Model loading method, electronic equipment and storage medium

Abstract

The invention discloses a model loading method, electronic equipment and a storage medium. The model loading method comprises the steps that a main card responds to starting of target equipment where the main card is located, whether a target network accessed by the target equipment is the same as a set preset network or not is judged, and the target equipment comprises the main card and at least one sub-card; in response to the fact that the target network is different from the preset network, establishing communication connection between the main card and each daughter card in the target equipment; and sending a first model configuration instruction to each daughter card, so that each daughter card performs data analysis by using an algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. According to the scheme, the effectiveness of model loading can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a model loading method, an electronic device, and a storage medium. Background Art

[0002] In the field of video surveillance, there are more and more types of intelligent analysis that need to be processed simultaneously, such as face analysis, fingerprint analysis, vehicle analysis, etc. If the above analysis process is only processed by a central processing unit (CPU), it is difficult to complete the analysis task. In this case, smart cards are usually used to perform intelligent analysis tasks. In the current common scheme, there are multiple smart cards used for intelligent analysis tasks, and each smart card only performs one type of intelligent analysis task. It should be added here that the smart card that processes one type of intelligent analysis task is called a smart sub-card, and each type of intelligent analysis task is pre-compressed into an intelligent model package and stored in the main card. During the implementation process, the processing order and general business logic of the above multiple smart sub-cards are performed by the main card. Generally, a master-slave architecture format is adopted, with one main card and multiple sub-cards. Multiple physical transmission lines are set between the main card and the multiple sub-cards. The main card uses an ordinary CPU to perform the processing order of the sub-cards and the overall scheduling logic and other tasks, and the smart sub-cards are used for different types of intelligent analysis tasks. It can be understood that in the master-slave architecture format, the main chip uses an ordinary CPU to perform general business, and the slave chip uses a smart card to perform intelligent algorithm analysis. Large-scale intelligent devices will have a main card with multiple intelligent sub-cards. Although there are multiple physical transmission lines between the main card and multiple sub-cards, the sub-cards can perform intelligent algorithm analysis. However, when performing intelligent algorithm analysis, the sub-cards may not work properly due to network conflicts, thus affecting the model loading of the sub-cards.

[0003] In view of the existing technical defects, how to provide an effective model loading solution is a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0004] The present application at least provides a model loading method, an electronic device and a storage medium.

[0005] The present application provides a model loading method, comprising: a main card is started in response to a target device where the main card is located, and determines whether a target network accessed by the target device is the same as a preset network, wherein the target device includes a main card and at least one sub-card; in response to the target network being different from the preset network, establishing a communication connection between the main card and each sub-card in the target device; and sending a first model configuration instruction to each sub-card so that each sub-card performs data analysis using an algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction.

[0006] The present application provides a model loading device, including: a judgment module, an establishment module and a sending module; the judgment module is used for a main card to respond to the target device where the main card is located to start up, and to judge whether a target network accessed by the target device is the same as a preset network, and the target device includes a main card and at least one sub-card; the establishment module is used to establish a communication connection between the main card and each sub-card in the target device in response to the target network being different from the preset network; the sending module is used to send a first model configuration instruction to each sub-card so that each sub-card performs data analysis using an algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction.

[0007] The present application provides an electronic device, including a memory and a processor, wherein the processor is used to execute program instructions stored in the memory to implement the above-mentioned model loading method.

[0008] The present application provides a computer-readable storage medium on which program instructions are stored. When the program instructions are executed by a processor, the above-mentioned model loading method is implemented.

[0009] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0010] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and are used together with the specification to illustrate the technical solution of the present application.

[0012] Figure 1 It is a flow chart of an embodiment of the model loading method of the present application; Figure 2 This is a schematic diagram of the effect of a virtual bridge in an embodiment of the model loading method of the present application; Figure 3 This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 1 ; Figure 4 This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 2 ; Figure 5This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 3 ; Figure 6 It is a structural schematic diagram of an embodiment of a model loading device of the present application; Figure 7 It is a structural schematic diagram of an embodiment of the electronic device of the present application; Figure 8 It is a structural diagram of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0013] The scheme of the embodiment of the present application is described in detail below in conjunction with the drawings of the specification.

[0014] In the following description, for the purpose of explanation rather than limitation, specific details such as specific system structures, interfaces, and technologies are provided to facilitate a thorough understanding of the present application.

[0015] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the objects associated before and after are in an "or" relationship. In addition, "many" in this article means two or more than two. In addition, the term "at least one" in this article means any combination of at least two of any one or more of a plurality of, for example, including at least one of A, B, and C, can mean including any one or more elements selected from the set consisting of A, B, and C.

[0016] The present application provides some model loading methods and model loading devices. The application scenarios of the model loading method include but are not limited to intelligent analysis tasks. The executor of the model loading method can be a model loading device. For example, the model loading device can be set in a terminal device or a server or other processing device, including user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, vehicle-mounted device, etc. In some possible implementations, the model loading method can be implemented by a processor calling computer-readable instructions stored in a memory.

[0017] See also Figure 1 , Figure 1 1 is a flow chart of an embodiment of a model loading method of the present application. Specifically, the model loading method may include the following steps: The existing solution has the problem of conflict between internal and external network IPs. When the virtual network segment of the sub-card is the same as the external network, the sub-card IP and the external network device IP may be the same. At this time, the conflicting sub-card in the device will not work properly.

[0018] In the existing framework, the sub-card has no permanent storage medium, and the intelligent model package of the sub-card is stored in the main card. When the sub-card is started, the main control storage medium is mounted to the sub-card through the master-slave network, and the sub-chip accesses the mount path to load the model. As the business scenarios increase, the types of intelligent algorithms are increasing, and the number of algorithm combinations required for different scenarios is increasing. Due to the limitation of device resources, a product cannot integrate all algorithms at the same time. Different combinations require different products. For new combinations, the product development and delivery time is longer.

[0019] Step S11: the main card responds to the target device where the main card is located being started, and determines whether the target network accessed by the target device is the same as the preset network. The target device includes the main card and at least one sub-card.

[0020] The target device may be a back-end analysis device capable of performing intelligent analysis tasks. The back-end analysis device may be a storage device with intelligent analysis capabilities. Specifically, the target device may be an intelligent video surveillance server IVSS. The target device may adopt a master-slave architecture format. The master chip of the target device may be a master card. The slave chips of the target device may be multiple daughter cards. The master card may be a processor of the target device. The master card may have a permanent storage medium. In the intelligent analysis task, the daughter card may be an intelligent module capable of processing a type of intelligent analysis task. Among them, there is a physical transmission line between each daughter card and the master card. The daughter card may be an intelligent module without a permanent storage medium. It can be understood that the main control may refer to the system control unit on the main card, and the main control is responsible for managing and coordinating the various parts of the entire computer system of the target device. Therefore, it can be said that the main control is the system on the main card. The system of the target device may be the main control. It can be understood that the execution subject of the model loading method may be a model loading device. Specifically, the execution subject of the model loading method may be the main control of the target device. Further, the execution subject of the model loading method may be the main card.

[0021] In some application scenarios, the main card can sense that the target device is started. The model loading device includes a judgment module. The judgment module can be used to judge whether the target network accessed by the target device is the same as the preset network set. The input of the judgment module can be the target network and the preset network. The output of the judgment module can be a judgment result. The judgment result can be that the target network is the same as the preset network or the target network is different from the preset network. After the main card responds to the start of the target device where the main card is located, the main card can judge whether the target network accessed by the target device is the same as the preset network set through the judgment module. In some application scenarios, the model loading device includes an acquisition module. The acquisition module can be used to obtain the target network accessed by the target device, and can also be used to obtain the preset network set. The target network accessed by the target device can be an external network of the target device. The preset network set by the target device can be an external network of the target device.

[0022] It can be understood that after the main card responds to the start-up of the target device where the main card is located, the main card uses the judgment module to judge whether the external network and the internal network of the target device are the same.

[0023] Step S12: In response to the target network being different from the preset network, establishing a communication connection between the main card and each sub-card in the target device.

[0024] The model loading device may include an establishment module. The establishment module is used to establish a communication connection between the main card and each sub-card in the target device. In the above step S11, the judgment result output by the judgment module is that the target network is different from the preset network. After the main card responds that the target network is different from the preset network, the main card establishes a communication connection between the main card and each sub-card in the target device. It can be understood that the main card realizes data exchange, resource sharing and control scheduling between the main card and each sub-card by establishing a communication connection between the main card and each sub-card. In some application scenarios, the data exchange between the main card and each sub-card can be that the communication connection between the main card and each sub-card allows data to be transmitted between them. Information, commands and status can be transmitted between the main card and each sub-card through the communication connection, thereby realizing the collaboration between the main card and each sub-card of the target device and the expansion of intelligent analysis functions. In some application scenarios, resource sharing between the main card and each sub-card can be that the main card establishes a communication connection, and the main card and each sub-card can share resources, such as memory, etc. The utilization and performance of the system can be improved, so that different sub-cards can work at the same time or use resources preferentially as needed. In some application scenarios, the control and scheduling between the main card and each sub-card can be a communication connection so that the main card can control and schedule each sub-card. The main card can send instructions and control signals to the sub-card to make it work in a predetermined manner, so as to allow intelligent control and collaborative work of the target device.

[0025] Step S13: Sending the first model configuration instruction to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the first model configuration instruction after receiving the first model configuration instruction.

[0026] The first model configuration instruction may be related to an already installed algorithm model. In some application scenarios, the sub-card may determine the already installed algorithm model corresponding to the first model configuration instruction based on the first model configuration instruction. The main card may send the first model configuration instruction to each sub-card. After the main card issues the first model configuration instruction to each sub-card, each sub-card may receive the first model configuration instruction issued by the main card. After each sub-card receives the first model configuration instruction, each sub-card may perform data analysis using the algorithm model corresponding to the first model configuration instruction. It is understandable that each sub-card may first load the algorithm model corresponding to the first model configuration instruction, and then each sub-card implements the algorithm model corresponding to the first model configuration instruction for data analysis.

[0027] It is understandable that before the main card receives the first model configuration instruction, the algorithm model corresponding to the first model configuration instruction in each sub-card may have been loaded. After each sub-card receives the first model configuration instruction, each sub-card may perform data analysis on the loaded algorithm model.

[0028] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0029] In some embodiments, the model loading method may further include the following steps: first, in response to the target network being the same as the preset network, adjusting the preset network to another network, the other network being different from the target network, and controlling the target device to restart, wherein the network accessed by the target device after restarting is the target network.

[0030] The model loading device may include an adjustment module. The adjustment module is used to adjust the preset network to another network. In the above step S11, the judgment result output by the judgment module is that the target network is the same as the preset network. After the main card responds that the target network is the same as the preset network, the main card uses the adjustment module to adjust the preset network to another network. Among them, the other network is a network that is different from the target network. In some application scenarios, after the main card adjusts the preset network to another network, the main card controls the target device to restart so that the network connected to the target device after the restart is the target network.

[0031] In some application scenarios, the adjustment module may be used to adjust the preset network to another network that is different from the target network accessed by the target device. Specifically, the adjustment module may modify the preset network corresponding to the default configuration of the target device, add one to the third bit of the gateway of the preset network, and obtain the other network, so that the other network is different from the target network corresponding to the external network.

[0032] In some application scenarios, after the main card adjusts the preset network to another network, the main card controls the target device to restart. In response to the target device where the main card is located restarting, the main card determines whether the target network accessed by the target device is the same as the other network. In response to the target network being different from the other network, the main card establishes a communication connection between the main card and each sub-card in the target device. The main card executes the above step S13.

[0033] It is understandable that at this time, the main card adjusts the internal network to a network different from the external network, so that there will be no conflict between the internal and external networks after the target device is restarted. The model loading device can execute the above step S13, and when there is no conflict between the internal and external networks, each sub-card in the target device can load the algorithm model normally to realize the intelligent analysis task of each sub-card.

[0034] In some embodiments, the above step S12 may include the following steps: first, the main card creates a virtual network bridge and virtual network cards corresponding to each sub-card. Second, each virtual network card is bound to each virtual port in the virtual network bridge. Finally, the configuration information of the virtual network card is sent to each sub-card, so that each sub-card establishes a connection with each virtual network card based on the configuration information of the virtual network card, wherein the main card communicates with each sub-card based on the virtual network bridge.

[0035] After the target device is started, the main card can create a virtual bridge and a virtual network card corresponding to each sub-card. The virtual bridge created by the main card can be a logical bridge created by connecting multiple network interfaces. The virtual bridge can connect multiple network interfaces together so that the main card and each sub-card on these network interfaces can communicate with each other. It can be understood that the main card can correspond to the main network interface. By creating a virtual bridge, the main card can connect the main network interface corresponding to the main card with multiple network interfaces corresponding to each sub-card to achieve communication and data transmission between different networks. It can be considered that the model loading device uses the virtual bridge created by the main card, and the target network accessed by the main card can communicate and transmit data between the preset network corresponding to the default configuration of the target device. In other application scenarios, after the above-mentioned step of controlling the target device to restart, the main card responds to the target network being different from other networks, and establishes a communication connection between the main card and each sub-card in the target device. At this time, the model loading device uses the virtual bridge created by the main card, and the target network accessed by the main card can communicate and transmit data between other networks corresponding to the default configuration of the target device.

[0036] See also Figure 2 , Figure 2 It is a schematic diagram of the effect of the virtual bridge in an embodiment of the model loading method of the present application.

[0037] After the main card creates a virtual network bridge, it can also create virtual network cards corresponding to each sub-card according to the number of sub-cards. Figure 2As shown, the virtual network cards corresponding to each subcard can be represented as virtual network card 0, virtual network card 1, and virtual network card 2, etc. The number of virtual network cards corresponding to each subcard can be the same as the number of each subcard. A virtual bridge contains multiple virtual ports. The virtual port in the virtual bridge can be a network interface of the subcard. The number of virtual ports in the virtual bridge can be the same as the number of each subcard. The virtual network card of a subcard refers to a virtualized network device associated with the subcard, which is a connection channel between the subcard and the virtual bridge. The virtual network card can be regarded as a physical interface of the subcard, which is used to exchange data with a virtual machine, a container or other virtual devices. In some application scenarios, each subcard has its own virtual network card for configuring and managing network connections. The virtual network card can have various network functions, such as IP settings, network policies, security controls, etc. The virtual port of a subcard refers to the network interface of the subcard when it is running in a virtual machine or a container. In some application scenarios, the virtual port is a virtual device created by virtualization technology, and the virtual port is associated with the virtual network card of the subcard. In a virtual machine or container, a virtual port is part of the network stack and is used to handle the sending and receiving of network data. Each virtual machine or container has its own virtual port, which acts as a bridge between the virtual machine or container and the virtual bridge, enabling communication between the virtual device and the physical network. It is understandable that a virtual machine, container, or other virtual device can be another subcard connected to the same virtual bridge as the subcard.

[0038] It can be understood that the virtual network card can be regarded as the physical interface of the daughter card, and the virtual port can be regarded as the network interface when the daughter card is running.

[0039] The main card binds each virtual network card to each virtual port in the virtual bridge. In some application scenarios, the main card binds the virtual network card to the virtual port list in the virtual bridge. Among them, the virtual port list can be used to represent each virtual port, and the virtual port list can be represented as a list port_lis. The virtual port list contains multiple virtual ports. Exemplarily, the virtual port list can be virtual port 1, virtual port 2, virtual port 3, etc. It can be understood that binding the virtual network card and virtual port of the subcard can achieve network isolation and management. In some application scenarios, in the virtual bridge, the main card needs to bind the virtual network card and virtual port of the subcard to establish the connection and communication between the subcard and the virtual bridge. The main card binds the virtual network card and virtual port of the subcard to realize data transmission and exchange. When the virtual network card of the subcard is bound to the virtual bridge, the virtual bridge can receive and send data from the subcard and forward it to other devices connected to the same virtual bridge. Among them, the devices on the same virtual bridge can include the main card, other subcards, and virtual network cards of other subcards. The binding relationship between each subcard and each virtual port in the virtual bridge establishes the connection and communication link between each device in the virtual network. In other application scenarios, the binding relationship between each subcard and each virtual port in the virtual bridge can also realize the configuration and management of network parameters. By binding the subcard with the virtual port of the virtual bridge, the virtual port of the subcard can exchange and configure network parameters such as IP address, subnet mask, gateway, etc. through the virtual bridge. In this way, the network stack running in the virtual machine or container can obtain the required network information through the virtual port and communicate with other devices. In other application scenarios, the administrator can configure independent network parameters such as IP address, subnet mask, gateway, etc. for each subcard, thereby realizing more fine-grained network management. In some application scenarios, by binding the virtual network card of the subcard with the virtual port of the virtual bridge, the network isolation of each subcard and the main card can be realized. In this way, the network traffic between different subcards can be effectively separated to avoid interference and conflict.

[0040] The main card can send the configuration information of the virtual network card to each sub-card, so that each sub-card can establish a connection with each virtual network card based on the configuration information of the virtual network card, wherein the main card and each sub-card communicate based on the virtual network bridge. After the sub-card is started, the sub-card can obtain the IP configuration information of the corresponding virtual network card from the main card through the bus, and configure the virtual network card of the sub-card (which may be the above-mentioned virtual network card 0). After the sub-card obtains the configuration information of the virtual network card, the virtual network card of the sub-card may be the same as or different from the above-mentioned virtual network card 0. After the sub-card obtains the configuration information of the virtual network card, each sub-card establishes a connection with each virtual network card based on the configuration information of the virtual network card, thereby realizing the communication connection between the main card and each sub-card for quality inspection.

[0041] In some embodiments, the model loading method may further include the following steps: First, the main card traverses the model directory to determine the first name of the algorithm model that the target device can run, and the model directory is used to record the first name of the algorithm model that the target device can run. Secondly, the first name of each algorithm model is sent to the preset recipient so that the preset recipient displays the first name of each algorithm model in the configuration interface of the preset recipient. Then, the first model configuration instruction sent by the preset recipient is received, and the first model configuration instruction is issued by the preset recipient in response to the selection operation of the displayed algorithm model.

[0042] The model directory may be related to the algorithm models that the target device can run. The model directory may be related to the algorithm models that have been installed. Among them, the model directory may include algorithm frameworks and algorithm packages. Among them, the algorithm framework may be the basis for the work of the algorithm package. The algorithm framework may contain the services of all algorithm models within the framework. Among them, the algorithm packages in the model directory may correspond to the dynamic libraries and algorithm models of the algorithm packages, and the file names of the algorithm packages are named using the major categories of intelligent services. An algorithm package in the model directory may contain multiple algorithm models of the same type of intelligent services. An algorithm package in the model directory may contain an algorithm model corresponding to an intelligent service.

[0043] In some application scenarios, the model directory is used to record the first name of the algorithm model that the target device can run. Specifically, the model directory can be a directory of the first name corresponding to the algorithm model and the installation path of the algorithm model. The preset recipient can be a client. There can be multiple preset recipients. The configuration interface of the preset recipient can be a display interface that can respond to user operations. The main card traverses the model directory to determine the first name of the algorithm model that the target device can run. The main card sends the first name of each algorithm model that the target device can run to the preset recipient, so that the preset recipient displays the first name of each algorithm model on the configuration interface of the preset recipient. The main card receives the first model configuration instruction sent by the preset recipient, and the first model configuration instruction is issued by the preset recipient in response to the selection operation of the displayed algorithm model. The algorithm model displayed on the configuration interface can be the first name of each algorithm model that the target device can run. Exemplarily, the first model configuration instruction sent by the preset recipient received by the main card is a configuration instruction of the installed algorithm model A, and the first model configuration instruction includes the first name of the algorithm model A. It can be understood that the algorithm model corresponding to the first name displayed on the configuration interface can be an algorithm model that can be run by the target device, or can be an algorithm model that can be loaded by each sub-card.

[0044] In some embodiments, the step of sending the first model configuration instruction to each daughter card may include the following steps: sending the first model configuration instruction to the virtual bridge so that the virtual bridge forwards the first model configuration instruction to each daughter card.

[0045] After the main card receives the first model configuration instruction sent by the preset recipient, the main card sends the first model configuration instruction to the virtual bridge. Exemplarily, the main card sends the first model configuration instruction of the installed algorithm model A to the virtual bridge. At this time, the virtual bridge can forward the first model configuration instruction to each sub-card. After the virtual bridge forwards the first model configuration instruction of the algorithm model A to each sub-card, each sub-card can first load the algorithm model A corresponding to the first model configuration instruction, and then each sub-card implements the algorithm model A corresponding to the first model configuration instruction to perform data analysis.

[0046] In some embodiments, the first model configuration instruction also includes parameter information of the data to be analyzed, and the model loading method may further include the following steps: first, receiving the data analysis results of each subcard, each data analysis result is obtained by each subcard performing data analysis on the data to be analyzed based on the parameter information after running the algorithm model. Secondly, forwarding each data analysis result to a preset recipient so that the preset recipient displays the data analysis result.

[0047] The first model configuration instruction also includes parameter information of the data to be analyzed. The data to be analyzed may be input data of the algorithm model corresponding to the first model configuration instruction. After the model loading device inputs the data to be analyzed into the algorithm model corresponding to the first model configuration instruction, the algorithm model may output a data analysis result. The parameter information of the data to be analyzed may be directly the data to be analyzed, or may be an index that can obtain the data to be analyzed. The model loading device may obtain the data to be analyzed corresponding to the parameter information based on the parameter information of the data to be analyzed. After the main card receives the data analysis results of each sub-card. The main card forwards each data analysis result to a preset recipient so that the preset recipient displays the data analysis result.

[0048] Exemplarily, each sub-card may first load the algorithm model A corresponding to the first model configuration instruction. After each sub-card receives the parameter information of the data to be analyzed in the first model configuration instruction, each sub-card obtains the data to be analyzed based on the parameter information. Each sub-card performs data analysis on the data to be analyzed in the algorithm model A to obtain a data analysis result. The algorithm model A may output the data analysis result. Each sub-card may output each data analysis result. Each sub-card may send each data analysis result to the main card. After the main card receives the data analysis results of each sub-card, the main card forwards each data analysis result to the preset receiving party client. After the client receives the data analysis results output by the algorithm model A, the client displays the data analysis results output by the algorithm model A. In some embodiments, the model loading method may further include the following steps: first, receiving a newly added algorithm model sent by a preset receiving party, the newly added algorithm model is an algorithm model imported by the preset receiving party. Secondly, installing the newly added algorithm model to the model directory to obtain a new model directory. Subsequently, the main card traverses the new model directory to determine the second name of the algorithm model that the target device can currently run. Finally, the name of the algorithm model that the target device can currently run is sent to the preset recipient, so that the preset recipient displays the second name on the configuration interface of the preset recipient.

[0049] According to the needs of the intelligent analysis task, the model loading device can add an algorithm model to meet the needs of obtaining more intelligent capabilities. The preset receiver responds to the user's import operation on the configuration display interface, and the preset receiver imports the algorithm model corresponding to the import operation. The newly added algorithm model is the algorithm model corresponding to the import operation. The preset receiver can send the newly added algorithm model to the main card. In response to receiving the newly added algorithm model sent by the preset receiver, the main card can install the newly added algorithm model to the model directory. The main card traverses the new model directory to determine the name of the algorithm model that the target device can currently run. In some application scenarios, the main card traverses the new model directory to determine that the second name of the algorithm model that the target device can currently run can add a second name corresponding to the algorithm model. Then, the main card sends the name of the algorithm model that the target device can currently run to the preset receiver, so that the preset receiver displays the first name and the second name on the configuration interface of the preset receiver. The first name can be the algorithm model that the target device can run before installing the newly added algorithm model to the model directory. The second name can be the name of the newly added algorithm model.

[0050] It can be understood that before the above-mentioned main card receives the newly added algorithm model sent by the preset recipient, the algorithm model corresponding to the second name displayed in the configuration interface may be an algorithm model that the target device cannot run, or an algorithm model that each sub-card cannot load.

[0051] In some embodiments, after the above-mentioned main card traverses the new model directory to determine the second name of the algorithm model that the target device can run, the model loading method may also include the following steps: first, send a new notification corresponding to the newly added algorithm model to each sub-card so that each sub-card loads the newly added algorithm model. And / or, receive a second model configuration instruction sent by a preset recipient, the second model configuration instruction is issued by the preset recipient in response to a second selection operation on the displayed algorithm model. Subsequently, send the second model configuration instruction to the virtual bridge so that the virtual bridge forwards the second model configuration instruction to each sub-card.

[0052] In some application scenarios, the main card sends a new notification corresponding to the new algorithm model to each sub-card so that each sub-card loads the new algorithm model. The new notification may be related to the new algorithm model. The new notification may represent the loading parameter information of each sub-card for the new algorithm model. Exemplarily, the new algorithm model may be algorithm model B. After each sub-card receives the new notification, each sub-card may obtain the loading parameter information of the algorithm model B stored in the main card through the communication connection with the virtual bridge. Each sub-card may implement the loading of the new algorithm model B in each sub-card based on the loading parameter information corresponding to the new algorithm model. It is understandable that after each sub-card receives the new notification corresponding to the new algorithm model, each sub-card may pre-load the new algorithm model, so that before the main card receives the second model configuration instruction sent by the preset recipient, or before the main card sends the second model configuration instruction to the virtual bridge, the new algorithm model in each sub-card is already loaded.

[0053] In some application scenarios, the main card receives a second model configuration instruction sent by a preset receiver, and the second model configuration instruction is issued by the preset receiver in response to a second selection operation on the displayed algorithm model. The configuration interface of the preset receiver can display a selection button corresponding to the newly added algorithm model B. In response to the second selection operation on the selection button corresponding to the newly added algorithm model B, the preset receiver sends the second model configuration instruction corresponding to the newly added algorithm model B to the main card. After receiving the second model configuration instruction sent by the preset receiver, the main card sends the second model configuration instruction to the virtual bridge so that the virtual bridge forwards the second model configuration instruction to each sub-card.

[0054] In some other application scenarios, the main card sends a new notification corresponding to the new algorithm model to each sub-card so that each sub-card loads the new algorithm model. The main card receives a second model configuration instruction sent by a preset receiver, and the second model configuration instruction is issued by the preset receiver in response to a second selection operation on the displayed algorithm model. The main card sends the second model configuration instruction to the virtual bridge so that the virtual bridge forwards the second model configuration instruction to each sub-card.

[0055] In some embodiments, before the above-mentioned step of the main card traversing the model directory to determine the first name of the algorithm model that the target device can run, the model loading method may also include the following steps: first, through the virtual network bridge, the main card mounts the model directory to each sub-card, so that each sub-card loads each algorithm model in the model directory.

[0056] The main card mounts the model directory to each sub-card. In some application scenarios, the main card mounts the algorithm models that the target device can run in the model directory to each sub-card, so that each sub-card loads each algorithm model in the model directory.

[0057] In some application scenarios, the main card can mount the loading parameter information in the model directory to each sub-card. The loading parameter information in the model directory corresponds to the algorithm model. After each sub-card obtains the target loading parameter information, each sub-card can load the algorithm model corresponding to the target loading parameter information in the model directory into each sub-card. Each sub-card can obtain the target loading parameter information sent by the main card through a virtual bridge. The above-mentioned new notification may include target loading parameter information. The target loading parameter information in the new notification may point to the newly added algorithm model in the new model directory. After each sub-card obtains the target loading parameter information sent by the virtual bridge, it loads the newly added algorithm model corresponding to the target loading parameter information mounted in each sub-card.

[0058] In the present application, the model loading device includes an intelligent related module. The intelligent related module can be a main control terminal or a main card involving an intelligent configuration module, a data display module and a storage related module. The intelligent related module can be used for the main card to receive various model configuration instructions issued by a preset recipient and various model configuration instructions issued by the main card to the virtual bridge. The data display module can be used for the main card to receive the data analysis results of each sub-card and the main card to forward each data analysis result to the preset recipient. The storage related module can be an algorithm model that can be stored in the main card and can be run by the target device related to the model target. Among them, different algorithm models in the model directory are packaged independently. Specifically, the algorithm models corresponding to the major categories of intelligent services are packaged independently.

[0059] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0060] See also Figure 3 , Figure 3This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 1 .

[0061] The main card sequentially executes step S31, step S32, and step S33. Step S31: The main card starts in response to the target device where the main card is located. Step S32: Obtain the target network to which the target device is connected. Step S33: The main card determines whether the target network is the same as the preset network. If the judgment result of step S33 is that the target network is different from the preset network, the main card can sequentially perform steps S34, step S35, and step S36. Step S34: The main card creates a virtual bridge and virtual network cards corresponding to each sub-card. Step S35: Bind each virtual network card to each virtual port in the virtual bridge. Step S36: Send the configuration information of the virtual network card to each sub-card. It can be understood that after step S36, the configuration information of the virtual network card is sent to each sub-card so that each sub-card establishes a connection with each virtual network card based on the configuration information of the virtual network card, wherein the main card communicates with each sub-card based on the virtual bridge.

[0062] If the result of step S33 is that the target network is the same as the preset network, the main card can execute step S37. Step S37: Adjust the preset network to another network. After the main card executes step S37, the main card can control the target device to restart. After the main card controls the target device to restart, the main card can execute steps S31 and S32 in sequence. After the main card executes step S32, the main card can execute step S33. At this time, in step S33, the target network can be the network accessed by the target device after the target device is restarted, and the preset network can be another network obtained after the preset network is adjusted after the processing of step S37. It can be understood that step S33 can also include: judging whether the target network accessed by the target device after the target device is restarted is the same as the other network obtained after the preset network is adjusted. That is, judging whether the target network is the same as the other network. If the result of step S33 is that the target network is different from the other network, the main card executes steps S34, S35 and S36 in sequence. Thereby, the main card establishes a communication connection between the main card and each sub-card in the target device. The main card communicates with each sub-card based on a virtual bridge.

[0063] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0064] See also Figure 4 , Figure 4 This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 2 .

[0065] Step S41: The main card traverses the current model directory to determine the name of the algorithm model that the target device can run. It is understandable that step S41 can be that the main control in the main card traverses the current model directory to determine the name of the algorithm model that the target device can run. The main control in the main card can obtain the intelligent capabilities corresponding to the algorithm model supported by the target device. In some application scenarios, after the main card determines the name of the algorithm model that the target device can run, the main card can send target loading parameter information to each sub-card so that each sub-card loads the target algorithm model based on the target loading parameter information. Among them, the target algorithm model can be the algorithm model that the main card determines that the target device can run. Specifically, the main card sends the target loading parameter information to the virtual bridge so that the virtual bridge forwards the target loading parameter information to each sub-card. It is understandable that the name of the algorithm model that the target device can run can be the above-mentioned first name or the above-mentioned second name. The target algorithm model can be the algorithm model corresponding to the first model configuration instruction or the newly added algorithm model corresponding to the second model configuration instruction. The target loading parameter information can be sent to each sub-card after the main card traverses the current model directory to determine the name of the algorithm model that the target device can run.

[0066] In some application scenarios, after step S41, the main card may determine the first name of the algorithm model. The main card may send a first model configuration instruction corresponding to the first name of the algorithm model to the sub-card. In other application scenarios, after step S41, the main card may determine the second name of the algorithm model. The main card may send a second model configuration instruction corresponding to the second name of the algorithm model to the sub-card.

[0067] The main card sends the first name of each algorithm model to the preset receiver, so that the preset receiver displays the first name of each algorithm model on the configuration interface of the preset receiver. The main card receives the first model configuration instruction sent by the preset receiver, and the first model configuration instruction is issued by the preset receiver in response to the selection operation of the displayed algorithm model. The main card sends the first model configuration instruction to the virtual bridge, so that the virtual bridge forwards the first model configuration instruction to each sub-card.

[0068] Step S42: Sending the target model configuration instruction to each sub-card, so that each sub-card performs data analysis using the algorithm model corresponding to the target model configuration instruction after receiving the target model configuration instruction.

[0069] It is understandable that the algorithm model corresponding to the target model configuration instruction in each sub-card of the target device can be pre-loaded. After waiting for the main card to send the target model configuration instruction to each sub-card, each sub-card can receive the target model configuration instruction. Each sub-card starts the algorithm model corresponding to the pre-loaded target model configuration instruction based on the target model configuration instruction, so that the algorithm model corresponding to the target model configuration instruction performs data analysis to obtain the target data analysis result. Each sub-card can send the target data analysis result output by the algorithm model corresponding to the target model configuration instruction to the virtual bridge, so that the virtual bridge forwards the target data analysis result to the main card.

[0070] It is understandable that the target model configuration instruction can be the first model configuration instruction or the second model configuration instruction. The algorithm model corresponding to the target model configuration instruction can be the target algorithm model. Specifically, the target algorithm model can be the initial algorithm model corresponding to the first model configuration instruction or the newly added algorithm model corresponding to the second model configuration instruction.

[0071] After the above step S42, the main card may also execute the step of receiving the target data analysis results of each sub-card by the main card. Wherein, each target data analysis result is obtained by each sub-card performing data analysis on the data to be analyzed based on the target parameter information after running the target algorithm model. The target model configuration instruction may include the target parameter information of the target algorithm model. The target parameter information may be the data input of the target algorithm model. The target parameter information may also be the call information of the data input of the target algorithm model. Each sub-card may use the data called from the relevant database based on the call information as the data input. Each sub-card may obtain the data stream required by the intelligent analysis task as the data input based on the call information. It is understandable that each sub-card may obtain the data input of the target algorithm model based on the target parameter information, so that the target algorithm model performs data analysis on the target algorithm model based on the data input to obtain the target data analysis result. Each sub-card may send the target data analysis result to the main card.

[0072] Step S43: the master card forwards each target data analysis result to a preset recipient, so that the preset recipient displays the target data analysis result.

[0073] The main card receives the target data analysis result. It is understandable that the target data analysis result received by the main card may be the target data analysis result sent by each sub-card forwarded to the main card by the virtual bridge. The main card may forward the target data analysis result to a preset recipient so that the preset recipient displays the target data analysis result on the display interface.

[0074] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0075] See also Figure 5 , Figure 5 This is a schematic diagram of the working of the main card in the first embodiment of the model loading method of this application. Figure 3 .

[0076] Step S501: The main card traverses the model directory to determine the first name of the algorithm model that the target device can run. Before step S501, the target device is started. In response to the start of the target device, each sub-card automatically starts and mounts the main control model directory. Specifically, each sub-card mounts the algorithm package path in the model directory. After mounting the model directory, each sub-card starts the intelligent related programs of each sub-card. It can be understood that after mounting the model directory, each sub-card starts the intelligent related programs of each sub-card, which can be the algorithm model that the target device can run, and the algorithm model supported by the target device at this time. Starting the intelligent related programs of each sub-card can be regarded as the pre-loading corresponding to the initial algorithm model that the target device can run. The initial algorithm model can be the algorithm model that each sub-card can pre-load based on the model directory mounted before the algorithm capability is added. After completing the pre-loading of the initial algorithm model, each sub-card waits for the target algorithm model configuration instruction to be issued to each sub-card, thereby realizing data analysis. Step S501 can be in response to the start of the target device, and the main card traverses the model directory. Specifically, the main card traverses the algorithm package installation path, and the main card obtains the device intelligent capability. It can be understood that, assuming that the target device has installed algorithm model A, the algorithm capability supported by the target device is algorithm package [A]. The main card traverses the model directory to determine that the first name of the algorithm model that the target device can run is A.

[0077] Step S502: the master card sends the first name of each algorithm model to a preset recipient.

[0078] The main card sends the first name of each algorithm model to the preset recipient. The main card can report the algorithm capability A supported by the target device to the preset recipient. It can be understood that the main card can send the name of the algorithm capability supported by the target device to the preset recipient, thereby completing the main card reporting the algorithm capability to the preset recipient. After step S502, the display interface of the preset recipient can display the first name A of the algorithm model currently supported by the target device. The preset recipient can generate a first model configuration instruction in response to the selection operation of the displayed algorithm model. The preset recipient can send the first model configuration instruction to the main card.

[0079] Step S503: the primary card receives a first model configuration instruction sent by a preset recipient.

[0080] Step S504: the main card sends a first model configuration instruction to the virtual network bridge, so that the virtual network bridge forwards the first model configuration instruction to each sub-card.

[0081] After step S504, each subcard can set the algorithm configuration corresponding to the algorithm model A based on the received first model configuration instruction, implement data analysis of the algorithm model A, and obtain data analysis results. After obtaining the data analysis results, each subcard can forward the data analysis results to the main card through the virtual bridge.

[0082] Step S505: The main card obtains the first data analysis result and sends the first data analysis result to the preset recipient. After the main card obtains the first data analysis result forwarded by the virtual bridge, it sends the first data analysis result to the preset recipient. After receiving the data analysis result, the preset recipient displays the first data analysis result on the display interface of the preset recipient.

[0083] Step S506: the primary card receives the newly added algorithm model sent by the preset recipient.

[0084] The preset receiver can import algorithm model B, that is, algorithm package B. The preset receiver sends and installs the algorithm model B to the master card. The master card receives the newly added algorithm model B sent by the preset receiver.

[0085] Step S507: The main card installs the newly added algorithm model into the model directory to obtain a new model directory.

[0086] The main card installs the newly added algorithm model B into the model directory to obtain a new model directory. It is understandable that here the main card installs the newly added algorithm model B into the model directory to specify the installation path. The newly added algorithm model B can be the corresponding algorithm package B. Each algorithm package is packaged independently in the installation path in the model directory. The algorithm package can correspond to the dynamic library and algorithm model of the algorithm, and the file name of the algorithm package is named using the major category of intelligent business.

[0087] Step S508: The master card traverses the new model directory to determine the second name of the algorithm model that the target device can currently run.

[0088] After the main card completes the installation of the newly added algorithm model B, it re-traverses the new model directory (that is, the algorithm package installation directory) to obtain the capabilities supported by the current target device. It is understandable that at this time, the algorithm capabilities that the target device can support are algorithm model A and the newly added algorithm model B. The main card reports the capabilities to the preset recipient, and the main card knows that the sub-card re-acquires the algorithm capabilities. In other application scenarios, after step S508, the display interface of the preset recipient can display the second name A and B of the algorithm model currently supported by the target device. The preset recipient can generate a second model configuration instruction in response to the selection operation of the displayed algorithm model. It is understandable that the second model configuration instruction can be to send down the configuration only for algorithm model A or to send down the configuration only for algorithm model B or to send down the configuration only for algorithm models A and B.

[0089] Step S509: the master card sends the name of the algorithm model that the target device can currently run to the preset recipient.

[0090] The algorithm capability reported by the master card to the preset recipient may specifically be that the algorithm capability supported by the target device is algorithm model A and a newly added algorithm model B.

[0091] Step S510: The main card sends a new notification corresponding to the new algorithm model to each sub-card so that each sub-card loads the new algorithm model. The new notification can be used for each sub-card to reacquire the algorithm capability of the target device. Specifically, each sub-card can be restarted to obtain the algorithm capability supported by the target device as the preload corresponding to the algorithm model A and the new algorithm model B. Each sub-card can also obtain the preload corresponding to the corresponding new algorithm model B based on the new notification only, without repeating the preload of the algorithm model A. Each sub-card loads the new algorithm model B. Each sub-card waits for the main card to issue the second model configuration instruction.

[0092] Step S511: the primary card receives a second model configuration instruction sent by a preset recipient.

[0093] After receiving the second model configuration instruction sent by the preset recipient, the main card sends the second model configuration instruction to the virtual network bridge.

[0094] Step S512: the main card sends a second model configuration instruction to the virtual network bridge, so that the virtual network bridge forwards the second model configuration instruction to each sub-card.

[0095] Exemplarily, the second model configuration instruction is used to send configurations to algorithm models A and B. After step S512, after each sub-card receives the second model configuration instruction, each sub-card can set the algorithm configuration corresponding to the algorithm model A and the newly added algorithm model B based on the received second model configuration instruction, implement data analysis of the algorithm model A and the newly added algorithm model B, and obtain second data analysis results respectively. After obtaining the second data analysis result, each sub-card can forward the second data analysis result to the virtual bridge to the main card.

[0096] Step S513: The main card obtains the second data analysis result and sends the second data analysis result to the preset recipient. After the main card obtains the second data analysis result forwarded by the virtual bridge, it sends the second data analysis result to the preset recipient. After receiving the data analysis result, the preset recipient displays the second data analysis result on the display interface of the preset recipient. The display interface of the preset recipient is to display the relevant interface corresponding to the second data analysis result corresponding to the algorithm model A and the newly added algorithm model B data.

[0097] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0098] See also Figure 6 , Figure 6 It is a structural diagram of an embodiment of the model loading device of the present application. The model loading device 60 includes a judgment module 61, an establishment module 62 and a sending module 63. The judgment module 61 is used for the main card to respond to the target device where the main card is located to start up and judge whether the target network connected to the target device is the same as the preset network set. The target device includes the main card and at least one sub-card; the establishment module 62 is used to establish a communication connection between the main card and each sub-card in the target device in response to the target network being different from the preset network; the sending module 63 is used to send a first model configuration instruction to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction.

[0099] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0100] Please refer to the model loading method for the functions performed by each module, which will not be repeated here.

[0101] See also Figure 7 , Figure 7 : is a schematic diagram of the structure of an embodiment of an electronic device of the present application. The electronic device 70 includes a memory 71 and a processor 72, and the processor 72 is used to execute program instructions stored in the memory 71 to implement the steps in the above-mentioned model loading method embodiment. In a specific implementation scenario, the electronic device 70 may include but is not limited to: a microcomputer, a server, and in addition, the electronic device 70 may also include a mobile device such as a laptop computer and a tablet computer, which is not limited here.

[0102] Specifically, the processor 72 is used to control itself and the memory 71 to implement the steps in the above-mentioned model loading method embodiment. The processor 72 can also be called a CPU (Central Processing Unit). The processor 72 may be an integrated circuit chip with signal processing capabilities. The processor 72 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 72 can be implemented by an integrated circuit chip.

[0103] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0104] See also Figure 8 , Figure 8 The computer-readable storage medium 80 stores program instructions 801 thereon, and when the program instructions 801 are executed by a processor, the steps in any of the above-mentioned model loading method embodiments are implemented.

[0105] In the above scheme, the target device includes a main card and at least one sub-card. The main card is started in response to the target device where the main card is located. In response to the target network connected to the target device being different from the preset network, a communication connection is established between the main card and each sub-card in the target device, and a first model configuration instruction is sent to each sub-card so that each sub-card performs data analysis using the algorithm model corresponding to the model configuration instruction after receiving the model configuration instruction. Compared with the situation where the sub-card cannot perform data analysis due to network conflicts, the present application can improve the effectiveness of model loading.

[0106] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0107] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0108] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0109] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0110] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

Claims

1. A model loading method, It is characterized in that The method comprises: The main card is responsive to the target device where the main card is located being started, and determines whether a target network accessed by the target device is the same as a preset network, wherein the target device includes the main card and at least one sub-card; In response to the target network being different from the preset network, establishing a communication connection between the main card and each of the sub-cards in the target device; The first model configuration instruction is sent to each of the sub-cards so that each of the sub-cards performs data analysis using the algorithm model corresponding to the first model configuration instruction after receiving the first model configuration instruction.

2. The method according to claim 1, It is characterized in that The method further comprises: In response to the target network being the same as the preset network, adjusting the preset network to another network, where the other network is different from the target network; The target device is controlled to restart, wherein the network to which the target device accesses after restarting is the target network.

3. The method according to claim 1, It is characterized in that The establishing of a communication connection between the main card and each of the sub-cards in the target device includes: The main card creates a virtual network bridge and virtual network cards corresponding to each of the sub-cards; Binding each of the virtual network cards to each virtual port in the virtual network bridge; The configuration information of the virtual network card is sent to each of the sub-cards, so that each of the sub-cards establishes a connection with each of the virtual network cards based on the configuration information of the virtual network card, wherein the main card communicates with each of the sub-cards based on the virtual network bridge.

4. The method according to claim 3, It is characterized in that The method further comprises: The master card traverses the model directory to determine the first name of the algorithm model that the target device can run, and the model directory is used to record the first name of the algorithm model that the target device can run; Sending the first name of each algorithm model to a preset recipient, so that the preset recipient displays the first name of each algorithm model on a configuration interface of the preset recipient; receiving a first model configuration instruction sent by the preset recipient, wherein the first model configuration instruction is issued by the preset recipient in response to a selection operation on a displayed algorithm model; The sending the first model configuration instruction to each of the sub-cards includes: The first model configuration instruction is sent to the virtual network bridge so that the virtual network bridge forwards the first model configuration instruction to each of the daughter cards.

5. The method according to claim 4, It is characterized in that The first model configuration instruction also includes parameter information of the data to be analyzed, and the method further includes: receiving data analysis results of each of the sub-cards, each of the data analysis results being obtained by each of the sub-cards performing data analysis on the data to be analyzed based on the parameter information after running an algorithm model; Forward each of the data analysis results to the preset recipient so that the preset recipient displays the data analysis result.

6. The method according to claim 4, It is characterized in that The method further comprises: Receiving a newly added algorithm model sent by the preset recipient, where the newly added algorithm model is an algorithm model imported by the preset recipient; Install the newly added algorithm model to the model directory to obtain a new model directory; The main card traverses the new model directory to determine the second name of the algorithm model that the target device can currently run; The name of the algorithm model that the target device can currently run is sent to a preset recipient, so that the preset recipient displays the second name on a configuration interface of the preset recipient.

7. The method according to claim 6, It is characterized in that After the main card traverses the new model directory to determine the second name of the algorithm model that the target device can run, the method further includes: Sending a new notification corresponding to the new algorithm model to each of the sub-cards so that each of the sub-cards loads the new algorithm model; And / or, receiving a second model configuration instruction sent by the preset recipient, wherein the second model configuration instruction is issued by the preset recipient in response to a second selection operation on the displayed algorithm model; sending the second model configuration instruction to the virtual bridge so that the virtual bridge forwards the second model configuration instruction to each of the sub-cards.

8. The method according to claim 4, It is characterized in that Before the main card traverses the model directory to determine the first name of the algorithm model that can be run by the target device, the method also includes: Through the virtual bridge, the main card mounts the model directory to each of the sub-cards, so that each of the sub-cards loads each algorithm model in the model directory.

9. An electronic device, It is characterized in that include: A memory and a processor, wherein the memory stores program instructions, and the processor retrieves the program instructions from the memory to execute the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, It is characterized in that include: A program file is stored, and when the program file is executed by a processor, it is used to implement the method according to any one of claims 1 to 8.

Images