Model deployment method and apparatus, electronic device, and readable storage medium

Abstract

The application provides a model deployment method, device, electronic equipment and readable storage medium. The model deployment method is applied to an edge computing system, the edge computing system comprises a management server, a configuration server and an edge device, and the method comprises the following steps: the configuration server calls an original model from a model database; the management server sends target configuration information to the configuration server, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information comprises category information of the edge device; the configuration server adjusts the original model according to the category information of the target device to generate model data matched with the target device; the configuration server sends the model data to the target device; and the edge device performs model deployment according to the model data. The embodiment of the application can improve the convenience of model deployment in the edge computing system.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of edge computing, and in particular to a model deployment method and device, electronic equipment and a readable storage medium. BACKGROUND

[0002] Edge computing refers to an open platform that integrates network, computing, storage and application core capabilities, and provides services at the nearest end. The application program is initiated at the edge side to produce faster network service response and meet the basic needs of real-time business, application intelligence, security and privacy protection. Since edge devices are usually distributed in different areas, the types of edge devices also have great differences, so the deployment and configuration process is relatively inconvenient. SUMMARY

[0003] Embodiments of the present application provide a model deployment method, device, electronic equipment and readable storage medium to improve the edge computing problem.

[0004] To solve the above problems, the present application is implemented as follows:

[0005] In a first aspect, the embodiments of the present application provide a model deployment method applied to an edge computing system, wherein the edge computing system includes a management server, a configuration server and edge devices, and the method includes the following steps:

[0006] The configuration server calls an original model from a model database;

[0007] The management server sends target configuration information to the configuration server, wherein the target configuration information is the configuration information of a target device in the edge device, and the configuration information includes the category information of the edge device;

[0008] The configuration server adjusts the original model according to the category information of the target device to generate model data matched with the target device;

[0009] The configuration server sends the model data to the target device;

[0010] The edge device performs model deployment according to the model data.

[0011] In some embodiments, the method further includes:

[0012] Each edge device sends the configuration information of the edge device to the management server;

[0013] The management server adds the edge device to an edge device pool according to the configuration information, wherein the target configuration information is called by the management server from the edge device pool.

[0014] In some embodiments, before the configuration server calls the original model from the model database, the method further includes:

[0015] The management server receives a service activation request, wherein the service activation request includes a service requirement;

[0016] A target device among the edge devices is determined according to the type of the service requirement and / or the area corresponding to the service requirement.

[0017] In some embodiments, the configuration server calls the original model from the model database, including:

[0018] The original model stored in the model database in a preset format is pulled to the configuration server through a pull code.

[0019] In some embodiments, the original model includes one of a torch model and a tensorflow model.

[0020] In some embodiments, the category information of the edge device includes at least one of bitmain, rockchip rknn, NVIDIA Jetson, NVIDIA RTX / GTX, X86 CPU and arm CPU.

[0021] In a second aspect, an embodiment of the present invention provides a model deployment method, which is applied to a management server in an edge computing system, wherein the edge computing system includes a management server, a configuration server, and an edge device. The method includes the following steps:

[0022] receiving a selection instruction for a target device in the edge device;

[0023] Target configuration information is sent to the configuration server according to the selection instruction, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information includes category information of the edge device.

[0024] In a third aspect, an embodiment of the present invention provides a model deployment method, which is applied to a configuration server in an edge computing system, wherein the edge computing system includes a management server, a configuration server, and an edge device. The method includes the following steps:

[0025] Call the original model from the model database;

[0026] receiving target configuration information sent by the management server, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information comprises category information of the edge device;

[0027] generating model data matched with the target device by adjusting the original model according to the category information of the target device;

[0028] sending the model data to the target device, so that the edge device performs model deployment according to the model data.

[0029] In a fourth aspect, an embodiment of the present application provides a model deployment apparatus applied to an edge computing system, wherein the edge computing system comprises a management server, a configuration server and an edge device.

[0030] The configuration server comprises:

[0031] a model calling module configured to call an original model from a model database;

[0032] The management server comprises:

[0033] an information sending module configured to send target configuration information to the configuration server, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information comprises category information of the edge device;

[0034] The configuration server further comprises:

[0035] a model data generating module configured to generate model data matched with the target device by adjusting the original model according to the category information of the target device;

[0036] a model data sending module configured to send the model data to the target device;

[0037] The edge device comprises:

[0038] a deployment module configured to perform model deployment according to the model data.

[0039] In a fifth aspect, an embodiment of the present application provides an electronic device, comprising a processor, a memory and a program stored in the memory and executable on the processor;

[0040] When the program is executed by the processor, steps of the method in any one of the preceding claims can be implemented.

[0041] In a sixth aspect, an embodiment of the present application provides a readable storage medium for storing a program, wherein the program is executed by a processor to implement steps of the method in any one of the preceding claims.

[0042] The configuration server in the technical solution of the embodiment of the present application can generate model data matched with the target device according to the category information of the target device, and send the model data to the target device, so that the edge device performs model deployment according to the model data. The cooperation of the cloud and the edge device is realized, the model conversion and adjustment can be completed at the configuration server in the cloud, and then the model is sent to the selected edge device, the management of the edge device is realized, and the convenience and reliability of model deployment and management are improved. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0044] Figure 1 is a flowchart of the model deployment method provided by the embodiment of the present application;

[0045] Figure 2 is another flowchart of the model deployment method provided by the embodiment of the present application;

[0046] Figure 3 is a structural schematic diagram of an electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0048] The terms "first", "second", and the like in the embodiments of the present application are used to distinguish similar objects, and do not necessarily mean a specific order or sequence. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those clearly listed steps or units, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices. In addition, "and / or" is used in the present application to represent at least one of the connected objects, for example, A and / or B and / or C, which represents 7 cases including A alone, B alone, C alone, A and B, B and C, A and C, and A, B and C.

[0049] The embodiment of the present application provides a model deployment method.

[0050] In one embodiment, the method is applied to an edge computing system, and the edge computing system comprises a management server, a configuration server and edge devices.

[0051] In some embodiments, an edge device pool is established first.

[0052] The method for establishing the edge device pool comprises the following steps:

[0053] Each of the edge devices sends configuration information of the edge device to the management server;

[0054] The management server adds the edge device to the edge device pool according to the configuration information.

[0055] In the technical scheme of the embodiment, the edge device can be added to the edge device pool in different ways.

[0056] In one exemplary embodiment, a plurality of edge devices can be deployed in the same machine room, so that when an edge device is added to the local area network of the machine room, the edge information of the edge device can be obtained, and the configuration information of the edge device is sent to the management server, so that the edge device is included in the edge device pool for unified management.

[0057] In another embodiment, the edge device can be registered through the management interface provided by the management server, so that the edge device is included in the management of the edge device pool.

[0058] In implementation, the address information (for example, hardware mac address) and category information of the device and other information related to the edge device can be written into the configuration information of the edge device, and then the edge device is added to the edge device pool.

[0059] After the edge device pool is established, the edge devices in the edge device pool can be managed through the management server or the configuration server, and subsequently, when a target device in the device is determined and the target configuration information of the target device needs to be called, the corresponding target configuration information can be called from the established edge device pool.

[0060] As shown in Figure 1 and Figure 2 In one embodiment, the model deployment method comprises the following steps:

[0061] Step 101: The configuration server calls an original model from a model database.

[0062] In the technical solution of the embodiment, the model saved in the model database is an original model, which can be understood as an initial model that has not been modified. Since the configurations of hardware and software of different edge devices are different, and the tasks to be performed by the edge devices are also different, the original model cannot be directly deployed on the edge devices, but must be adjusted accordingly before being deployed on the edge devices to perform corresponding work tasks.

[0063] In the embodiment, a model database is established in advance. For example, the model database can be a gitlab model management database. In an embodiment, the original model includes a torch model and a tensorflow model. Obviously, in some other embodiments, the original model that can be deployed or used can also be saved in the model database according to needs.

[0064] In some embodiments, the step 101 includes:

[0065] The original model saved in the model database in a preset format is pulled to the configuration server by pulling the code.

[0066] In the embodiment, the model database is taken as an example of gitlab. In implementation, the model can be pulled by pulling the code such as git pull, and the original model is pulled from the gitlab to the configuration server. The original model can be saved in the form of a specific specification model and model code, for example, in the form of pth, pb, etc.

[0067] Step 102: The management server sends target configuration information to the configuration server.

[0068] In an embodiment, before step 102, the method further includes:

[0069] The management server receives a selection instruction for a target device in the edge device.

[0070] The management server sends target configuration information to the configuration server according to the selection instruction, wherein the target configuration information is configuration information of the target device in the edge device, and the configuration information includes category information of the edge device.

[0071] In the technical solution of the embodiment, the management server and the configuration server can be the same device or different devices.

[0072] In the embodiment, the management server is used to provide management information of the edge device. In an example embodiment, a web page or a program page can be used to provide an instruction for a user to input the management information.

[0073] In some embodiments, the selection instruction includes selection of a model type, which may, for example, include specific functions required to be implemented by the face recognition, face detection, face quality detection, etc. model, and further may include selection instruction for selection of the structure of the model itself.

[0074] In some embodiments, the method further includes:

[0075] The management server receives a service opening request, wherein the service opening request includes a service requirement;

[0076] According to the type of the service requirement and / or the area corresponding to the service requirement, a target device in the edge device is determined.

[0077] It should be understood that the user in the present embodiment mainly refers to a B-end user, more specifically, can be a service provider, and the B-end user can be a bank, an industrial park, a mobile health service, a smart health service, a public service, a smart retail service, etc. The B-end user can provide corresponding services to C-end customers (for example, can be ordinary consumers). These specific service contents can be realized by relying on cloud computing services to provide stronger computing power support, and further, in the present embodiment, edge computing services are realized, which can make the services provided to the C-end customers have faster response speed.

[0078] When it is necessary to provide services to C-end customers, for example, to open new services, the B-end user can determine the service requirement through the management server, for example, through the operation interface of the web page provided by the management server. After the service requirement is determined, the required edge service is determined according to the type of the service requirement.

[0079] In the present embodiment, a more basic selection method can be provided for the B-end user, for example, the user can select the corresponding service requirement as picture processing, data processing or face recognition, etc. In this way, the management server can allocate the type of the target device according to the corresponding service requirement, for example, the edge device including a display chip type is allocated for the picture processing service, and the edge device including a central processing unit type is allocated for the data processing service. In this way, the B-end user does not need to have professional knowledge to identify different types of platform functions, only needs to put forward the service requirement, and the management server can allocate the appropriate type of edge device for the B-end user, thereby improving the convenience of the operation of the B-end user.

[0080] Further, the service requirement in the embodiment can also include area information corresponding to the service, so that the edge device more suitable for the service can be allocated according to the area information. For example, an idle edge device closer to the service can be allocated, so as to further shorten the service response time and improve the use experience of the C-end customer.

[0081] The selection instruction can also include the selection of the device. Generally, the number of edge devices in the edge computing system is multiple. In implementation, the target device for model deployment can be determined from the multiple edge devices according to the selection instruction input by the user.

[0082] For example, each edge device has a unique device name or device id (number), and the device name or device id is bound to the configuration information of the edge device. The user can select the target device in the edge device by selecting the device name or device id. Further, the target configuration information is determined through the binding relationship between the device name or device id and the configuration information, and then the target configuration information is sent to the configuration server.

[0083] In some embodiments, the configuration information includes category information of the edge device. In some embodiments, the category information of the edge device includes bitmain, rknn, NVIDIA Jetson (an embedded system), NVIDIARTX / GTX (a display chip), X86 CPU (X86 architecture central processing unit), and arm CPU (arm architecture central processing unit), etc. Other categories of edge devices can also be selected as needed in implementation.

[0084] After the target device is determined, the target configuration information of the target device is called from the edge device pool and sent to the configuration server.

[0085] Step 103: The configuration server adjusts the original model according to the category information of the target device to generate model data matched with the target device.

[0086] After obtaining the target configuration information, the configuration server can determine the category information of the target device. Further, the original model can be converted by the model conversion tool according to the category information of the target device, so as to convert the original model into model data matched with the category of the target device. More specifically, the model conversion and evaluation can be performed according to the category information of the target device, so as to generate a customized model and inference code and send them to the target device.

[0087] It needs to be understood that the original model in the model database is a model with a fixed format, however, different edge devices are different in type and different in specific tasks required to be performed, therefore, the original model cannot be directly deployed on each edge device, in the embodiment, the original model is further adjusted and converted by the configuration server according to the type of the target device in the edge device, so as to adapt to edge devices of different types and meet the differentiated requirements of the model for different tasks.

[0088] Among them, the model conversion tool can select existing or improved model conversion tools such as bmNN, RKNN, tensorRT and MNN, which are not limited here.

[0089] For example, the configuration information can be issued to the configuration server for model conversion in the format of json, and the conversion command starts to be executed, the model conversion tool can be configured through json, after receiving the json issued by the management server, update the json configuration of itself, and then start the model conversion.

[0090] Step 104: The configuration server sends the model data to the target device.

[0091] Step 105: The edge device deploys the model according to the model data.

[0092] After the configuration server completes the model conversion, the model data is sent to the target device, and after the target device receives the model data, the model is deployed through the obtained model data, so that the model that has been converted and matched with the target device can be deployed on the target device.

[0093] The configuration server in the technical scheme of the embodiment of the application can adjust the original model to generate model data matched with the target device according to the category information of the target device and send the model data to the target device, so that the edge device can deploy the model according to the model data. The cooperation of the cloud and the edge device is realized, the model conversion and adjustment can be completed at the configuration server in the cloud and then issued to the selected edge device, the management of the edge device is realized, and the convenience and reliability of the model deployment and management are improved.

[0094] It needs to be understood that different types of edge devices are suitable for performing different tasks, for example, a CPU platform is more suitable for performing floating point operation and integer operation tasks, and a display chip platform is more suitable for performing floating point operation tasks. In the related art, if a user needs to perform a certain specific task, the user needs to send an application to the administrator of the edge computing service, and then the administrator adjusts the model according to the user's needs, and then performs model deployment on the corresponding edge computing platform. After the model deployment is completed, the edge device is connected to the network to provide edge services, and then the user is notified, and the user can use the corresponding edge computing service. As can be seen, in the related art, the deployment and use process of the edge device for a specific requirement is relatively complex.

[0095] In the technical scheme of the embodiment of the application, the user can directly determine the type of the edge device according to his own needs, and then submit an application to the cloud. The configuration server configures the model according to the type of the target device and deploys the model on the corresponding edge device, thereby realizing cloud-edge coordination and improving the convenience of using the edge computing service.

[0096] The embodiment of the application provides a model deployment method, which is applied to the management server, the configuration server or the edge device, and can refer to the steps of the management server, the configuration server or the edge device in the model deployment method for implementation, and can achieve the same or similar technical effects, which will not be described here.

[0097] The embodiment of the application provides a model deployment device, which is applied to an edge computing system, and the edge computing system comprises a management server, a configuration server and an edge device.

[0098] The configuration server comprises:

[0099] The model calling module is configured to call an original model from a model database;

[0100] The management server comprises:

[0101] The information sending module is configured to send target configuration information to the configuration server, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information comprises category information of the edge device.

[0102] The configuration server further comprises:

[0103] The model data generation module is configured to adjust the original model to generate model data matched with the target device according to the category information of the target device;

[0104] The model data sending module is configured to send the model data to the target device.

[0105] The edge device comprises:

[0106] A deployment module is configured to perform model deployment according to the model data.

[0107] In some embodiments, the edge device further comprises:

[0108] A configuration information uploading module is configured to send configuration information of the edge device to the management server.

[0109] The management server comprises:

[0110] An adding module is configured to add the configuration information to an edge device pool, wherein the target configuration information is called by the management server from the edge device pool.

[0111] In some embodiments, the management server further comprises:

[0112] A service request receiving module is configured to receive a service opening request, wherein the service opening request comprises service requirements.

[0113] A device allocation module is configured to determine a target device in the edge device according to a type of the service requirements and / or a region corresponding to the service requirements.

[0114] In some embodiments, the model calling module of the configuration server is specifically configured to pull an original model saved in the model database in a preset format to the configuration server through a pull code.

[0115] The model deployment apparatus of the embodiment of the present application can realize each step of the model deployment method embodiment described above, which will not be repeated here.

[0116] The embodiment of the present application further provides an electronic device. The electronic device can be specifically the management server, the configuration server or the edge device in the edge computing system described above.

[0117] Please refer to Figure 3 , the electronic device can include a processor 301, a memory 302, and a program 3021 stored in the memory 302 and executable on the processor 301. When the program 3021 is executed by the processor 301, any step in the above method embodiments and the same beneficial effects can be achieved, which will not be repeated here.

[0118] Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiment methods can be completed by program instruction related hardware, and the program can be stored in a readable medium.

[0119] The embodiment of the present application further provides a readable storage medium, wherein the readable storage medium stores a computer program, the computer program is executed by a processor to realize any step of the above method embodiment and achieve the same technical effect, and details are not repeated here.

[0120] The storage medium is, for example, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, etc.

[0121] It should be noted that the division of the above modules is only a logical functional division, and all or part of them can be integrated into one physical entity or physically separated. These modules can all be implemented in the form of software called by a processing element; all can be implemented in the form of hardware; or part of the modules can be implemented in the form of software called by a processing element, and part of the modules can be implemented in the form of hardware. For example, the determination module can be a separately established processing element, or can be integrated into a chip of the above device, and in addition, the determination module can be stored in the form of program code in the memory of the above device, and the function of the determination module can be called and executed by a processing element of the above device. The implementation of other modules is similar. In addition, all or part of the modules can be integrated together or independently implemented. The processing element described herein can be an integrated circuit having a signal processing capability. In the implementation process, each step of the above method or each module can be completed by the integrated logic circuit of the hardware or the instruction of the software in the processing element.

[0122] For example, each module, unit, sub-unit or sub-module can be one or more integrated circuits configured to implement the above method, such as one or more Application Specific Integrated Circuits (ASICs), or one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), etc. For another example, when a certain module above is implemented in the form of scheduling program code by a processing element, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor that can call program code. For another example, these modules can be integrated together to implement in the form of system-on-a-chip (SOC).

[0123] The above is the preferred embodiment of the present application, it should be noted that for those skilled in the art, without departing from the principles of the present application, can make several improvements and refinements, these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A model deployment method, characterized in that: Applied to an edge computing system, the edge computing system includes a management server, a configuration server, and an edge device, and the method includes the following steps: The configuration server calls the original model from the model database; The management server receives a selection instruction for a target device in the edge device, and sends target configuration information to the configuration server according to the selection instruction, wherein the target configuration information is configuration information of the target device in the edge device, and the configuration information includes category information of the edge device; the category information of the edge device includes at least one of Bitmain, Rockchip RKNN, NVIDIA Jetson, NVIDIA RTX / GTX, X86 CPU, and ARM CPU; The configuration server adjusts the original model to generate model data matching the target device according to the category information of the target device; wherein the original model in the model database is a model with a fixed format, and the execution tasks of the original model when deployed on the target device with different category information are different; The configuration server sends the model data to the target device; The edge device performs model deployment according to the model data; The method further comprises: Each of the edge devices sends the configuration information of the edge device to the management server; The management server adds the edge device to an edge device pool according to the configuration information, wherein the target configuration information is called by the management server from the edge device pool; The configuration server adjusts the original model to generate model data matching the target device according to the category information of the target device, including: The configuration server converts the original model using a model conversion tool according to the category information of the target device, converting the original model into model data matching the category information of the target device; the original model includes one of a torch model and a tensorflow model; In which, the configuration information is sent to the configuration server that performs model conversion in the json format, and the model conversion tool is configured through json; after obtaining the configuration information in the json format, the configuration server updates the json configuration of the model conversion tool, and then converts the original model through the model conversion tool.

2. The method according to claim 1, wherein Before the configuration server calls the original model from the model database, the method further includes: The management server receives a service activation request, wherein the service activation request includes a service requirement; A target device among the edge devices is determined according to the type of the service requirement and / or the area corresponding to the service requirement.

3. The method according to claim 1, wherein The configuration server calls the original model from the model database, including: The original model stored in the model database in a preset format is pulled to the configuration server through a pull code.

4. A model deployment method, characterized in that: A method for applying a management server to an edge computing system, wherein the edge computing system includes a management server, a configuration server, and an edge device, and comprising the following steps: receiving a selection instruction for a target device in the edge device; Sending target configuration information to the configuration server according to the selection instruction, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information includes category information of the edge device; the category information of the edge device includes at least one of Bitmain, Rockchip RKNN, NVIDIA Jetson, NVIDIA RTX / GTX, X86 CPU, and ARM CPU; The target configuration information is used to enable the configuration server to adjust the original model called from the model database by the configuration server according to the category information of the target device, and generate model data matching the target device; wherein the original model in the model database is a model with a fixed format, and the execution tasks of the original model deployed on the target device with different category information are different; The method further comprises: The management server receives configuration information sent by each edge device; The management server adds the edge device to an edge device pool according to the configuration information, wherein the target configuration information is called by the management server from the edge device pool.

5. A model deployment method, characterized in that: A configuration server is applied to an edge computing system, wherein the edge computing system includes a management server, a configuration server, and an edge device. The method includes the following steps: Call the original model from the model database; Receive target configuration information sent by the management server, wherein the target configuration information is configuration information of a target device in the edge device, and the configuration information includes category information of the edge device; the target configuration information is called by the management server from an edge device pool, and the management server adds the edge device to the edge device pool according to the configuration information sent by each edge device; the category information of the edge device includes at least one of Bitmain, Rockchip RKNN, NVIDIA Jetson, NVIDIA RTX / GTX, X86 CPU, and ARM CPU; Adjusting the original model according to the category information of the target device to generate model data matching the target device; wherein the original model in the model database is a model with a fixed format, and the execution tasks of the original model when deployed on the target device with different category information are different; Sending the model data to the target device so that the edge device can perform model deployment according to the model data; The step of adjusting the original model according to the category information of the target device to generate model data matching the target device includes: According to the category information of the target device, converting the original model through a model conversion tool to convert the original model into model data matching the category information of the target device; the original model includes one of a torch model and a tensorflow model; In which, the configuration information is sent to the configuration server that performs model conversion in the json format, and the model conversion tool is configured through json; after obtaining the configuration information in the json format, the configuration server updates the json configuration of the model conversion tool, and then converts the original model through the model conversion tool.

6. A model deployment device, characterized in that: Applied to edge computing systems, which include management servers, configuration servers, and edge devices: The configuration server includes: Model calling module, used to call the original model from the model database; The management server includes: an information sending module, configured to receive a selection instruction for a target device in an edge device, and send target configuration information to the configuration server according to the selection instruction, wherein the target configuration information is configuration information of the target device in the edge device, and the configuration information includes category information of the edge device; the category information of the edge device includes at least one of Bitmain, Rockchip RKNN, NVIDIA Jetson, NVIDIA RTX / GTX, X86 CPU, and ARM CPU; The configuration server also includes: a model data generation module, configured to adjust the original model according to the category information of the target device to generate model data matching the target device; wherein the original model in the model database is a model with a fixed format, and the execution tasks of the original model when deployed on the target device with different category information are different; A model data sending module, configured to send the model data to the target device; The edge device includes: A deployment module, configured to deploy the model according to the model data; The edge device further includes: A configuration information uploading module, configured to send the configuration information of the edge device to the management server; The management server also includes: An adding module is used to add the configuration information to the edge device pool, wherein the target configuration information is called by the management server from the edge device pool The model data generation module adjusts the original model to generate model data matching the target device according to the category information of the target device, including: The model data generation module converts the original model using a model conversion tool according to the category information of the target device, and converts the original model into model data that matches the category information of the target device; the original model includes one of a torch model and a tensorflow model; In which, the configuration information is sent to the configuration server that performs model conversion in the json format, and the model conversion tool is configured through json; after obtaining the configuration information in the json format, the configuration server updates the json configuration of the model conversion tool, and then converts the original model through the model conversion tool.

7. An electronic device, characterized in that: including a processor, a memory, and a program stored in the memory and executable on the processor; When the program is executed by a processor, the steps of the method according to any one of claims 1 to 3 can be implemented; or When the program is executed by a processor, the steps of the method according to claim 4 can be implemented; or When the program is executed by a processor, the steps of the method according to claim 5 can be implemented.

8. A readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented; or Implement the steps of the method according to claim 4; or Implement the steps of the method according to claim 5.

Images