WO2023138234A1

WO2023138234A1 - Model management method and apparatus, networking architecture, electronic device and storage medium

Info

Publication number: WO2023138234A1
Application number: PCT/CN2022/136416
Authority: WO
Inventors: 许晓东; 原英婷; 董辰; 韩书君; 王碧舳
Original assignee: 北京邮电大学
Priority date: 2022-01-20
Filing date: 2022-12-04
Publication date: 2023-07-27
Also published as: CN116527497A

Abstract

Provided are a model management method and apparatus, a networking architecture, an electronic device and a storage medium. The specific technical solution comprises: acquiring a model propagation request initiated by a receiving end node, and querying a storage position of a target model as a sending end node according to the model propagation request (S101); establishing at least one transmission path for transmitting the target model between the sending end node and the receiving end node (S102); and in response to the receiving end node training a pre-stored old version model into a new version model by utilizing the target model, storing model information of the new version model in a model library (S103).

Description

Model management method, device, networking architecture, electronic equipment and storage medium

technical field

The present disclosure relates to the technical field of communication, and in particular, to a model management method, device, networking architecture, electronic equipment, and storage medium.

Background technique

In the future intelligent network of all things, network nodes tend to be intelligent. The intelligentization of network nodes has led to rapid expansion of information space, and even dimensional disasters, which has exacerbated the difficulty of representing information carrying space, making it difficult to match traditional network service capabilities with high-dimensional information space. The amount of data transmitted through communication is too large, and the information business service system cannot continue to meet people's needs for complex, diverse, and intelligent information transmission. Using artificial intelligence models to encode, disseminate, and decode business information can significantly reduce the amount of data transmission in communication services and greatly improve the efficiency of information transmission. These models are relatively stable, and have reusability and dissemination. The dissemination and reuse of models will help to enhance network intelligence while reducing overhead and resource waste, forming an intelligent network with extremely intelligent nodes and a minimal network.

Different from traditional communication, the core of Intent-Driven Network communication lies in the model. Therefore, the management process of model training, dissemination, and storage is crucial to the I-Driven network communication system. There is an urgent need for a management method that can improve the capabilities of model training, operation, dissemination, and storage in the I-Driven network, as well as the network architecture. Improve the communication system’s ability to manage models, and then improve the overall communication efficiency of the communication system.

Contents of the invention

The present disclosure provides a method, device, networking architecture, electronic equipment, and storage medium for managing models in an Intent-Driven Network.

According to an aspect of the present disclosure, a model management method is provided, including:

Acquiring the model propagation request initiated by the receiving end node, and querying the storage location of the target model according to the model propagation request as the sending end node;

Establishing at least one transmission path for transmitting the target model between the sending end node and the receiving end node, for the sending end node to transmit the target model to the receiving end node;

In response to the receiving end node using the target model to train a pre-stored old version model into a new version model, storing model information of the new version model in a model library.

Optionally, the model management method further includes: after the new version model is formed, controlling the receiving end node to transmit the new version model to its cluster member nodes, and updating the model information corresponding to the new version model in the model library.

Optionally, the model management method further includes: after updating the model information of the new version model in the model library, query whether there is still a node to be updated that stores the old version model, and transmit the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate the new version model.

Optionally, after the node to be updated updates and generates the new version model, the model library updates the model information corresponding to the old version model and the model information corresponding to the new version model respectively.

Optionally, the model information includes model generation location and/or model storage location and/or model encoding and/or model function information and/or model version information.

Optionally, the step of establishing at least one transmission path in the model library specifically includes:

determining routing nodes between the sending end node and the receiving end node;

It is determined that the sending end node transmits part of the object model or all of the object model to the receiving end node.

Optionally, the model library includes one or more memories.

According to another aspect of the present disclosure, a model management device is provided, including:

The model query module is configured to acquire the model propagation request initiated by the receiving end node, and query the storage location of the target model according to the model propagation request as the sending end node;

A model transmission module configured to establish at least one transmission path for transmitting the target model between the sending node and the receiving node, so that the transmitting node transmits the target model to the receiving node;

The model training module is configured to store model information of the new version model in a model library in response to the receiving end node using the target model to train a pre-stored old version model into a new version model.

Optionally, the model management device further includes: a first update module configured to control the receiving end node to actively transmit the new version model to its cluster member nodes after the new version model is formed, and update the model information corresponding to the new version model in the model library.

Optionally, the model management device further includes: a second update module configured to query whether there is still a node to be updated that stores the old version model after updating the model information of the new version model in the model library, and transmit the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate the new version model, and update the model information corresponding to the new version model and the model information corresponding to the old version model in the model library.

The present disclosure also provides a network architecture of an intelligent simplified network, which is applied to execute the above model management method, including:

The first type of nodes, the second type of nodes, the third type of nodes and the model library establish communication connections with each other; wherein, the first type of nodes and the second type of nodes are configured to train, run and store models, and the third type of nodes are configured to run and store the models; the model library is configured to store and update model information corresponding to the models.

The present disclosure also provides an electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the model management method described in any one of the above technical solutions.

The present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to make the computer execute the model management method according to any one of the above embodiments.

The present disclosure also provides a computer program product, including a computer program, when the computer program is executed by a processor, the model management method according to any one of the above embodiments is realized.

Based on the model management method, device, networking architecture, electronic equipment, and storage medium in the above-mentioned technical solution, the disclosure improves the management ability of the model, improves the sharing of the model in the smart network, and realizes a more efficient and stable communication system.

It should be understood that what is described in this section is not intended to identify key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will be readily understood through the following description.

Description of drawings

The accompanying drawings are used to better understand the present solution, and do not constitute a limitation to the present disclosure. in:

FIG. 1 is a step diagram of a model management method in an embodiment of the present disclosure;

FIG. 2 is a schematic flowchart of a model management method in an embodiment of the present disclosure;

Fig. 3 is a functional block diagram of the first model management device in an embodiment of the present disclosure;

FIG. 4 is a functional block diagram of a second model management device in an embodiment of the present disclosure;

FIG. 5 is a functional block diagram of a third model management device in an embodiment of the present disclosure;

Fig. 6 is a diagram of a network architecture that can implement the model management method in the embodiment of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and they should be regarded as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

In the I-Driven Network, business information is mainly disseminated through the artificial intelligence model. By using the artificial intelligence model to compress the first business information to be disseminated into the second business information related to the artificial intelligence model, the data traffic in the network is greatly reduced, and the compression efficiency far exceeds the traditional compression algorithm. Wherein, the sending-end device extracts the first service information by using a pre-configured first model to obtain the second service information to be transmitted; the sending-end device transmits the second service information to the receiving-end device. The receiving end device receives the second service information, and uses the pre-configured second model to restore the second service information to obtain the third service information; the third service information restored by the second model may have a slight difference in quality compared with the original first service information, but the two are consistent in content, and the user experience is almost the same. Before the transmitting end device transmits the second service information to the receiving end device, it further includes: an update module judges whether the receiving end device needs to update the second model, and transmits a preconfigured third model to the receiving end device when it is judged that an update is required, and the receiving end device uses the third model to update the second model. Processing business information through pre-trained artificial intelligence models can significantly reduce the amount of data transmission in communication services and greatly improve the efficiency of information transmission. These models are relatively stable, and have reusability and dissemination. Model propagation and reuse will help enhance network intelligence while reducing overhead and resource waste. The model can be divided into several model slices according to different segmentation rules, the above model slices can also be transmitted between different network nodes, and the model slices can be assembled into a model. Model slices can be distributed and stored on multiple network nodes. When a network node finds that it lacks or needs to update a certain model or a certain model slice, it can make a request to the surrounding nodes that may have the slice.

Both the transmission of the business information and the transmission of the model take place in the communication network, and the communication transmission is performed based on a network protocol. The network nodes passed on the path for transmitting the service information and the model include an intelligent simplified router. The functions of the I-Driven Router include but are not limited to business information transmission, model transmission, absorbing model self-update, security protection and other functions. The transmission function of the Intelligent-Driven Router involves the transmission of business information or models from the source node to the sink node, and there are multiple paths between the source node and the sink node. The model transmission function of the Smart-Driven Router can transmit model slices. By rationally arranging model slices to take multiple paths, multiple transmission model slices can be used to improve the model transmission rate.

The present disclosure provides a model management method, as shown in Figure 1, including:

Step S101, obtain the model propagation request initiated by the receiving end node, and query the storage location of the target model as the sending end node according to the model propagation request;

Step S102, establishing at least one transmission path for transmitting the target model between the sending end node and the receiving end node, so that the sending end node transmits the target model to the receiving end node;

Step S103, in response to the receiving end node using the target model to train the pre-stored old version model into a new version model, storing the model information of the new version model in the model library.

Specifically, this embodiment provides a model management method, including processes such as model query, transmission, training, and information storage. Exemplarily, assuming that node A needs to use the models of other nodes to improve its stored model 3.1, it can initiate a model propagation request to the model library. The model library queries the model storage in its storage table according to the model propagation request of the receiving end node, that is, the storage node of the target model, and selects the node closest to the receiving end node from these model storages as the sending end node. As shown in Figure 2, based on the consideration of channel conditions, privacy protection, energy consumption, delay constraints and other factors, node B (large server) and node C (ordinary node 1) are selected to transmit model 1 and model 2 to node A respectively.

Further, after the sending end nodes B and C are determined, the transmission path between the sending end node and the receiving end node is calculated, and there may be one transmission path or multiple transmission paths. The network establishes a transmission path transmission model between the sending end nodes B and C and the receiving end node A. The transmission path may be a direct connection, or may be connected via one or more relay nodes, and the connection may include one link or multiple links. As an optional implementation, on the one hand, this step needs to determine the routing node between the sending node and the receiving node; in addition, it also needs to determine that the sending node transmits part of the target model or all target models to the receiving node. For example, if the existing model of some nodes has similar parts with the target model, then only the different parts of the model need to be transmitted to complete the model update; if some nodes do not have the target model at all, all models need to be transmitted. In this way, occupied network resources can be reduced, repeated transmission can be avoided, and communication efficiency can be improved.

After the transmission of models 1 and 2 is completed, the receiving end node A sends a signaling to the model library, and the model library records the receiving end node A as the storer of model 1 and model 2. The receiving end node A uses model synthesis technology to combine model 1 and model 2 with its own old version model 3.1 to train to generate a new version of model 3.2, and generates a number of model 3.2, and sends it to the model library for new model registration. The model library stores the model information of the new version of model 3.2. The model information includes but is not limited to the model generation location (the node that trains the model) and/or the storage location (i.e. the storage node) and/or the model code and/or the function information of the model and/or the version information of the model.

As an optional implementation, the model management method further includes: after the new version model is formed, controlling the receiving end node to transmit the new version model to its cluster member nodes, and updating the model information corresponding to the new version model in the model library. For example, after receiving node A forms a new version of model 3.2, it needs to transmit the new version of model 3.2 to its bound cluster members. When receiving node A is the cluster head, node D and node E are its cluster members, and receiving node A transmits model 3.2 to node D and node E. After receiving model 3.2, node D and node E send signaling to the model library, and the model library registers node D and node E as the storers of model 3.2.

As an optional implementation, the model management method further includes: after updating the model information of the new version model in the model library, query whether there is still a node to be updated that stores the old version model, and transmit the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate a new version model.

Exemplarily, after receiving the model information of the new version model 3.2, the model library inquires about the existence of the lower version of the same model, that is, model 3.1. Notify the storage nodes of model 3.1. For example, node F stores model 3.1. At the same time, node F is not a cluster member of node A, so it has not been updated in time. At this time, the model library lists node F as a node to be updated, and actively sends a model update notification to node F. After receiving the model update notification, node F can initiate a model propagation request to the model library to update the local model 3.1. After the model library receives the model dissemination request, the model library can select node D to transmit the part of model 3.2 to node F, for example, the part of model 3.2 that is different from model 3.1, so as to reduce the amount of model transmission and improve transmission efficiency. Node F combines the received partial model 3.2 with the local model 3.1 to obtain model 3.2, deletes the locally stored model 3.1, and sends information to the model library, the model library registers node F as the storer of model 3.2, deletes its storer record in model 3.1, that is, updates the model information of model 3.1 and model 3.2 respectively.

It should be noted that, the model library in any of the above embodiments may include a centralized storage or be composed of multiple distributed storages. This disclosure aims to manage the specific processes of model training, transmission, and storage with the assistance of the model library, so that each node can share the model and update its own model in time, so that the content transmitted by other nodes can be processed to obtain the required business information. With the assistance of the model library, the communication system can quickly query the model-related information, thereby improving the speed of model transmission, training, storage and other processes, and improving the overall communication efficiency.

The present disclosure also provides a model management device, as shown in FIG. 3 , including:

The model query module 301 is configured to obtain the model propagation request initiated by the receiving end node, and query the storage location of the target model as the sending end node according to the model propagation request;

The model transmission module 302 is configured to establish at least one transmission path for transmitting the target model between the sending end node and the receiving end node, so that the sending end node transmits the target model to the receiving end node;

The model training module 303 is configured to store the model information of the new version model in the model library 304 in response to the receiving end node using the target model to train the pre-stored old version model into a new version model.

Specifically, this embodiment provides a model management device, including processes such as model query, transmission, training, and information storage. Exemplarily, assuming that node A needs to use the models of other nodes to improve its stored model 3.1, it can initiate a model propagation request. The model query module 301 queries the model storage in its storage table according to the model propagation request of receiving node A. The model storage is the storage node of the target model, and selects the node closest to the receiving node from these model storages as the sending node. As shown in Figure 2, based on the consideration of channel conditions, privacy protection, energy consumption, delay constraints and other factors, it is assumed that node B (large server) and node C (ordinary node 1) are selected to transmit model 1 and model 2 to node A respectively.

Further, after the sending end nodes B and C are determined, the transmission path between the sending end node and the receiving end node is calculated, and there may be one transmission path or multiple transmission paths. The model transmission module 302 establishes a transmission path transmission model between the sending end nodes B and C and the receiving end node A. The transmission path may be formed by a direct connection between the sending end node and the receiving end node, or may be connected via one or more relay nodes, and the connection may include one link or multiple links. As an optional implementation, on the one hand, this step needs to determine the routing node between the sending node and the receiving node; in addition, it also needs to determine that the sending node transmits part of the target model or all target models to the receiving node. For example, if the existing model of some nodes has similar parts with the target model, then only the different parts of the model need to be transmitted to complete the model update; if some nodes do not have the target model at all, all models need to be transmitted. In this way, occupied network resources can be reduced, repeated transmission can be avoided, and communication efficiency can be improved.

After the transmission of models 1 and 2 is completed, the receiving end node A sends a signaling to the model library, and the model library records the receiving end node A as the storer of model 1 and model 2. The model training module 303 uses the model synthesis technology to combine model 1 and model 2 with the old version model 3.1 of the receiving end node A to train to generate a new version model 3.2, and generates a number of the model 3.2, and sends it to the model library for new model registration. The model library stores the model information of the new version model 3.2. The model information includes but is not limited to the model generation location (i.e. the node for training the model) and/or the storage location (i.e. the storage node) and/or the model code and/or the function information of the model and/or the version information of the model.

As an optional implementation, as shown in FIG. 4 , the model management device further includes: a first update module 305, which controls the receiving end node to transmit the new version model to its cluster member nodes after the new version model is formed, and updates the model information corresponding to the new version model in the model library. For example, after receiving node A forms a new version of model 3.2, it needs to transmit the new version of model 3.2 to its bound cluster members. When receiving node A is the cluster head, node D and node E are its cluster members, and receiving node A transmits model 3.2 to node D and node E. After receiving model 3.2, node D and node E send signaling to the model library, and the model library registers node D and node E as the storers of model 3.2.

As an optional implementation, as shown in FIG. 5 , the model management device further includes: a second update module 306. After updating the model information of the new version model in the model library, query whether there is still a node to be updated that stores the old version model, and transmit the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate a new version model.

It should be noted that the model library in any of the above embodiments may include a centralized storage memory or multiple distributed memories, which may be deployed in the core network or in the access network, and the storage range of the model library may be determined according to factors such as specific application scenarios and storage capabilities. This disclosure aims to manage the specific processes of model training, transmission, and storage with the assistance of the model library, so that each node can share the model and update its own model in time, so that the content transmitted by other nodes can be processed to obtain the required business information. With the assistance of the model library, the communication system can quickly query the model-related information, thereby improving the speed of model transmission, training, storage and other processes, and improving the overall communication efficiency.

The present disclosure also provides a network architecture of an intelligent simplified network, which is applied to implement the model management method described in any one of the above embodiments, including:

The first type of nodes, the second type of nodes, the third type of nodes and the model library establish communication connections with each other; among them, the first type of nodes and the second type of nodes are configured to train, run and store models, and the third type of nodes are configured to run and store models; the model library is configured to store and update model information corresponding to the model.

Specifically, the nodes in the network are divided according to their functions. The IDN can include large-scale servers (i.e., the first type of nodes), ordinary nodes (i.e., the second type of nodes), and deployment nodes (i.e., the third type of nodes). The large-scale servers can generate large models, which can be devices with strong computing capabilities in the network, such as cloud computing servers, cloudlet small cloud slices, etc.; ordinary nodes can be used for migration learning and training small and medium models, including but not limited to dedicated computing servers and satellites with computing capabilities deployed on the access network, and devices with strong computing capabilities in the network, such as personal Computers, smart phones, smart vehicles, smart ships, drones, etc.; deployment nodes only have the ability to perceive and collect data, but not the ability to train models, including but not limited to mobile phones, cameras, smart bracelets, smart TVs and other devices.

As shown in Figure 6, this embodiment is divided into three types of nodes according to the training capability of the model, that is, large-scale server 601, common node 602, and deployment node 603. These three types of nodes form an intelligent network architecture, and these three types of nodes are capable of running, storing, and transmitting models. Large-scale servers communicate with other large-scale servers and some ordinary nodes using wired connections. In addition, there is no direct connection between large-scale servers and other nodes. Ordinary nodes can communicate with other ordinary nodes using wired or wireless connections. Deployment nodes can communicate with other deployment nodes or ordinary nodes using wireless connections.

In the network architecture of this embodiment, large-scale servers and ordinary nodes can use locally stored data to directly train and generate models; they can also use locally stored models to generate new models through model processing technology combined with locally stored data training. Model processing technologies include but are not limited to model synthesis technology and model compression technology, such as knowledge distillation, transfer learning, stacking methods, etc.

After the model is generated, the node that generates the model will generate the model information of the model, which includes the model number, function information and version information. The node sends the model number to the model library for storage, and the model library records the node as a model contributor and a model storer. The model library can include a centralized storage memory or multiple distributed memories, which can be deployed in the core network or in the access network. The storage range of the model library can be determined according to specific application scenarios and storage capabilities.

Exemplarily, the model propagation process initiated by the model demander (i.e., the receiving end node) includes: the model library receives the model information of the new version of the model, and can query whether there is an old version of the model in the node. If it exists, it will be listed as a node to be updated, and send a model update notification to the node to be updated. model) Actively query the existing models in the model library, and initiate a propagation request for one or more required models; when a node receives a model expiration notification, it can initiate a model propagation request to request a new version of the model.

As an optional implementation, in the above technical solution, the model update notification does not need to be sent to each node to be updated, but only needs to be sent to the node to be updated as the cluster head in the bound cluster, and then the cluster head transmits the model to other cluster member nodes.

In another embodiment, the model propagation process initiated by the model generator (i.e., the sending end node) includes: after a large server or a common node generates a new model, it can directly send the model to a specific node, requiring the node to receive the new model, so that the models between the nodes remain unified. The premise of using this mode is that it needs to be agreed in advance between the nodes, for example, to form a binding cluster between the nodes, the cluster head should be the uppermost node, and when the model in the cluster head is updated, the new model will be actively transmitted to the cluster members.

After determining the sender node and the receiver node of the target model, it is necessary to select a transmission path between the sender node and the receiver node. The connection between the sending end node and the receiving end node can be a direct connection, or can be connected through one or more relay nodes; the transmission path can include one link or multiple links; you can also choose to transmit part of the target model or all of the target model to the receiving end node.

After the target model is transmitted to the receiving node, the node stores the target model. At this time, the receiving node needs to send information to the model library, and record the node as the model storer. Since a node may only store parts of a model, the model information can include which part of the model the node stores. The propagation of the model is limited by the time limit. For different versions of the model, after the time limit expires, the model library will send an expiration notification to the node storing the lower version model, suggesting that the node delete the old version model, which can reduce the memory resources occupied by the old version model, update the model of each node in time, and also improve the information processing capability of each node, thereby improving communication efficiency. For example, a node uses a new version model to extract and encode the first business information to form the second business information. If the node receiving the second business information of the node still uses the old version model, the node may not be able to restore the second business information to the third business information, or the restored quality is too low to meet the needs of users. Therefore, updating the models of each node in time is also conducive to improving user experience. It should be noted that the time limit can be set and can be determined according to actual application scenarios. After the node deletes the model, it needs to send relevant messages to the model library, and the model library will delete the model storage information. However, for the sake of security, the model contributor information will not be deleted, that is, the model library will not delete the generation location of the model when updating the model information, but only update the storage location of the model. With the assistance of the model library, you can know the generation location, storage location, function and version of each model in the entire system. When any node needs a related model, the model library can quickly search for the target model and transmit it to the model demander through the fastest transmission path. .

According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

In particular, electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are by way of example only, and are not intended to limit implementations of the disclosure described and/or claimed herein.

The device includes a computing unit, which can perform various appropriate actions and processes according to computer programs stored in a read-only memory (ROM, Read-Only Memory) or loaded from a storage unit into a random access memory (RAM, Random Access Memory). In RAM, various programs and data necessary for device operation are also stored. The computing unit, ROM, and RAM are connected to each other through a bus. An input/output (I/O, Input/Output) interface is also connected to the bus.

Multiple components in the device are connected to the I/O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as magnetic disks, optical discs, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information/data with other devices over a computer network such as the Internet and/or various telecommunication networks.

Computing units may be various general and/or special purpose processing components having processing and computing capabilities. Some examples of computing units include, but are not limited to, central processing units (CPU, Central Processing Unit), graphics processing units (GPU, Graphics Processing Unit), various dedicated artificial intelligence (AI, Artificial Intelligence) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSP, Digital Signal Processing), and any appropriate processors, controllers, microcontrollers, etc. The calculation unit executes various methods and processes described above, such as the model management method in the above embodiments. For example, in some embodiments, the model management method can be implemented as a computer software program tangibly embodied on a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and/or installed on the device via a ROM and/or a communication unit. When the computer program is loaded into RAM and executed by the computing unit, one or more steps of the model management method described above may be performed. Alternatively, in other embodiments, the computing unit may be configured to execute the model management method in any other suitable manner (eg, by means of firmware).

Various implementations of the systems and technologies described above in this paper can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGA, Field Programmable Gate Array), application specific integrated circuits (ASIC, Application Specific Integrated Circuit), application specific standard products (ASSP, Application Specific Standard Product), systems on chips (SOC, System on Chip), load programmable logic devices (CPLD, Comp lex Programmable Logic Device), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include being implemented in one or more computer programs executable and/or interpreted on a programmable system comprising at least one programmable processor, which may be a special purpose or general purpose programmable processor, capable of receiving data and instructions from and transmitting data and instructions to a storage system, at least one input device, and at least one output device.

Program codes for implementing the model management method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to processors or controllers of general-purpose computers, special purpose computers, or other programmable data processing devices, so that the program codes cause the functions/operations specified in the flowcharts and/or block diagrams to be implemented when executed by the processors or controllers. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), EPROM (Electrical Programmable Read-Only Memory or flash memory), optical fiber, Compact Disc Read-Only Memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. combination.

To provide for interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (Cathode Ray Tube, cathode ray tube) or LCD (Liquid Crystal Display, liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, voice input, or tactile input.

The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, eg, a communication network. Examples of communication networks include: Local Area Network (LAN, Local Area Network), Wide Area Network (WAN, Wide Area Network) and the Internet.

A computer system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

It should be understood that steps may be reordered, added or deleted using the various forms of flow shown above. For example, each step described in the present disclosure may be executed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution disclosed in the present disclosure can be achieved, no limitation is imposed herein.

The specific implementation manners described above do not limit the protection scope of the present disclosure. It should be apparent to those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

A model management method, characterized by comprising:

Obtain the model propagation request initiated by the receiving end node, and query the storage location of the target model as the sending end node according to the model propagation request (S101);

Establishing at least one transmission path for transmitting the target model between the sending end node and the receiving end node, so that the sending end node transmits the target model to the receiving end node (S102);

In response to the receiving end node using the target model to train the pre-stored old version model into a new version model, store the model information of the new version model in a model library (S103).
The model management method according to claim 1, further comprising: after forming the new version model, controlling the receiving end node to transmit the new version model to its cluster member nodes, and updating the model information corresponding to the new version model in the model library.
The model management method according to claim 1, further comprising: after updating the model information of the new version model in the model library, querying whether there is still a node to be updated that stores the old version model, and transmitting the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate the new version model.
The model management method according to claim 3, wherein after the node to be updated updates and generates the new version model, the model library updates the model information corresponding to the old version model and the model information corresponding to the new version model respectively.
The model management method according to claim 1, wherein the model information includes model generation location and/or model storage location and/or model coding and/or model function information and/or model version information.
The model management method according to claim 1, wherein the step of establishing at least one transmission path in the model library specifically includes:

determining routing nodes between the sending end node and the receiving end node;

It is determined that the sending end node transmits part of the object model or all of the object model to the receiving end node.
The model management method according to any one of claims 1-6, wherein the model library includes one or more storages.
A model management device, characterized by comprising:

A model query module (301), configured to obtain a model propagation request initiated by a receiving end node, and query the storage location of a target model as a sending end node according to the model propagation request;

A model transmission module (302), configured to establish at least one transmission path for transmitting the target model between the sending end node and the receiving end node, so that the sending end node transmits the target model to the receiving end node;

The model training module (303) is configured to store model information of the new version model in a model library (304) in response to the receiving end node using the target model to train a pre-stored old version model into a new version model.
The model management device according to claim 8, characterized in that the model management device further comprises: a first update module (305), configured to control the receiving end node to actively transmit the new version model to its cluster member nodes after the new version model is formed, and update the model information corresponding to the new version model in the model library.
The model management device according to claim 8, characterized in that the model management device further comprises: a second update module (306), configured to query whether there is still a node to be updated that stores the old version model after updating the model information of the new version model in the model library, and transmit the new version model to the node to be updated that stores the old version model, so as to update the old version model in the node to be updated to generate the new version model, and update the model information corresponding to the new version model and the old version model in the model library Corresponding model information.
A networking architecture of an intelligent network, characterized in that it is applied to the model management method described in any one of claims 1-7, including:

The first type of nodes (601), the second type of nodes (602), the third type of nodes (603) and the model library establish communication connections with each other; wherein, the first type of nodes (601) and the second type of nodes (602) are configured to train, run and store models, and the third type of nodes (603) are configured to run and store the models; the model library is configured to store and update model information corresponding to the models.
An electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the model management method according to any one of claims 1-7.
A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to make the computer execute the model management method according to any one of claims 1-7.
A computer program product, comprising a computer program, the computer program implements the model management method according to any one of claims 1-7 when executed by a processor.