Industrial model platform implementation method and system

Abstract

The application provides an industrial model platform implementation method and system, comprising the following steps: step 1: submitting industrial model running code, model parameter files and image construction command files to a code warehouse according to preset specifications; step 2: compiling the industrial model running code and the model parameter files in the code warehouse into images and storing the images in an image warehouse of the platform; step 3: auditing industrial models that need to be shared on the platform, and putting the industrial models on the platform and displaying the industrial models after the auditing; and step 4: running the industrial models on the platform, and calling and managing the industrial models through a representational state transfer application program interface. The method solves the problem that industrial models are separated from each other and independent, and the same mechanism industrial models are repeatedly developed, enables the platform to integrate the achievements of various industrial model research and development units, and realizes co-creation, sharing and intercommunication of the industrial models.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of model platformization, in particular, to an industrial model platformization implementation method and system. BACKGROUND

[0002] Enterprises manage digital assets through various information systems, but there is no unified system for unified management. The data of each information system is separated from each other, and there is no suitable way to make the digital assets of the enterprise become the iterative update of the technical inheritance and experience of the enterprise.

[0003] Patent document CN109754014A (application number: CN201811654236.1) discloses an industrial model training method, device, equipment and medium, including obtaining an industrial model and training data, configuring initial parameters of the industrial model; input the training data into the industrial model, train the industrial model, and obtain the training result; determine whether the training result meets the preset condition, if not, adjust the parameters of the industrial model according to the training result, retrain the industrial model after adjusting the parameters, until the training result meets the preset condition. However, this patent does not describe the process of model platformization and the running process of industrial model.

[0004] Patent document CN105549982B (application number: CN201610024028.8) discloses an automatic development platform based on model configuration, including a visual work engine module and a database engine. The visual work engine module is internally divided according to the system implementation function, and each part contains front-end and back-end related logic. It is implemented according to the agile development mode, and the parts are loosely coupled. The database engine includes a database connection manager, a sql manager, a transaction manager and an external interface, which is used to create and manage database connections and maintain data persistence. The sql manager is used to generate corresponding sql objects according to the user configured data model and pass them to the tbl class in the visual work engine module. The tbl class executes the sql and responds accordingly. However, this patent does not describe the process of model platformization, and cannot realize the co-creation and sharing of industrial models. SUMMARY

[0005] In view of the defects in the prior art, the purpose of the present application is to provide an industrial model platformization implementation method and system.

[0006] According to the industrial model platformization implementation method provided by the present application, the following steps are included:

[0007] Step 1: submit the industrial model running code, model parameter file and image construction command file to the code repository according to the preset specification;

[0008] Step 2: Compile the industrial model running code and model parameter file in the code repository into an image and store it in the image repository of the platform;

[0009] Step 3: Review the industrial model to be shared on the platform, and after review, put the industrial model on the shelf and display it;

[0010] Step 4: Run the industrial model on the platform and call and manage the industrial model through the representational state transfer application interface.

[0011] Preferably, the step 1 comprises:

[0012] Before submitting the industrial model to the code repository, record the parameters of the industrial model and submit the application form to the upload platform;

[0013] The parameters of the industrial model include the use scenario, accuracy, throughput performance, interface logic, interface example and environment dependence of the industrial model;

[0014] The content of the application form includes the interface person and contact information of the application unit.

[0015] Preferably, the step 2 comprises:

[0016] Step 2.1: Compile the industrial model running code and model parameter file into an image according to the image construction command file in the code repository, and if the compilation fails, notify the application unit interface person to modify the code;

[0017] Step 2.2: After successful image compilation, push the image to the image repository of the platform for storage.

[0018] Preferably, the step 3 comprises:

[0019] Step 3.1: Manage the uploaded industrial model running code and image. The industrial model state includes shelving, pending review and shelving. The state of the uploaded industrial model is shelving and is not displayed on the platform. Select the model to be shared on the platform and submit the shelving application. The shelving application includes the CPU, memory, GPU core and video memory resource size required for image running. Update the state of the industrial model submitted for shelving to pending review;

[0020] Step 3.2: Run test and interface test on the image. If the test fails, notify the application unit interface person to modify the code;

[0021] Step 3.3: After the test passes, update the model state to shelving and display the industrial model information on the platform for users to try or purchase the shelved industrial model on the platform.

[0022] Preferably, the step 4 comprises:

[0023] Step 4.1: the user runs the industrial model uploaded by himself or tries and purchases the industrial model uploaded by other users, if the resources required by the industrial model running are less than the remaining resources of the user, the user needs to purchase resources to restart the model;

[0024] Step 4.2: the image is run on the available running resources through the platform resource scheduling engine;

[0025] Step 4.3: after the running is successful, the running state of the image is continuously monitored;

[0026] Step 4.4: the user obtains the model inference result through the interface call.

[0027] According to the industrial model platform implementation system provided by the application, the system comprises:

[0028] Module M1: the industrial model running code, the model parameter file and the image construction command file are submitted to the code warehouse according to the preset specification;

[0029] Module M2: the industrial model running code and the model parameter file in the code warehouse are compiled into an image and stored in the image warehouse of the platform;

[0030] Module M3: the industrial model to be shared on the platform is audited, and after the audit, the industrial model is put on the shelf and displayed;

[0031] Module M4: the industrial model is run on the platform, and the calling and management of the industrial model are performed through the representational state transfer application program interface.

[0032] Preferably, the module M1 comprises:

[0033] Before the industrial model is submitted to the code warehouse, the parameters of the industrial model are recorded and an application form of the uploading platform is submitted;

[0034] The parameters of the industrial model comprise the use scene, the accuracy, the throughput performance, the interface logic, the interface example and the environment dependence of the industrial model;

[0035] The content of the application form comprises the interface person and the contact information of the application unit.

[0036] Preferably, the module M2 comprises:

[0037] Module M2.1: according to the image construction command file in the code warehouse, the industrial model running code and the model parameter file are compiled into an image, if the compilation fails, the application unit interface person is informed, and the code is modified by the application unit;

[0038] Module M2.2: after the mirror image is compiled successfully, the mirror image is pushed into the mirror image warehouse of the platform for storage.

[0039] Preferably, the module M3 comprises:

[0040] Module M3.1: the uploaded industrial model running code and mirror image are managed, the industrial model state comprises delisting, waiting for auditing and shelving, the state of the uploaded industrial model is delisting, and the industrial model is not displayed on the platform; the model to be shared on the platform is selected and a shelving application is submitted; the shelving application comprises the CPU, memory, GPU core and video memory resource size required by the mirror image running; the state of the industrial model submitted in the shelving application is updated to waiting for auditing;

[0041] Module M3.2: the mirror image is tested for running and interface, if the test fails, the interface person of the application unit is informed, and the code is modified by the application unit;

[0042] Module M3.3: after the test passes, the model state is updated to shelving, and the industrial model information is displayed on the platform for users to try or purchase the shelved industrial model on the platform.

[0043] Preferably, the module M4 comprises:

[0044] Module M4.1: the user runs the industrial model uploaded by himself or tries and purchases the industrial model uploaded by other users; if the resource required by the industrial model running is less than the remaining resource of the user, the resource needs to be purchased to restart the model;

[0045] Module M4.2: the mirror image is run on the available running resource through the platform resource scheduling engine;

[0046] Module M4.3: after the running is successful, the mirror image running state is continuously monitored;

[0047] Module M4.4: the user obtains the model inference result through the interface call.

[0048] Compared with the prior art, the present application has the following beneficial effects:

[0049] The present application adopts a standardized interface format, a data processing method and a mirror image development specification, and unifies the registration, management and running of industrial models under different programming languages and different frameworks on the platform, solves the problem that industrial models are separated from each other and independent, and the same mechanism industrial models are repeatedly developed, enables the platform to integrate the achievements of various industrial model research and development units, and realizes the co-creation and sharing of industrial models. BRIEF DESCRIPTION OF DRAWINGS

[0050] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0051] Figure 1 For model platformization flowchart;

[0052] Figure 2 For model running service flowchart;

[0053] Figure 3 For platform architecture diagram. DETAILED DESCRIPTION

[0054] The application will be described in detail below with specific embodiments. The following examples will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the application. These are within the scope of the present application.

[0055] Embodiment:

[0056] The application provides an industrial model platformization system, which adopts standardized interface format, data processing method, code development specification, and unified code management and image management components, so that industrial models under different programming languages and different frameworks can be uniformly registered, managed and run on the platform.

[0057] As Figure 1 and Figure 2 The specific implementation steps are as follows:

[0058] Step 1: submit the industrial model running code, model parameter file and image construction related command file to the code repository according to the specification;

[0059] Step 2: the platform compiles the code and model parameter file in the code repository into an image and stores it in the image repository of the platform;

[0060] Step 3: manage the industrial models on the platform, select the models to be shared on the platform, and after the platform administrator's review, the industrial models are displayed on the platform;

[0061] Step 4: run the industrial model on the platform and call the industrial model through the standard RESTAPI.

[0062] The step 1 includes: after an industrial model research and development unit develops an industrial model, the industrial model platform administrator submits an industrial model platform application form, the application form contains the interface person of the application unit, the contact information and the use scene, the accuracy, the throughput performance, the interface logic, the interface example and the environment dependence of the industrial model; after the administrator approves the application, the research and development unit develops the industrial model image code according to the platform specification, and after the code development is completed, the research and development unit uploads the code to the code repository at the specified address on the platform.

[0063] Code contains:

[0064] 1. The industrial model running code is based on the REST API interface calling service, the platform has complete service development specifications and interface calling service code that can be quickly integrated, the applicant only needs to complete the model loading, data preprocessing, model calling, and data post-processing part of the code, and integrate it into the interface calling service code provided by the platform according to the specification;

[0065] 2. The model parameter file is a configuration variable inside the model required for model loading, and the output result is obtained by calculating the input data according to the model parameters when calling the model;

[0066] 3. The image building command file contains a text file used to build an image, and the text content contains instructions and descriptions required for building an image. Optionally, it can contain installation packages required for environment installation, modules and codes required for program running, etc.

[0067] Step 2 includes: the platform compiles the service running code and the model parameter file into an image according to the image building related command file in the code repository. If the compilation fails, the system automatically notifies the interface person of the applicant unit, and the applicant unit modifies the code. After the image compilation is successful, the image is pushed to the image repository of the platform for storage.

[0068] Step 3 includes: the applicant manages the uploaded industrial model code and image, and the industrial model state has the following states: unlisted, pending review, and listed. The model after uploading is in the unlisted state and is not displayed on the platform for other users. The applicant selects the model to be shared on the platform and submits a listing application. The application includes the CPU, memory, GPU core, and video memory resource size required for image running. The model state after submitting the listing application is pending review. The platform image tester performs running test and interface test on the image to ensure the availability of the image service and interface. If the test fails, the system automatically notifies the interface person of the applicant unit, and the applicant unit modifies the code. After the platform administrator approves, the model state is updated to listed, and the industrial model information is displayed on the platform. Other users can try or purchase the listed industrial model on the page.

[0069] As Figure 3 , the platform includes a front-end interface for displaying industrial models; the platform has permission management to manage code repositories, compilation environments, image repositories, and container scheduling;

[0070] The step 4 comprises that the platform user can run the industrial model uploaded by himself or the industrial model uploaded by other users for trial and purchase; if the resources required by the industrial model running are less than the remaining resources of the user, the platform user needs to purchase resources, and restarts the model after the resource purchase is successful; starting the industrial model is to run the image on the available running resources by the platform resource scheduling engine; after the running is successful, the platform continuously monitors the image running state; the user obtains the model inference result by calling the interface.

[0071] The industrial model platform implementation system provided by the application comprises modules M1, M2, M3 and M4.

[0072] The module M1 comprises the following steps: before the industrial model is submitted to the code repository, the parameters of the industrial model are recorded, and an application form is submitted to the uploading platform; the parameters of the industrial model comprise the use scene, accuracy, throughput performance, interface logic, interface example and environment dependence of the industrial model; the content of the application form comprises the interface person and contact information of the application unit.

[0073] The module M2 comprises the following modules M2.1 and M2.2.

[0074] The module M3 comprises the following modules M3.1, M3.2 and M3.3.

[0075] The module M4 includes: module M4.1: a user runs an industrial model uploaded by himself or tries and purchases an industrial model uploaded by another user, if the resource required by the industrial model running is less than the remaining resource of the user, the user needs to purchase resource to restart the model; module M4.2: running the image on the available running resource through the platform resource scheduling engine; module M4.3: continuously monitoring the running state of the image after the running is successful; module M4.4: the user obtains the model inference result through the interface call.

[0076] Those skilled in the art know that, in addition to implementing the system, device and each module thereof provided by the present application in the form of pure computer readable program code, the same program can also be realized in the form of logic gate, switch, application specific integrated circuit, programmable logic controller and embedded microcontroller by logically programming the method steps. Therefore, the system, device and each module thereof provided by the present application can be considered as a hardware component, and the modules included therein for realizing various programs can also be considered as structures in the hardware component; the modules for realizing various functions can also be considered as both software programs for realizing methods and structures in the hardware component.

[0077] The specific embodiments of the present application are described above. It needs to be understood that the present application is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essential content of the present application. The embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily without conflict.

Claims

1. An industrial model platform implementation method, characterized by, The method comprises the following steps: Step 1: submitting industrial model running code, model parameter files and image construction command files to a code repository according to preset specifications; Step 2: compiling the industrial model running code and the model parameter files in the code repository into an image and storing the image in an image repository of the platform; Step 3: auditing the industrial model to be shared on the platform, and uploading and displaying the industrial model after the auditing; Step 4: running the industrial model on the platform, and calling and managing the industrial model through a representational state transfer application program interface; The step 1 comprises: Before the industrial model is submitted to the code repository, the parameters of the industrial model are recorded, and an application form of an uploading platform is submitted; The parameters of the industrial model comprise a use scenario, an accuracy, a throughput performance, an interface logic, an interface example and an environmental dependence of the industrial model; The content of the application form comprises an interface person and a contact method of an application unit; The step 2 comprises: Step 2.1: compiling the industrial model running code and the model parameter files into an image according to the image construction command files in the code repository, and notifying the interface person of the application unit to modify the code if the compilation fails; Step 2.2: pushing the image to the image repository of the platform for storage after the image compilation succeeds; The step 3 comprises: Step 3.1: managing the uploaded industrial model running code and the image, and the state of the industrial model comprises the following: unloading, auditing and uploading, the state of the uploaded industrial model is unloading, the uploaded industrial model is not displayed on the platform, the model to be shared on the platform is selected and an uploading application is submitted, the uploading application comprises CPU, memory, GPU core and GPU memory resource sizes required for image running, and the state of the industrial model submitted in the uploading application is updated to auditing; Step 3.2: performing running test and interface test on the image, and notifying the interface person of the application unit to modify the code if the test fails; Step 3.3: updating the state of the model to uploading after the test succeeds, and displaying the industrial model information on the platform, so that a user can try or purchase the uploaded industrial model on the platform; The step 4 comprises: Step 4.1: the user runs the industrial model uploaded by the user or tries and purchases the industrial model uploaded by another user, and the user needs to purchase resources to restart the model if the resources required for running the industrial model are less than the remaining resources of the user; Step 4.2: running the image on available running resources through a platform resource scheduling engine; Step 4.3: continuously monitoring the running state of the image after the running succeeds; Step 4.4: obtaining a model inference result through an interface call by the user.

2. An industrial model platform implementation system, characterized by, The method comprises the following steps: Module M1: submitting industrial model running code, model parameter files and image construction command files to a code repository according to preset specifications; Module M2: compiling the industrial model running code and the model parameter files in the code repository into an image and storing the image in an image repository of the platform; Module M3: auditing the industrial model to be shared on the platform, and uploading and displaying the industrial model after the auditing; Module M4: running the industrial model on the platform, and calling and managing the industrial model through a representational state transfer application program interface; The module M1 comprises: Before submitting the industrial model to the code repository, the parameters of the industrial model are recorded, and an application form of the uploading platform is submitted; The parameters of the industrial model include the use scene, accuracy, throughput performance, interface logic, interface example and environment dependence of the industrial model; The content of the application form includes the interface person and contact information of the application unit; The module M2 includes: Module M2.1: According to the mirror building command file in the code repository, the industrial model running code and the model parameter file are compiled into an image, and if the compilation fails, the application unit interface person is notified, and the code is modified by the application unit; Module M2.2: After the image compilation is successful, the image is pushed to the image repository of the platform for storage; The module M3 includes: Module M3.1: The uploaded industrial model running code and image are managed, and the industrial model state includes off the shelf, pending review and on the shelf, the state of the uploaded industrial model is off the shelf, and it is not displayed on the platform, the model to be shared on the platform is selected and the on-shelf application is submitted, the on-shelf application includes the CPU, memory, GPU core and video memory resource size required for image running, and the state of the industrial model submitted for on-shelf application is updated to pending review; Module M3.2: The image is tested for running and interface, and if the test fails, the application unit interface person is notified, and the code is modified by the application unit; Module M3.3: After the test is passed, the model state is updated to on the shelf, and the industrial model information is displayed on the platform for users to try or purchase the on-shelf industrial model on the platform; The module M4 includes: Module M4.1: The user runs the industrial model uploaded by himself or tries and purchases the industrial model uploaded by other users, and if the resource required for the industrial model running is less than the remaining resource of the user, the resource needs to be purchased to restart the model; Module M4.2: The image is run on the available running resource through the platform resource scheduling engine; Module M4.3: After running successfully, the running state of the image is continuously monitored; Module M4.4: The user gets the model inference result through the interface call.

Images