Method and device for managing running of algorithm model, storage medium and electronic device

Abstract

Embodiments of the present application provide a method and device for managing and running an algorithm model. The method comprises: in response to a processing instruction for a target algorithm model, obtaining a model file of the target algorithm model from a model warehouse, wherein the model warehouse is a database storing model files of algorithm models; encapsulating the model file of the target algorithm model into a target model service according to an encapsulation template; and submitting the target model service as a target task of a big data platform through an interactive interface, so as to call the target task through a task interface of the big data platform to run the target algorithm model. Therefore, the problem that algorithm models cannot be uniformly managed and run due to manual encapsulation of algorithm models in related technologies can be solved, and the effect of uniformly managing and running algorithm models is achieved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of computers, and in particular, to a method and apparatus for managing and running an algorithm model. BACKGROUND

[0002] With the expansion of the business of financial companies and the continuous iteration of financial technology, financial businesses have introduced a large number of algorithm models based on machine learning and deep learning for risk control, pricing, credit review, anti-fraud, insurance underwriting, and other businesses to facilitate the convenient and accurate development of businesses. Due to the explosive growth of algorithm models, the management and running based on algorithm models have become a problem to be solved.

[0003] Currently, the model management engine usually encapsulates algorithm models into services manually. The manual encapsulation method requires algorithm engineers to deliver trained algorithm models to development engineers, so as to manually encapsulate the models into services and expose interfaces to users. This method not only consumes time and effort, but also disperses algorithm models, which cannot be uniformly managed and versioned. SUMMARY

[0004] Embodiments of the present application provide a method and apparatus for managing and running an algorithm model to at least solve the problem that algorithm models cannot be uniformly managed and run due to manual encapsulation of algorithm models in related technologies.

[0005] According to an embodiment of the present application, a method for managing and running an algorithm model is provided, comprising: in response to a processing instruction for a target algorithm model, obtaining a model file of the target algorithm model from a model warehouse, wherein the model warehouse is a database storing model files of algorithm models; encapsulating the model file of the target algorithm model into a target model service according to an encapsulation template; submitting the target model service as a target task of a big data platform through an interactive interface, so as to call the target task through a task interface of the big data platform to run the target algorithm model.

[0006] In an exemplary embodiment, before obtaining the model file of the target algorithm model from the model warehouse in response to the processing instruction for the target algorithm model, it further comprises: in response to a model submission request generated by an operation interface, obtaining the model file of the target algorithm model carried in the submission request, and submitting the model file of the target algorithm model to the model warehouse through a target interactive protocol, wherein the model warehouse lists and stores the model file according to the model version.

[0007] In an example embodiment, after the model file of the target algorithm model is submitted to the model warehouse through the target interaction protocol, further comprising: in response to a model audit operation received by the operation interface, obtaining the model file of the target algorithm model stored in the model warehouse, and displaying the model file of the target algorithm model in the operation interface; receiving an audit pass instruction generated by the operation interface, instructing the model warehouse to update the state of the model file of the target algorithm model to an audit pass state, wherein the state of the model file of the target algorithm model is stored in the state database of the model warehouse.

[0008] In an example embodiment, encapsulating the model file of the target algorithm model into a target model service according to the packaging template comprises: calling the corresponding packaging template according to the model file of the target algorithm model; parsing the model file of the target algorithm model using the model markup language indicated by the packaging template to obtain a model parsing file; determining a corresponding prediction function according to the file type of the model parsing file, wherein the prediction function is a test function used when initializing the target algorithm model before the big data platform runs the target algorithm model; loading the prediction function into an initialization class of the model parsing file to obtain the target model service.

[0009] In an example embodiment, loading the prediction function into the initialization class of the model parsing file comprises: loading the model parsing file when the target framework starts and executing the initialization of the model parsing file; loading the prediction function into the initialization class using the target framework, wherein the initialization class is a class file related to the initialization of the model parsing file, and the prediction function is a prediction function corresponding to the model type to which the target algorithm model belongs.

[0010] In an example embodiment, submitting the target model service as a target task of the big data platform through an interaction interface comprises: sending a creation request of the target task to the big data platform through the interaction interface; in the case of receiving the target task creation success, submitting the target model service to the target task through the interaction interface.

[0011] In an example embodiment, after submitting the target model service as a target task of the big data platform through an interaction interface, comprising: receiving the port number of the task interface returned by the big data platform through the interaction interface, wherein the task interface is a calling port allocated by the big data platform for the target task.

[0012] According to another embodiment of the present application, there is provided a management and running device of an algorithm model, comprising: an acquisition unit configured to acquire a model file of a target algorithm model from a model warehouse in response to a processing instruction of the target algorithm model, wherein the model warehouse is a database storing model files of algorithm models; an encapsulation unit configured to encapsulate the model file of the target algorithm model into a target model service according to an encapsulation template; and a submission unit configured to submit the target model service as a target task of a big data platform through an interactive interface, so as to call the target task through a task interface of the big data platform to run the target algorithm model.

[0013] According to still another embodiment of the present application, there is also provided a computer readable storage medium having a computer program stored therein, wherein the computer program is configured to perform the steps of any one of the method embodiments when executed.

[0014] According to still another embodiment of the present application, there is also provided an electronic device comprising a memory and a processor, wherein the memory has a computer program stored therein, and the processor is configured to execute the computer program to perform the steps of any one of the method embodiments.

[0015] According to the present application, the model file of the target algorithm model is acquired from the model warehouse in response to the processing instruction of the target algorithm model, wherein the model warehouse is a database storing model files of algorithm models, the model file of the target algorithm model is encapsulated into a target model service according to an encapsulation template, the target model service is submitted as a target task of a big data platform through an interactive interface, and the target task is called through a task interface of the big data platform to run the target algorithm model. The algorithm model submitted to the model warehouse can be automatically encapsulated and submitted as a task of the big data platform, so as to run the algorithm model through the interface calling mode. The algorithm model is uniformly stored, encapsulated and managed, and the algorithm model is run in the form of a task in the big data platform, so as to solve the problem that the algorithm model cannot be uniformly managed and run due to manual encapsulation of the algorithm model in the related art, and achieve the effect of uniformly managing and running the algorithm model. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a hardware structure block diagram of a computer terminal of a management and running method of an algorithm model according to an embodiment of the present application;

[0017] Figure 2 is a flowchart of a management and running method of an algorithm model according to an embodiment of the present application;

[0018] Figure 3 is a flowchart of a management and running method of an algorithm model according to an embodiment of the present application;

[0019] Figure 4 This is a flowchart of the management and operation method of the algorithm model according to an embodiment of the present invention;

[0020] Figure 5 This is a structural block diagram of the management and operation device of the algorithm model according to another embodiment of the present invention. Detailed Implementation

[0021] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0023] To facilitate understanding of the technical solutions provided by this invention, the technical terms involved in the embodiments of this invention will be explained below.

[0024] Model Operation Management Platform: A platform for managing and running machine learning and deep learning algorithm models for consumer finance scenarios.

[0025] Spark is a fast and general-purpose computing engine designed specifically for large-scale data processing, possessing the advantages of Hadoop MapReduce.

[0026] Cloud native is a software architecture concept based on the cloud, and a set of methodologies for software development practices based on the cloud.

[0027] The methods and embodiments provided in this application can be executed on a mobile terminal, a computer terminal, or a similar computing device. Taking running on a computer terminal as an example, Figure 1 This is a hardware structure block diagram of a computer terminal for managing and running the algorithm model according to an embodiment of the present invention. Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor (MCU) or a field-programmable gate array (FPGA)) and a memory 104 for storing data are also shown. The computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1The illustrated structure is only a schematic, and does not limit the structure of the computer terminal described above. For example, the computer terminal can further include more or fewer components than those shown, or have a different configuration of components than those shown. Figure 1 The illustrated structure is only a schematic, and does not limit the structure of the computer terminal described above. For example, the computer terminal can further include more or fewer components than those shown, or have a different configuration of components than those shown. Figure 1 The illustrated structure is only a schematic, and does not limit the structure of the computer terminal described above. For example, the computer terminal can further include more or fewer components than those shown, or have a different configuration of components than those shown.

[0028] The memory 104 can be used to store computer programs, such as software programs of application software and modules, such as the computer program corresponding to the subject behavior permission management method in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, implements the method described above. The memory 104 can include a high-speed random access memory, and can further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor 102, which can be connected to the computer terminal through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0029] The transmission device 106 is used to receive or send data via a network. Specific examples of the network can include a wireless network provided by a communication provider of the computer terminal. In one example, the transmission device 106 includes a network adapter (NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet in a wireless manner.

[0030] Figure 2 The flowchart of the management and operation method of the algorithm model of the embodiments of the present application is shown in FIG. 2, which includes the following steps: Figure 3

[0031] Step S202, in response to a processing instruction of a target algorithm model, a model file of the target algorithm model is obtained from a model warehouse, wherein the model warehouse is a database storing model files of algorithm models;

[0032] Step S204, the model file of the target algorithm model is packaged into a target model service according to a packaging template;

[0033] Step S206, the target model service is submitted as a target task of a big data platform through an interactive interface, so as to call the target task through a task interface of the big data platform to run the target algorithm model.

[0034] ​The management operation of the algorithm model is not limited to the algorithm model management platform, and the original model file of the algorithm model is obtained by interacting with the model warehouse, and is automatically packaged into a model service, and the packaged model service is submitted to the big data platform.

[0035] The model warehouse is used to store the model file of the trained algorithm model, and the algorithm model management platform is not limited to providing the submission and review functions of the model file, and is used to store the trained model file to the model warehouse and review the submitted model file to determine whether the model file is a runnable algorithm model.

[0036] In an example embodiment, before obtaining the model file of the target algorithm model from the model warehouse in response to the processing instruction of the target algorithm model, further comprising:

[0037] S11, in response to the model submission request generated by the operation interface, obtaining the model file of the target algorithm model carried in the submission request,

[0038] S12, submitting the model file of the target algorithm model to the model warehouse through the target interaction protocol, wherein the model warehouse lists and stores the model file according to the model version.

[0039] The operation interface is not limited to the front-end visual interface provided by the algorithm model management platform, and is used to upload the model file of the algorithm model to the management platform. After the management platform receives the model file transmitted by the front-end, the model file of the algorithm model is submitted to the model warehouse through SSH or HTTP protocol.

[0040] The model warehouse is not limited to selecting various object storage, for example, selecting GitLib as the storage warehouse of the algorithm model, and is not limited to simultaneously using a state database to record the storage state of the model file, for example, recording whether the model file is in an unreviewed state, a passed state or a failed state.

[0041] In an example embodiment, after submitting the model file of the target algorithm model to the model warehouse through the target interaction protocol, further comprising:

[0042] S21, in response to the model review operation received by the operation interface, obtaining the model file of the target algorithm model stored in the model warehouse, and displaying the model file of the target algorithm model in the operation interface;

[0043] S22, receiving the review pass instruction generated by the operation interface, instructing the model warehouse to update the state of the model file of the target algorithm model to the review pass state, wherein the state of the model file of the target algorithm model is stored in the state database of the model warehouse.

[0044] The model auditing of the algorithm model is not limited to being performed through the auditing visual interface of the management platform, the model file of the algorithm model is called from the model warehouse, the model file is displayed through the auditing interface, and a download or online review function is not limited to being provided to perform the auditing of the model file.

[0045] When the model file is submitted to the model warehouse through the management platform, that is, when the model file is not audited, the model file is not limited to being automatically marked as an unaudited state, the state of the model file is updated according to an auditing result of the model file, the model file is updated to a passed state in a case where the auditing is passed, and the model file is updated to a failed state in a case where the auditing is failed. The state database is not limited to using a Mysql database.

[0046] In one example embodiment, as shown in FIG. 2, step S204 encapsulates the model file of the target algorithm model into a target model service according to an encapsulation template, including: Figure 3

[0047] S204-1, calling a corresponding encapsulation template according to the model file of the target algorithm model;

[0048] S204-2, parsing the model file of the target algorithm model by using a model markup language indicated by the encapsulation template to obtain a model parsing file;

[0049] S204-3, determining a corresponding prediction function according to a file type of the model parsing file, wherein the prediction function is a test function used when the target algorithm model is initialized before the target algorithm model is run on the big data platform;

[0050] S204-4, loading the prediction function into an initialization class of the model parsing file to obtain the target model service.

[0051] The encapsulation template is not limited to being a template corresponding to the file type of the model file, and is used to encapsulate the model file according to the encapsulation template. The model file is parsed by using the model markup language indicated by the encapsulation template, and the prediction function is recorded when the model file is initialized, so that the model file is encapsulated into the model service.

[0052] The model file is not limited to being pulled from the model warehouse by the backend service of the model management engine through an SSH or HTTP protocol, and the PMML format model file is parsed by using a JPMML class library of Java.

[0053] In one example embodiment, step S204-4 loads the prediction function into the initialization class of the model parsing file, including:

[0054] S204-41, loading the model parsing file when the target framework is started, and performing initialization of the model parsing file;

[0055] ​S204-42, loading the prediction function into the initialization class by using the target framework, wherein the initialization class is a class file related to the initialization of the model parsing file, and the prediction function is a prediction function corresponding to the model type of the target algorithm model.

[0056] The target framework is not limited to selecting the springboot framework. The PMML format parsed model file is loaded and initialized when the springboot is started. The prediction function is defined to load the parsed model file, and the prediction function is loaded into the initialization class.

[0057] In an example embodiment, step S206 submits the target model service as a target task of the big data platform through the interactive interface, including:

[0058] S31, sending a creation request of the target task to the big data platform through the interactive interface;

[0059] S32, in the case of receiving the target task creation success, the target model service is submitted to the target task through the interactive interface.

[0060] The interactive interface is an interface for the management platform to interact with the big data platform. The model service encapsulated is uploaded as a task in the big data platform through the interactive interface. Specifically, the encapsulated model service is added to the created task by initiating a creation request through the interactive interface. The big data platform is not limited to running the model service in the form of a task by using the model service, and determining the task interface corresponding to the called model service, so as to realize the form of calling the model service through the task interface.

[0061] In an example embodiment, after the target model service is submitted as a target task of the big data platform through the interactive interface, the port number of the task interface returned by the big data platform is received through the interactive interface, wherein the task interface is the calling port allocated by the big data platform for the target task.

[0062] Specifically, taking the big data platform selecting Spark as an example, the automatically encapsulated model service is submitted as a task by using the API of SparkLauncher provided by Spark through the management platform, and the model service is started by Spark to run in Spark, and is exposed as the port number specified by the model service, facilitating the calling of the model service through the specified port number. When predicting batch data extracted from the big data platform or processing multiple tasks, MapReduce operation can be used for distributed computing.

[0063] The management and running process of the algorithm model is not limited to as Figure 4The management platform 400 receives the trained model file and stores the received model file to the model warehouse 410. The model file is audited through the audit interface, which is not limited to calling the model file from the model warehouse and updating the state of the model file after the audit, from the un-audited state to the audit passed state or the audit failed state. The model file in the audit failed state is not limited to not being automatically packaged. The model file is pulled from the model warehouse for the model file that passes the audit, and is packaged into a model service according to a packaging template. The management platform 400 submits the model service to a spark task in the big data platform 420 after packaging the model file into the model service, the big data platform 420 runs the model service, and exposes the corresponding task interface. The model management engine is built on the big data platform, and combines the advantages of the big data platform in processing large-scale data to realize good fitting of the model service and the data, thereby reducing the difficulty of data collection, preprocessing and batch data transmission and processing.

[0064] In the above embodiments of the present application, the model file of the target algorithm model is obtained from the model warehouse based on the processing instruction for the target algorithm model, the model warehouse is a database for storing the model file of the algorithm model, the model file of the target algorithm model is packaged into a target model service according to a packaging template, the target model service is submitted as a target task of the big data platform through an interactive interface, and the target task is called through the task interface of the big data platform to run the target algorithm model. The algorithm model submitted to the model warehouse can be automatically packaged and submitted as a task of the big data platform, and the algorithm model is run in the form of a task in the big data platform, so that the problem that the algorithm model cannot be uniformly managed and run due to manual packaging of the algorithm model in the related art can be solved, and the effect of uniformly managing and running the algorithm model is achieved.

[0065] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and a general hardware platform as necessary, and of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory (Read-Only Memory / Random Access Memory, ROM / RAM), a magnetic disk, an optical disk), and includes a plurality of instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the method described in each embodiment of the present application.

[0066] There is also provided in the embodiment an algorithm model management and running device for implementing the above embodiments and preferred embodiments, which has been described and will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware or a combination of software and hardware is also possible and contemplated.

[0067] Figure 5 is a structural block diagram of an algorithm model management and running device according to an embodiment of the application, as shown in the figure, the device comprises: Figure 5

[0068] An acquisition unit 502 is configured to acquire a model file of a target algorithm model from a model warehouse in response to a processing instruction for the target algorithm model, wherein the model warehouse is a database storing model files of algorithm models;

[0069] A packaging unit 504 is configured to package the model file of the target algorithm model into a target model service according to a packaging template;

[0070] A submission unit 506 is configured to submit the target model service as a target task of a big data platform through an interactive interface, so as to call the target task through a task interface of the big data platform to run the target algorithm model.

[0071] The algorithm model management and running device comprises all the modules shown in Figure 5 , and further comprises a storage unit configured to, before acquiring the model file of the target algorithm model from the model warehouse in response to the processing instruction for the target algorithm model, acquire a model file of a target algorithm model carried in a model submission request generated by an operation interface in response to the model submission request, and submit the model file of the target algorithm model to the model warehouse through a target interactive protocol, wherein the model warehouse lists and stores the model files according to model versions.

[0072] The algorithm model management and running device comprises all the modules shown in Figure 5 , and further comprises an audit unit configured to, after submitting the model file of the target algorithm model to the model warehouse through the target interactive protocol, acquire the model file of the target algorithm model stored in the model warehouse in response to a model audit operation received by the operation interface, and display the model file of the target algorithm model in the operation interface; receive an audit pass instruction generated by the operation interface, and instruct the model warehouse to update a state of the model file of the target algorithm model to an audit pass state, wherein the state of the model file of the target algorithm model is stored in a state database of the model warehouse in association.

[0073] The packaging unit in the algorithm model management and running device further comprises: ​

[0074] The calling module is used to call the corresponding encapsulation template based on the model file of the target algorithm model;

[0075] The parsing module is used to parse the model file of the target algorithm model using the model markup language indicated by the encapsulated template, and obtain the model parsing file;

[0076] The determination module is used to determine the corresponding prediction function based on the file type of the model parsing file. The prediction function is a test function used when initializing the target algorithm model before the big data platform runs the target algorithm model.

[0077] The loading module is used to load the prediction function into the initialization class of the model parsing file to obtain the target model service.

[0078] The loading module in the management and operation device of the algorithm model is also used to load the model parsing file when the target framework starts and to perform the initialization of the model parsing file; the target framework is used to load the prediction function into the initialization class, where the initialization class is the class file related to the initialization of the model parsing file, and the prediction function is the prediction function corresponding to the model type to which the target algorithm model belongs.

[0079] The submission unit in the management and operation device of the algorithm model includes: sending a request to create the target task to the big data platform through an interactive interface; and, upon receiving a successful request to create the target task, providing the target model service to the target task through the interactive interface.

[0080] The management and operation device of this algorithm model includes, in addition to Figure 5 In addition to all the modules shown, there is also an audit unit and a receiving unit, which are used to receive the port number of the task interface returned by the big data platform after the target model service is submitted as the target task of the big data platform through the interaction interface. The task interface is the calling port allocated by the big data platform for the target task.

[0081] In the above embodiment of the present application, the model file of the target algorithm model is acquired from the model warehouse based on the processing instruction for the target algorithm model, the model file of the target algorithm model is encapsulated into the target model service according to the encapsulation template, the target model service is submitted as the target task of the big data platform through the interactive interface, and the target task is called through the task interface of the big data platform to run the target algorithm model. The algorithm model submitted to the model warehouse can be automatically encapsulated and submitted as the task of the big data platform, and the algorithm model is run in the form of the task through the interface calling mode. The algorithm model is uniformly stored, encapsulated and run, and the algorithm model is run in the form of the task in the big data platform. Therefore, the problem that the algorithm model cannot be uniformly managed and run due to manual encapsulation of the algorithm model in the related art can be solved, and the effect of uniformly managing and running the algorithm model is achieved.

[0082] It should be noted that the above modules can be implemented by software or hardware. For the latter, the following implementation manners can be used, but are not limited thereto: the above modules are located in the same processor; or the above modules are located in different processors in any combination.

[0083] Embodiments of the present application also provide a computer readable storage medium, which stores a computer program. The computer program is configured to execute the steps in any of the above method embodiments when running.

[0084] In an example embodiment, the above computer readable storage medium can include, but is not limited to, a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0085] Embodiments of the present application also provide an electronic device, which includes a memory storing a computer program and a processor configured to execute the computer program to perform the steps in any of the above method embodiments.

[0086] In an example embodiment, the above electronic device can further include a transmission device connected to the processor and an input / output device connected to the processor.

[0087] The specific examples in the present embodiment can refer to the examples described in the above embodiments and example embodiments, which will not be described herein again.

[0088] It should be apparent to those skilled in the art that the modules or steps of the application described above can be implemented with general computing devices, which can be centralized on a single computing device or distributed on a network of multiple computing devices, which can be implemented with program codes executable by the computing devices, so that they can be stored in storage devices and executed by the computing devices, and in some cases, the steps shown or described can be executed in different orders than shown, or made into individual integrated circuit modules, or made into a single integrated circuit module. Thus, the application is not limited to any particular combination of hardware and software.

[0089] The preferred embodiments of the application described above are intended to be merely exemplary and those skilled in the art will readily suggest modifications and variations to the above-described embodiments. Any modification, equivalent replacement, improvement, etc. made within the principles of the application should be included in the protection scope of the application.

Claims

1. A method for managing and running an algorithm model, characterized in that, include: In response to a processing instruction for a target algorithm model, the model file of the target algorithm model is retrieved from a model repository, wherein the model repository is a database that stores model files of algorithm models; The target algorithm model's model file is packaged into a target model service according to the packaging template; The target model service is submitted as a target task of the big data platform through the interactive interface, so that the target task can be called through the task interface of the big data platform to run the target algorithm model; The step of encapsulating the model file of the target algorithm model into a target model service according to the encapsulation template includes: Based on the model file of the target algorithm model, call the corresponding encapsulation template; The model file of the target algorithm model is parsed using the model markup language indicated by the encapsulation template to obtain the model parsing file; The corresponding prediction function is determined based on the file type of the model parsing file, wherein the prediction function is a test function used when initializing the target algorithm model before the big data platform runs the target algorithm model; The prediction function is loaded into the initialization class of the model parsing file to obtain the target model service; The process of loading the prediction function into the initialization class of the model parsing file includes: The model parsing file is loaded when the target framework starts, and the initialization of the model parsing file is performed. The prediction function is loaded into the initialization class using the target framework, wherein the initialization class is a class file related to the initialization of the model parsing file, and the prediction function is the prediction function corresponding to the model type to which the target algorithm model belongs.

2. The method according to claim 1, characterized in that, Before retrieving the model file of the target algorithm model from the model repository in response to the processing instruction for the target algorithm model, the method further includes: In response to a model submission request generated by the user interface, the model file of the target algorithm model carried in the submission request is obtained. The target algorithm model's model file is submitted to the model repository through the target interaction protocol, wherein the model repository manages and stores the model file in a list-style manner according to the model version.

3. The method according to claim 2, characterized in that, After submitting the model file of the target algorithm model to the model repository via the target interaction protocol, the process also includes: In response to the model review operation received by the operation interface, the model file of the target algorithm model stored in the model repository is obtained and displayed in the operation interface; The system receives an approval instruction generated by the operation interface, instructing the model repository to update the status of the model file of the target algorithm model to the approval status. The status of the model file of the target algorithm model is associated and stored in the status database of the model repository.

4. The method according to claim 1, characterized in that, Submitting the target model service as a target task of the big data platform through the interactive interface includes: The creation request for the target task is sent to the big data platform through the interactive interface; Upon receiving confirmation that the target task has been successfully created, the target model service is provided to the target task through the interaction interface.

5. The method according to claim 1, characterized in that, After submitting the target model service as a target task of the big data platform through the interactive interface, the process includes: The interaction interface receives the port number of the task interface returned by the big data platform, wherein the task interface is the calling port allocated by the big data platform for the target task.

6. A management and operation device for an algorithm model, characterized in that, include: The acquisition unit is configured to, in response to a processing instruction for the target algorithm model, acquire the model file of the target algorithm model from a model repository, wherein the model repository is a database storing model files of algorithm models; The encapsulation unit is used to encapsulate the model file of the target algorithm model into a target model service according to the encapsulation template; The submission unit is used to submit the target model service as a target task of the big data platform through an interactive interface, so as to call the target task through the task interface of the big data platform to run the target algorithm model; The encapsulation unit is further configured to: call the corresponding encapsulation template according to the model file of the target algorithm model; The model file of the target algorithm model is parsed using the model markup language indicated by the encapsulation template to obtain the model parsing file; The corresponding prediction function is determined based on the file type of the model parsing file, wherein the prediction function is a test function used when initializing the target algorithm model before the big data platform runs the target algorithm model; The prediction function is loaded into the initialization class of the model parsing file to obtain the target model service; The process of loading the prediction function into the initialization class of the model parsing file includes: The model parsing file is loaded when the target framework starts, and the initialization of the model parsing file is performed. The prediction function is loaded into the initialization class using the target framework, wherein the initialization class is a class file related to the initialization of the model parsing file, and the prediction function is the prediction function corresponding to the model type to which the target algorithm model belongs.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to perform the method described in any one of claims 1 to 5 when executed.

8. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 5.

Images