CN115658129A

CN115658129A - Management method, device, equipment and medium for full-life-cycle learning model

Info

Publication number: CN115658129A
Application number: CN202211358184.XA
Authority: CN
Inventors: 李安杰
Original assignee: China Resources Digital Technology Co Ltd
Current assignee: China Resources Digital Technology Co Ltd
Priority date: 2022-11-01
Filing date: 2022-11-01
Publication date: 2023-01-31

Abstract

The invention relates to the technical field of model management, and discloses a method, a device, equipment and a medium for managing a learning model in a full life cycle. The method comprises initializing a pre-stored system configuration; carrying out approval chain configuration on the system according to the pre-stored system configuration; if receiving the data to be processed input by the user, judging whether the data to be processed is contained in a pre-stored historical database; if the data to be processed is not contained in a pre-stored historical database, a new demand flow is established; acquiring a corresponding model file according to a demand flow; if the intelligent model corresponding to the model file receives the application form of the user, judging whether the judgment result output by the intelligent model to the application form is yes; if the judgment result is yes, outputting a related report and sending the related report; judging whether the effect of the intelligent model meets the offline requirement or not; and if the effect of the intelligent model meets the offline requirement, completing the offline of the intelligent model. The invention standardizes model management and improves the real-time property of model data updating and the safety of model use.

Description

Management method, device, equipment and medium for learning model of full life cycle

Technical Field

The invention relates to the technical field of model management, in particular to a management method of an intelligent model with a full life cycle.

Background

With the rapid development of science and technology, computer learning is more and more helpful to people's work and life at present, the number of models learned by computers also increases as geometric multiples along with the development of business and the transition of time, and meanwhile, a small challenge is brought to model management and model iteration.

At present, most of computer learning records historical models in excels or wiki of each group, accurate notification cannot be performed when model statistics and model offline notification are performed due to non-real-time data, data statistics analysis is kept locally, and risks of untimely statistics, inconsistent multi-person management data, data leakage, model loss and the like exist in the models. Therefore, the problem of model management confusion is to be solved, so as to improve the real-time performance of model data updating and the safety of model use.

Disclosure of Invention

The embodiment of the invention provides a method, a device, equipment and a medium for managing a full-life-cycle learning model, aiming at solving the technical problem of standardizing model management so as to improve the real-time property of model data updating and the safety of model use.

In a first aspect, an embodiment of the present invention provides a management method for a learning model in a full lifecycle, where the method includes:

initializing a pre-stored system configuration;

carrying out approval chain configuration on the system according to the pre-stored system configuration;

if the data to be processed input by a user is received, judging whether the data to be processed is contained in a pre-stored historical database;

if the data to be processed is not contained in a pre-stored historical database, a demand flow corresponding to the data to be processed is newly established;

acquiring a corresponding model file according to the demand flow;

if the intelligent model corresponding to the model file receives an application form of the user for the approval chain, judging whether a judgment result output by the intelligent model for the application form is yes;

if the judgment result output by the intelligent model is yes, outputting a related report and sending the related report;

judging whether the effect of the intelligent model meets the preset offline requirement or not;

and if the effect of the intelligent model meets the offline requirement, completing the offline of the intelligent model.

Preferably, the configuring the system according to the pre-stored system configuration by an approval chain includes:

dividing system levels according to the pre-stored system configuration to obtain a plurality of system spaces;

and respectively configuring an approval chain corresponding to each system space.

Preferably, the determining whether the data to be processed is contained in a pre-stored historical database includes:

analyzing the data to be processed to obtain a corresponding retrieval identifier;

and searching the search identification in the historical database to judge whether the historical database contains data matched with the search identification.

Preferably, the obtaining the corresponding model file according to the demand flow includes:

the requirement process is associated to the approval chain to dynamically generate an approval chain example;

sending the examination and approval chain example to a corresponding approver for examination and approval;

if the flow end information of the examination and approval chain example is acquired, completing the dynamic global marking of the demand flow and the model so as to complete the new construction of the demand flow;

constructing an intelligent model according to the demand flow;

and training the intelligent model so as to output a model file.

Preferably, the intelligent model judges whether the application form meets the approval requirements of the approval chain;

if the application form meets the approval requirement of the approval chain, the judgment result output by the intelligent model is yes, and the approval chain receives confirmation information input by a user end and finishes approval;

triggering automatic service expansion after the approval of the approval chain is passed;

and if the application form does not accord with the approval requirement of the approval chain, judging whether the output result of the intelligent model is negative or not, and receiving judging information whether the judging result is negative or not and returning the application form to the user side by the approval chain.

Preferably, whether the attenuation effect of the intelligent model meets the first effect condition of the offline requirement is judged to obtain a first judgment result;

judging whether the iterative effect of the intelligent model meets a second effect condition of the offline requirement or not to obtain a second judgment result;

and if the first judgment result is yes or the second judgment result is yes, judging that the effect of the intelligent model meets the preset offline requirement.

Preferably, the completing the offline of the intelligent model includes:

the pre-stored system configuration initiates a destroy inference example request to a bottom layer inference engine model pre-stored in the system;

the bottom layer reasoning engine model requests through the destruction reasoning instance;

and the system calls a destructor corresponding to the intelligent model to destroy the intelligent model, and outputs a model offline report and a flow monitoring report to finish the model offline.

In a second aspect, an embodiment of the present invention provides a management apparatus for a full-life-cycle intelligent model, including:

the configuration unit is used for dividing the system hierarchy according to the pre-stored system configuration; configuring a preset approval chain of the model series;

the receiving and judging unit is used for receiving the data to be processed input by the user side and judging whether the data to be processed is contained in a pre-stored historical database;

a flow establishing unit, configured to establish a required flow corresponding to the to-be-processed data if the to-be-processed data is not included in a pre-stored historical database;

the file acquisition unit is used for acquiring a corresponding model file according to the demand flow;

the approval unit is used for allowing the user to approve the approval chain according to the judgment result if the application table meets the approval requirement of the approval chain and the judgment result output by the model is yes;

the model offline unit is used for initiating a reasoning instance destroying request to a bottom layer reasoning engine model prestored in the system by the prestored system configuration; the bottom layer reasoning engine model requests through the destruction reasoning instance; and the system calls a destructor corresponding to the model to destroy the model.

In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the computer program, implements the management method for the full-life-cycle intelligent model described in the first aspect.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the processor is caused to execute the method for managing a full-lifecycle intelligent model according to the first aspect.

The embodiment of the invention provides a method, a device, equipment and a medium for managing a full-life-cycle learning model, wherein the method comprises the following steps: initializing a pre-stored system configuration; carrying out approval chain configuration on the system according to the pre-stored system configuration; if receiving the data to be processed input by the user, judging whether the data to be processed is contained in a pre-stored historical database; if the data to be processed is not contained in the pre-stored historical database, a demand flow corresponding to the data to be processed is newly established; acquiring a corresponding model file according to a demand flow; if the intelligent model corresponding to the model file receives the approval information of the user to the approval chain, outputting a related report and sending the related report; judging whether the effect of the intelligent model meets the preset offline requirement or not; and if the effect of the intelligent model meets the offline requirement, completing the offline of the intelligent model. The management method of the full-life-cycle intelligent model can be applied to model management of a user group consisting of multiple people such as enterprises and schools, so that the model management is standardized, and the real-time performance of model data updating and the safety of model use are improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic flow chart of a method for managing a full-life-cycle intelligent model according to the present invention;

fig. 2 is a schematic view of an application scenario of a management method of a full-life-cycle intelligent model according to an embodiment of the present invention;

FIG. 3 is a sub-flow diagram of a method for managing a full-lifecycle intelligent model according to an embodiment of the present invention;

FIG. 4 is a sub-flow diagram illustrating a management method of a full-lifecycle intelligent model according to an embodiment of the present invention;

FIG. 5 is a schematic view of another sub-flow of a management method of a full-life-cycle intelligent model according to an embodiment of the present invention;

FIG. 6 is a schematic view of another sub-flow of a management method of a full-life-cycle intelligent model according to an embodiment of the present invention;

FIG. 7 is a schematic view of another sub-flow of a management method of a full-life-cycle intelligent model according to an embodiment of the present invention;

FIG. 8 is a schematic diagram illustrating a later sub-flow of a management method for a full-lifecycle intelligent model according to an embodiment of the present invention;

FIG. 9 is a schematic block diagram of a management apparatus for a full lifecycle intelligence model provided by an embodiment of the present invention;

fig. 10 is a schematic block diagram of a computer device provided in an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.

It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

Referring to fig. 1 and fig. 2, fig. 1 is a schematic flowchart of a management method of an intelligent model in a full lifecycle provided in an embodiment of the present invention, and fig. 2 is a schematic application scenario diagram of the management method of an intelligent model in a full lifecycle provided in an embodiment of the present invention. The management method of the full-life-cycle intelligent model is applied to a management server 10, and a user side 20 is connected with the management server 10 through a network to transmit data information; the management method of the full-lifecycle intelligent model is executed through application software installed in a management server 10, the management server 10 is a server side for executing the full-lifecycle intelligent model to analyze data to be processed from a user side 20 to execute the full lifecycle, such as a server side configured in an enterprise, the user side 20 is a terminal device for collecting the data to be processed input by a user and uploading the data to the management server 10, and the user side 20 can be configured in various intelligent devices, such as a tablet computer, a notebook computer, a desktop computer, a mobile phone, and the like.

The user terminal 20 and the management server 10 can directly establish wireless communication connection through 2G/3G/4G/5G communication; the user end 20 may also establish a wireless communication connection with the management server 10 through other terminal devices, for example, after the user end 20 establishes a connection with a terminal device such as a computer, the user end 20 and the management server 10 are connected through the computer, so as to realize bridging between the user end 20 and the management server 10 and transmit data information.

As shown in fig. 1, the management method of the full-life-cycle intelligent model can be applied to model management of a user group consisting of multiple people, such as an enterprise, a school, and the like, to standardize model management and improve the real-time performance of model data update and the safety of model use.

And S110, initializing the pre-stored system configuration.

The pre-storage system configuration comprises six core processes of model establishment, model development, model review, model release, model access and model offline, and torsion judgment rules corresponding to the six core processes respectively; meanwhile, in each core flow, the pre-storage system is configured with three level models for classifying and configuring product lines, the three level models are respectively a project class, a service class and a detailed classification, each level model belongs to a specific product line, and the specific classification in each level model corresponds to an approval chain. In this case, the pre-stored system configuration is initialized by the initialization method function, thereby improving the stability of the model management.

And S120, carrying out approval chain configuration on the system according to the pre-stored system configuration.

An enterprise user firstly combs a business strategy party, a participation team and the examination and approval hierarchical relation of each process node, whether an algorithm evaluation team, a verification team and a model engineering team exist or not needs to be determined, then an initial information file is formed by combing, and the initial information file is input into a system in a management server; the system receives and classifies initial information files, then corresponding processes are configured in six core processes of model establishment, model development, model review, model release, model access and model offline, corresponding product lines, model types and service modes of the product lines are configured in the process of the approval chain, the system generates a specific approval chain according to the configured processes, and then automatic approval nodes are generated, and accordingly the approval chain of the specific product lines is output.

In an embodiment, as shown in fig. 3, step S120 includes substeps S121 and S122.

And S121, obtaining a plurality of system spaces according to the system hierarchy of the pre-stored system configuration. The system receives and classifies the initial information files to obtain a plurality of system spaces.

And S122, respectively configuring an approval chain corresponding to each system space. The system configures corresponding processes in six core processes of model establishment, model development, model review, model release, model access and model offline according to the classification of each system space, configures corresponding product lines and model types and service modes of the product lines in the process of the approval chain, and generates a specific approval chain according to the configured processes.

S130, if the data to be processed input by the user is received, judging whether the data to be processed is contained in a pre-stored historical database.

When the enterprise user uses the intelligent model, the data to be processed is input firstly, in this embodiment, the data to be processed can be a model number, a fuzzy model series, a product name and the like, and after the intelligent model receives the data to be processed, the intelligent model searches a historical database stored in the system to judge whether the historical database contains data matched with the search identifier.

In one embodiment, as shown in fig. 4, step S130 includes sub-steps S131 and S132.

S131, analyzing the data to be processed to obtain a corresponding retrieval identifier. The retrieval identification is identification information which is uniquely corresponding to the data to be processed, wherein the data to be processed comprises a data head, a data main body and a data tail, and a certain section or several sections of the data head, the data main body and the data tail are identified by analyzing the data to be processed so as to obtain the retrieval identification.

S132, retrieving the retrieval identification in the historical database to judge whether the historical database contains data matched with the retrieval identification. The historical database comprises identification information preset by the system and identification information input by a user and received in the using process, and if the historical database contains data matched with the acquired retrieval identification, the identification information in the historical database corresponds to the acquired retrieval identification.

And S140, if the data to be processed is not contained in a pre-stored historical database, establishing a demand flow corresponding to the data to be processed. A submission requirement page preset in the system comprises three hierarchical model options of a model series, a business category and a detailed classification, wherein the three hierarchical model options correspond to a product line respectively and are associated to a corresponding approval chain, directory items of product requirements and specific product names are preset on the approval chain, a starting node of the approval chain receives information such as requirement types, model types, using modes, requirement backgrounds, model application scenes, requirement targets and the like input by a user, relevant requirement information is automatically identified, an approval chain example is dynamically generated, and enterprise approvers such as an algorithm responsible person, an algorithm developer, a verification person and a business butt person related to the requirement information input by the user are automatically matched at a subsequent approval node so as to complete approval of the approval chain.

And S150, acquiring a corresponding model file according to the demand flow.

In one embodiment, as shown in FIG. 5, step S150 includes sub-steps S151-S155.

And S151, associating the requirement process with the approval chain to dynamically generate the approval chain example. Here, the approval chain is a specific approval chain generated by the system according to the configured process, and the approval chain example is an approval chain for actual circulation generated after associating the specific requirement process.

And S152, sending the examination and approval chain example to a corresponding approver for examination and approval. The input end of each approval node receives approval information input by a corresponding approver and identifies the approval information. If the result of the identified examination and approval information is 'pass', the examination and approval chain instance is transferred to the next process node for continuous examination and approval; if the identified examination and approval information result is 'fail', ending the example process of the examination and approval chain; and if the identified result of the examination and approval information is 'returning to a certain approver', and information such as a return reason of remarks is received, the examination and approval chain instance is returned to the corresponding approver to be re-approved.

S153, if the process end information of the examination and approval chain example is obtained, completing the dynamic global marking of the demand process and the model so as to complete the new construction of the demand process. After the requirement process is newly built, a model construction report is formed in the system, and the model construction report comprises a requirement background, a model requirement target, a modeling requirement specification, a modeling scheme and a modeling schedule which are input by an algorithm team. The dynamic global marker can mark which stage of model establishment, model development, model review, model release, model access and model offline of the model, and mark which specific approvers are associated with the model.

And S154, constructing an intelligent model according to the demand flow. The system analyzes the intelligent model construction report to construct an intelligent model, so that a user can conveniently manage the intelligent model.

And S155, training the intelligent model so as to output a model file. In this embodiment, the model training uses a tensrflow, which is a symbolic mathematical system, where the training is performed by an iterator. Training parameters are preset in the intelligent model, a loss function is defined, the loss function can be used for describing errors between predicted values and true values, so that the convergence direction of the intelligent model is guided, and common loss functions comprise mean square error, cross entropy and the like; the intelligent model is also preset with iteration turns, receives data input by a user one by one to perform gradient descent optimization operation steps, and then performs each iteration turn to obtain a model curve. TensorFlow training may also be based on tensor, which is a definition of an array in mathematics, which represents an array. Specifically, the intelligent model receives all data input by a user, then calls a model _ fit function, and specifies a parameter batch _ size to train all data in batches to obtain a model curve. In other embodiments, jupiter Lab, which is the latest web-based interactive development environment, may also be used for model training. Therefore, the use safety of the intelligent model can be improved.

And S160, if the intelligent model corresponding to the model file receives the application form of the approval chain from the user, judging whether the judgment result output by the intelligent model to the application form is yes.

In one embodiment, as shown in FIG. 6, step S160 includes sub-steps S161-S164.

S161, the intelligent model judges whether the application form meets the approval requirement of the approval chain. When an enterprise user needs to use the intelligent model, an application form needs to be filled in at a user side of the system, wherein the content of the application form comprises service request quantity and concurrency quantity, and the intelligent model receives the application form input by the user.

And S162, if the application form meets the approval requirement of the approval chain, the judgment result output by the intelligent model is yes, and the approval chain receives confirmation information input by the user end and completes approval.

And S163, triggering automatic service expansion after the approval chain passes approval. In this embodiment, full lifecycle management of an intelligent model is achieved through kubernets (K8S for short), where K8S is a container organizer, is an open source, and is used to manage containerized applications on multiple hosts in a cloud platform, and K8S makes deploying containerized applications simpler and more efficient. The K8S horizontal automatic scaling manager can realize a set of simple automatic scaling logic, under the default condition, the index is detected every 30S, and as long as the K8S detects the target value of the configuration horizontal automatic scaling manager, the number of copies of the expected workload is calculated, and then the scaling operation is carried out. Meanwhile, in order to avoid too frequent capacity expansion and contraction, K8S defaults that the capacity expansion and contraction can be triggered under the condition that the capacity is not expanded again within five minutes.

And S164, if the application form does not meet the approval requirements of the approval chain, judging whether the result output by the intelligent model is negative or not, and receiving judging information whether the result is negative or not and returning the application form to the user side by the approval chain.

And S170, if the judgment result output by the intelligent model is yes, outputting a relevant report and sending the report. Here, the relevant reports include model review materials and model verification reports.

And S180, judging whether the effect of the intelligent model meets the preset offline requirement or not.

In one embodiment, as shown in FIG. 7, step S180 includes sub-steps S181-S183.

S181, judging whether the attenuation effect of the intelligent model meets the first effect condition of the offline requirement or not, and obtaining a first judgment result.

If the attenuation threshold value can be preset in the first effect condition, whether the attenuation coefficient of the intelligent model is smaller than the attenuation threshold value in the first effect condition can be judged, and whether the attenuation effect of the intelligent model meets the first effect condition is judged.

S182, judging whether the iterative effect of the intelligent model meets a second effect condition of the offline requirement or not to obtain a second judgment result.

If an iteration threshold value can be preset in the second effect condition, whether the iteration coefficient of the intelligent model is larger than the iteration threshold value in the second effect condition can be judged, and whether the iteration effect of the intelligent model meets the second effect condition is judged.

And S183, if the first judgment result is yes or the second judgment result is yes, judging that the effect of the intelligent model meets the preset offline requirement. And if the first judgment result is negative or the second judgment result is negative, the intelligent model does not meet the preset offline requirement, and the intelligent model can be continuously used.

And if the first judgment result is yes or the second judgment result is yes, namely the attenuation coefficient of the intelligent model is smaller than the attenuation threshold value in the first effect condition or the iteration coefficient of the intelligent model is larger than the iteration threshold value in the second effect condition, the effect of the intelligent model meets the preset offline requirement.

And S190, if the effect of the intelligent model meets the offline requirement, completing the offline of the intelligent model.

In one embodiment, as shown in FIG. 8, step S190 includes substeps S191-S193.

S191, the pre-stored system configuration initiates a destroy inference instance request to a bottom layer inference engine model pre-stored by the system. When the effect of the intelligent model meets the offline requirement, a bottom layer inference engine model prestored in the system is triggered, the bottom layer inference engine model is developed based on a cloud native mode and is exposed based on an RESTAPI mode, wherein RESTAPI is a set of architecture rules, standards or guidance on how to construct a network application program interface, RESTAPI follows the architecture style of an application program interface principle, the complexity of system development is reduced, and the scalability of the system is improved.

And S192, the bottom layer inference engine model requests through the destruction inference example. Here, the underlying inference engine model is adapted to the mirror to mark whether the model is an LR, XGB or TF logical algorithm; and dynamically selecting the model file from the S3 storage and registering the model file for service. S3 is an automatic supervision learning model, the adoption of the S3 automatic supervision learning model facilitates the fusion of context characteristics, and the relevance of the object and the context characteristics can be captured from other angles when other training targets (such as mutual information maximization) are adopted; the association exposes K8S Ingress, wherein Ingress is a concept of K8S and can be called when other business systems perform model inference service.

And S193, the system calls a destructor corresponding to the intelligent model to destroy the intelligent model, the intelligent model is offline, and a model offline report and a flow monitoring report are output. Because each intelligent model created by the system occupies certain system resources, the system calls the intelligent model with the destructor effect attenuated corresponding to the intelligent model, releases certain system resources, improves the real-time performance of model data updating, and ensures the operating efficiency of the system.

The embodiment of the present invention further provides a management apparatus for a full-lifecycle intelligent model, where the management apparatus for a full-lifecycle intelligent model can be configured in a management server, and the management apparatus for a full-lifecycle intelligent model is configured to execute any one of the embodiments of the management method for a full-lifecycle intelligent model described above. Specifically, referring to fig. 9, fig. 9 is a schematic block diagram of a management apparatus for a full-lifecycle intelligent model according to an embodiment of the present invention.

As shown in fig. 9, the management apparatus 200 for a full-life-cycle intelligent model includes a configuration unit 210 for dividing a system hierarchy according to the pre-stored system configuration; configuring a preset approval chain of the model series; a receiving and determining unit 220, configured to receive data to be processed input by the user end 20, and determine whether the data to be processed is included in a pre-stored historical database; a flow establishing unit 230, configured to establish a required flow corresponding to the to-be-processed data if the to-be-processed data is not included in a pre-stored historical database; a file obtaining unit 240, configured to obtain a corresponding model file according to the demand flow; the approval unit 250 is configured to, if the application form meets the approval requirement of the approval chain, obtain a yes judgment result output by the model, and approve the approval chain by the user according to the judgment result; the model offline unit 260 is used for initiating a destroy inference example request to a bottom layer inference engine model prestored in the system by the prestored system configuration; the bottom layer reasoning engine model requests through the destruction reasoning instance; and the system calls a destructor corresponding to the model to destroy the model.

An embodiment of the present invention further provides a computer device 500, where the computer device 500 includes a memory, a processor 502 and a computer program 5032 stored in the memory and capable of running on the processor 502, and the management apparatus of the full-lifecycle intelligent model may be implemented in the form of the computer program 5032, and the computer program 5032 may run on the computer device 500 shown in fig. X.

Referring to fig. 10, fig. 10 is a schematic block diagram of a computer device 500 according to an embodiment of the present invention. The computer device 500 may be a management server 10 for a management method of a full life cycle intelligent model to manage the intelligent model.

Referring to fig. 10, the computer device 500 includes a processor 502, memory, and a network interface 505 connected by a system bus 501, where the memory may include a storage medium 503 and an internal memory 504.

The storage medium 503 may store an operating system 5031 and computer programs 5032. The computer program 5032, when executed, may cause the processor 502 to perform a method of managing a full lifecycle intelligence model, where the storage medium 503 may be a volatile storage medium 503 or a non-volatile storage medium 503.

The processor 502 is used to provide computing and control capabilities that support the operation of the overall computer device 500.

The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503, and when the computer program 5032 is executed by the processor 502, the processor 502 can be enabled to execute the management method of the full-life intelligent model.

The network interface 505 is used for network communication, such as providing transmission of data information. Those skilled in the art will appreciate that the configuration shown in fig. 10 is a block diagram of only a portion of the configuration associated with aspects of the present invention and is not intended to limit the computing device 500 to which aspects of the present invention may be applied, and that a particular computing device 500 may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

The processor 502 is configured to run the computer program 5032 stored in the memory to implement the corresponding functions in the management method of the full-lifecycle intelligent model described above.

Those skilled in the art will appreciate that the embodiment of computer device 500 illustrated in FIG. 10 is not intended to limit the specific configuration of computer device 500, and that in other embodiments, computer device 500 may include more or less components than those illustrated, or some components may be combined, or a different arrangement of components. For example, in some embodiments, the computer device 500 may only include the memory and the processor 502, and in such embodiments, the structure and function of the memory and the processor 502 are the same as those of the embodiment shown in fig. 10, and are not described herein again.

It should be understood that, in the embodiment of the present invention, the Processor 502 may be a Central Processing Unit (CPU), and the Processor 502 may also be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field-Programmable gate arrays (FPGAs) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

In another embodiment of the invention, a computer-readable storage medium 503 is provided. The computer-readable storage medium 503 may be a volatile or non-volatile computer-readable storage medium 503. The computer readable storage medium 503 stores a computer program 5032, wherein the computer program 5032 when executed by the processor 502 implements the steps included in the method for managing a full-life intelligent model as described above.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses, devices and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again. Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

In the embodiments provided by the present invention, it should be understood that the disclosed apparatus, device and method can be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only a logical division, and there may be other divisions in actual implementation, or units with the same function may be grouped into one unit, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may also be an electric, mechanical or other form of connection.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment of the present invention.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium 503. Based on such understanding, the technical solution of the present invention essentially or partly contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a computer-readable storage medium 503 and includes several instructions for causing a computer device 500 (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned computer-readable storage medium 503 includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.

The implementation principles of the method, the device, the equipment and the medium for managing the learning model of the full life cycle in the embodiment of the invention are as follows: the management method comprises initializing a pre-stored system configuration; carrying out approval chain configuration on the system according to the pre-stored system configuration; if the data to be processed input by the user is received, judging whether the data to be processed is contained in a pre-stored historical database; if the data to be processed is not contained in the pre-stored historical database, a demand flow corresponding to the data to be processed is newly established; acquiring a corresponding model file according to a demand flow; if the intelligent model corresponding to the model file receives the approval information of the user to the approval chain, outputting a related report and sending the related report; judging whether the effect of the intelligent model meets the preset offline requirement or not; and if the effect of the intelligent model meets the offline requirement, completing the offline of the intelligent model. The management method of the full-life-cycle intelligent model can be applied to model management of a user group consisting of a plurality of people, such as enterprises, schools and the like, so that the model management is standardized, and the real-time performance of model data updating and the safety of model use are improved.

While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A management method of a full-life-cycle intelligent model is applied to a management server, and a user side and the management server are connected through a network to transmit data information, and is characterized by comprising the following steps:

initializing a pre-stored system configuration;

acquiring a corresponding model file according to the demand flow;

judging whether the effect of the intelligent model meets a preset offline requirement or not;

2. The method for managing full-life intelligent models according to claim 1, wherein the configuring the system according to the pre-stored system configuration for approval chain comprises:

3. The method as claimed in claim 1, wherein the determining whether the data to be processed is contained in a pre-stored historical database comprises:

and searching the retrieval identification in the historical database to judge whether the historical database contains data matched with the retrieval identification.

4. The method for managing a full-life-cycle intelligent model according to claim 1, wherein the obtaining of the corresponding model file according to the demand flow includes:

constructing an intelligent model according to the demand flow;

and the intelligent model is trained so as to output a model file.

5. The method for managing full-life intelligent model according to claim 1, wherein the determining whether the determination result of the model on the application form output is yes comprises:

the intelligent model judges whether the application form meets the approval requirements of the approval chain;

if the application table meets the approval requirement of the approval chain, the judgment result output by the intelligent model is yes, and the approval chain receives confirmation information input by a user end and finishes approval;

and if the application form does not meet the approval requirements of the approval chain, the judgment result output by the intelligent model is negative, and the approval chain receives the judgment information that the judgment result is negative and returns the application form to the user side.

6. The method for managing the full-life-cycle intelligent model according to claim 1, wherein the step of judging whether the effect of the intelligent model meets the preset offline requirement comprises the following steps:

judging whether the attenuation effect of the intelligent model meets a first effect condition of the offline requirement or not to obtain a first judgment result;

judging whether the iteration effect of the intelligent model meets a second effect condition of the offline requirement or not to obtain a second judgment result;

7. The method of claim 1, wherein said completing the intelligent model offline comprises:

the pre-stored system configuration initiates a destroy inference example request to a bottom layer inference engine model pre-stored by the system;

8. An apparatus for managing a full-life intelligent model, wherein the apparatus is configured in a management server, and the management server establishes a network connection with a user terminal to implement data information transmission, the apparatus comprising:

the flow newly-establishing unit is used for newly establishing a demand flow corresponding to the data to be processed if the data to be processed is not contained in a pre-stored historical database;

the model offline unit is used for initiating a destroy inference example request to a bottom layer inference engine model prestored in the system by the prestored system configuration; the bottom layer reasoning engine model requests through the destruction reasoning instance; and the system calls a destructor corresponding to the model to destroy the model.

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements a method of managing a full lifecycle intelligence model as claimed in any of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that it stores a computer program which, when executed by a processor, implements the method of managing a full-life intelligent model according to any one of claims 1 to 7.