is used to instruct the sixth entity to update a model or to perform collaborative resource adjustment; A third response message is sent to the third entity.
59. A model training device, characterized in that: Including memory, transceiver, processor: The memory is used to store a computer program; the transceiver is used to send and receive data under the control of the processor; the processor is used to read the computer program in the memory and perform the following operations: Sending a model training request message to the first entity; Receive a second response message sent by the first entity, where the second response message includes a model training result.
60. The device according to claim 59, characterized in that The model training request message includes at least one of the following: training duration, first model accuracy, first model complexity, first model reasoning overhead, training data volume, training data source, first model generalization, and algorithm source.
61. A non-transitory readable storage medium, characterized in that The non-transitory readable storage medium stores a computer program, and the computer program is used to enable a processor to execute the method according to any one of claims 1 to 10.
62. A non-transitory readable storage medium, characterized in that: The non-transitory readable storage medium stores a computer program, and the computer program is used to enable a processor to execute the method according to any one of claims 11 to 23.
63. A non-transitory readable storage medium, characterized in that The non-transitory readable storage medium stores a computer program, and the computer program is used to cause a processor to execute the method according to any one of claims 24 to 27.
64. A non-transitory readable storage medium, characterized in that The non-transitory readable storage medium stores a computer program, and the computer program is used to cause a processor to execute the method according to any one of claims 28 to 31.
Citation Information
Cited By
Model training method and apparatus, and storage medium
EP4697783A1
