CN116070716A - Model verification realization method and device and server - Google Patents
Model verification realization method and device and server Download PDFInfo
- Publication number
- CN116070716A CN116070716A CN202310125263.4A CN202310125263A CN116070716A CN 116070716 A CN116070716 A CN 116070716A CN 202310125263 A CN202310125263 A CN 202310125263A CN 116070716 A CN116070716 A CN 116070716A
- Authority
- CN
- China
- Prior art keywords
- model
- preset
- machine learning
- verification
- package
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012795 verification Methods 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013507 mapping Methods 0.000 claims abstract description 74
- 238000010801 machine learning Methods 0.000 claims abstract description 68
- 238000004590 computer program Methods 0.000 claims description 15
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 4
- 230000000007 visual effect Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 238000012800 visualization Methods 0.000 description 4
- 238000012549 training Methods 0.000 description 3
- INFDPOAKFNIJBF-UHFFFAOYSA-N paraquat Chemical compound C1=C[N+](C)=CC=C1C1=CC=[N+](C)C=C1 INFDPOAKFNIJBF-UHFFFAOYSA-N 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/10—Requirements analysis; Specification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/40—Transformation of program code
- G06F8/41—Compilation
- G06F8/42—Syntactic analysis
- G06F8/427—Parsing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Stored Programmes (AREA)
Abstract
The present disclosure provides a method, an apparatus and a server for implementing model verification, which relate to an artificial intelligence technology, and include: acquiring an application calling request sent by a terminal, wherein the application calling request comprises a model package of a first model to be checked, a first machine learning platform and a first frame corresponding to the first model; analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model; determining target model specifications corresponding to a first machine learning platform and a first framework according to a preset specification mapping table; verifying a model structure corresponding to the first model by using a target model specification to obtain a verification result of the first model; and feeding back the verification result to the terminal. According to the scheme, the model and the corresponding model specifications can be analyzed, and the model and the corresponding model specifications are converted into a simpler mode compared with the prior art and then are checked, so that the model is easier to understand and maintain.
Description
Technical Field
The disclosure relates to artificial intelligence technology, and in particular relates to a method, a device and a server for realizing model verification.
Background
With the advent of artificial intelligence, machine learning has rolled up the entire technology industry in recent years with the potential of starfire burning. Machine learning platforms are dedicated to providing one-stop services to machine learning practitioners, defining model specifications suitable for the platform itself to implement a machine learning process. These machine learning platforms are popular with many businesses and open source communities, where update iterations are frequent. A problem that follows is that model specifications defined by respective machine learning platforms often have large differences, and even model specifications of different versions of the same platform are incompatible. When the development, application and management of the model are carried out in an enterprise, unified verification is required to be carried out on the development personnel of different technical stacks and the models produced by different machine learning platforms, so that the model files, the model data and the model engineering are ensured to accord with corresponding specifications.
In the prior art, in order to support model specifications of different versions of each machine learning platform, an abstraction is often made at a code level, and different model specifications are changed and expanded in a mode of adding a new code. In a specific implementation, the model structure is verified by hard coding and writing fixed rules. The relatively elegant method is to store the rules in a database and provide a visual interface for rule editing.
However, due to a great deal of complicated work in engineering of machine learning, the model structure is complex, and accordingly, the verification method is complex and is not easy to understand and maintain.
Disclosure of Invention
The disclosure provides a method, a device and a server for realizing model verification, which are used for solving the problems that a model verification method in the prior art is complex and is not easy to understand and maintain.
According to a first aspect of the present disclosure, there is provided a method for implementing model verification, including:
acquiring an application calling request sent by a terminal, wherein the application calling request comprises a model package of a first model to be checked, and a first machine learning platform and a first frame corresponding to the first model;
analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model;
determining target model specifications corresponding to the first machine learning platform and the first framework according to a preset specification mapping table; the preset specification mapping table comprises a machine learning platform, a framework and a mapping relation of model specifications; the model specification is determined by analyzing an original model specification;
verifying a model structure corresponding to the first model by using the target model specification, and obtaining a verification result of the first model; and feeding back the verification result to the terminal.
According to a second aspect of the present disclosure, there is provided an implementation apparatus for model verification, including:
the system comprises an acquisition unit, a verification unit and a verification unit, wherein the acquisition unit is used for acquiring an application calling request sent by a terminal, and the application calling request comprises a model package of a first model to be verified, a first machine learning platform and a first frame corresponding to the first model;
the model analysis unit is used for analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model;
a specification mapping unit, configured to determine, according to a preset specification mapping table, a target model specification corresponding to the first machine learning platform and the first framework; the preset specification mapping table comprises a machine learning platform, a framework and a mapping relation of model specifications; the model specification is determined by analyzing an original model specification;
the model verification unit is used for verifying the model structure corresponding to the first model by utilizing the target model specification and obtaining a verification result of the first model; and feeding back the verification result to the terminal.
According to a third aspect of the present disclosure, there is provided a server comprising a memory and a processor; wherein,,
the memory is used for storing a computer program;
the processor is configured to read the computer program stored in the memory, and execute the implementation method of model verification according to the first aspect according to the computer program in the memory.
According to a fourth aspect of the present disclosure, there is provided a computer-readable storage medium having stored therein computer-executable instructions which, when executed by a processor, implement a method of implementing model verification as described in the first aspect.
According to a fifth aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements a method of implementing model verification as described in the first aspect.
The method, the device and the server for realizing model verification provided by the disclosure comprise the following steps: acquiring an application calling request sent by a terminal, wherein the application calling request comprises a model package of a first model to be checked, a first machine learning platform and a first frame corresponding to the first model; analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model; determining target model specifications corresponding to a first machine learning platform and a first framework according to a preset specification mapping table; the preset standard mapping table comprises a machine learning platform, a frame and a mapping relation of model standards; the model specification is determined by analyzing the original model specification; verifying a model structure corresponding to the first model by using a target model specification to obtain a verification result of the first model; and feeding back the verification result to the terminal. In the method, the device and the server for realizing model verification, the model and the corresponding model specifications can be analyzed, and the model and the corresponding model specifications are converted into a simpler mode compared with the prior art and then are verified, so that the method, the device and the server are easier to understand and maintain. And a preset standard mapping table is adopted, so that the universality of model standards is improved.
Drawings
In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present disclosure, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for implementing model verification according to an exemplary embodiment of the present disclosure;
FIG. 2 is a flow chart of a method of implementing model verification as shown in another exemplary embodiment of the present disclosure;
FIG. 3 is a schematic diagram of an implementation process of model verification shown in an exemplary embodiment of the present disclosure;
FIG. 4 is a block diagram of an implementation of model verification shown in an exemplary embodiment of the present disclosure;
fig. 5 is a block diagram of a server shown in an exemplary embodiment of the present disclosure.
Detailed Description
With the advent of artificial intelligence, machine learning has rolled up the entire technology industry in recent years with the potential of starfire burning. Whereas in a machine-learned workflow, the code for model training is only a small part of it. Besides, the monitoring of training tasks, the recovery of logs, the selection and optimization of super parameters, the release and integration of models, the data cleaning, the feature extraction and the like are all indispensable parts in the process. Because of the great amount of complicated work in engineering of machine learning, there are many company products and open source products, helping users to better complete the landing of their machine learning business. Currently, there are Kubeflow, MLflow, argo, airFlow, seldon, comet, floydHub, riseML, sageMaker machine learning platforms that are being adopted by various major technologies. Machine learning platforms are dedicated to providing one-stop services to machine learning practitioners, defining model specifications suitable for the platform itself to implement a machine learning process. These machine learning platforms are popular with many businesses and open source communities, where update iterations are frequent. A problem that follows is that model specifications defined by respective machine learning platforms often have large differences, and even model specifications of different versions of the same platform are incompatible. When the development, application and management of the model are carried out in an enterprise, unified verification is required to be carried out on the development personnel of different technical stacks and the models produced by different machine learning platforms, so that the model files, the model data and the model engineering are ensured to accord with corresponding specifications.
In the prior art, in order to support model specifications of different versions of each machine learning platform, an abstraction is often made at a code level, and different model specifications are changed and expanded in a mode of adding a new code. In a specific implementation, the model structure is verified by hard coding and writing fixed rules. A relatively elegant approach is to save rules to a database, provide a visual interface for rule editing, but still fix the rules.
However, due to a great deal of complicated work in engineering of machine learning, the model structure is complex, and accordingly, the verification method is complex and is not easy to understand and maintain. In addition, when facing models produced by research personnel of different machine learning platforms and different technical stacks, the prior art cannot be flexibly expanded, and a new code is often required to expand a verification module.
In order to solve the technical problems, in the scheme provided by the disclosure, the model and the corresponding model specification can be analyzed, and the model and the corresponding model specification are converted into a simpler mode compared with the prior art and then are checked, so that the model and the model specification are easier to understand and maintain. And a preset standard mapping table is adopted, so that the universality of model standards is improved.
It should be noted that, the user information (including, but not limited to, user equipment information, user personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) related to the present disclosure are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region, and be provided with corresponding operation entries for the user to select authorization or rejection.
The following describes the technical solutions of the present disclosure and how the technical solutions of the present disclosure solve the above technical problems in detail with specific embodiments. The following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. Embodiments of the present disclosure will be described below with reference to the accompanying drawings.
Fig. 1 is a flow chart illustrating a method for implementing model verification according to an exemplary embodiment of the present disclosure.
As shown in fig. 1, the method for implementing model verification provided in this embodiment includes:
The method provided by the present disclosure may be performed by a server, among other things.
An application program interface (Application Programming Interface, API) for realizing model verification can be preset in the server for the terminal to call, and a software development kit (Software Development Kit, SDK) corresponding to the application interface for realizing model verification can be deployed in the terminal. Specifically, the server may obtain an application call request sent by the terminal, so as to call the application for implementing the model verification.
Specifically, the application call request may further include a model package of the first model to be verified, and a first machine learning platform and a first framework corresponding to the first model.
Specifically, according to the specified model specification, a model file, model data, model engineering, configuration file and the like corresponding to the first model can be packaged into a model package of the first model.
The model refers to a method or system for converting input data into numerical values of numerical value type prediction conclusion by using statistical, economic, financial or mathematical theory and technology, and also includes a method or system for outputting numerical values according to partial or complete qualitative input or based on expert judgment. A model consists of three parts: one is an information input section which inputs data into the model; one is a processing section that converts the input into a predicted conclusion; one is a reporting section that translates the predicted conclusions into business information that can be used.
The model file is used for serializing the trained model in the memory and persisting the model to the hard disk in a file form.
Where model data refers to training data sets, test data sets and other accessories typically required to contain the model.
The model engineering refers to engineering files and configuration files required by subsequent operations such as deployment, retraining and the like of the model.
The model checking refers to checking whether the model accords with the model specification of the platform or the running environment before the model is used. That is, it is checked whether the first model meets the model specification corresponding to the first machine learning platform and the first framework.
And 102, analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model.
Specifically, a preset mode may be adopted to analyze a model package of the first model, convert the content in the model package into a simpler data format, and determine the converted data format as a model structure corresponding to the first model.
The preset specification mapping table is preset according to actual conditions. The preset standard mapping table comprises a machine learning platform, a framework and a mapping relation of model standards.
The model specification is determined by analyzing the original model specification in a mode corresponding to the analysis model package.
Specifically, a preset specification mapping table can be matched according to the first machine learning platform and the first framework to obtain a model specification corresponding to the first machine learning platform and the first framework, and the model specification is determined to be a target model specification of the first model.
104, verifying a model structure corresponding to the first model by using the target model specification, and obtaining a verification result of the first model; and feeding back the verification result to the terminal.
Specifically, the model structure corresponding to the first model can be verified by using the target model specification, and a verification result of the first model is obtained. The verification result lists the contents of the model structure which do not accord with the target model specification.
The verification result may then be fed back to the terminal to enable the terminal to adjust the first model with the verification result.
The implementation method for model verification provided by the disclosure comprises the following steps: acquiring an application calling request sent by a terminal, wherein the application calling request comprises a model package of a first model to be checked, a first machine learning platform and a first frame corresponding to the first model; analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model; determining target model specifications corresponding to a first machine learning platform and a first framework according to a preset specification mapping table; the preset standard mapping table comprises a machine learning platform, a frame and a mapping relation of model standards; the model specification is determined by analyzing the original model specification; verifying a model structure corresponding to the first model by using a target model specification to obtain a verification result of the first model; and feeding back the verification result to the terminal. In the method adopted by the disclosure, the model and the corresponding model specification can be analyzed, and the model and the corresponding model specification are converted into a simpler mode compared with the prior art and then are checked, so that the method is easier to understand and maintain. And a preset standard mapping table is adopted, so that the universality of model standards is improved.
Fig. 2 is a flow chart illustrating a method for implementing model verification according to another exemplary embodiment of the present disclosure.
As shown in fig. 2, the method for implementing model verification provided in this embodiment includes:
Specifically, the principle and implementation of step 201 are similar to those of step 101, and will not be described again.
Specifically, the directory structure, the file format, the file content, and the attribute value included in the model package of the first model may be converted into a data exchange format (JavaScript Object Notation, JSON) string, and the JSON string may be determined as a model structure corresponding to the first model.
Among them, JSON is a lightweight data exchange format, which uses text formats that are completely independent of language.
Specifically, a preset specification mapping table can be matched according to the first machine learning platform and the first framework to obtain a model specification corresponding to the first machine learning platform and the first framework, and the model specification is determined to be a target model specification of the first model.
Specifically, the model specification is determined by parsing the original model specification in a manner corresponding to the parsing model package.
For example, the model specification may be JSON Schema (JSON Schema). JSON Schema is used to describe the JSON data format, defining one criterion for JSON data constraints. According to the convention mode, two parties exchanging data can understand the requirements and constraints of JSON data, and can verify the data according to the requirements, so that the correctness of data exchange is ensured.
In one implementation, a preset specification map is visually displayed.
In one implementation, a preset specification mapping table is edited in response to a mapping editing instruction of a user.
Specifically, the preset specification map may be visually displayed. Furthermore, the user can edit the preset specification mapping table in a visual mode.
In one implementation, model specifications are managed using a preset version management table.
Specifically, the model specification may be managed using a version management table set in advance. And can build a connection between the preset version management table and a preset specification mapping table. The preset specification mapping table can extract the corresponding model specification from the preset version management table according to the name of the model specification.
In one implementation, the preset version management table is visually displayed.
In one implementation, the preset version management table is edited in response to a user's specification edit instruction.
Specifically, the preset version management table may be visually displayed. Further, the user can edit the preset version management table in a visual manner.
Specifically, the version management and visual editing can be provided for JSON Schema corresponding to the original model specifications of different versions of each machine learning platform.
Specifically, when the model package of the first model is analyzed, configuration information in the model package of the first model can be identified at the same time. The configuration information may be a configuration item in a configuration file included in the model package. For example, the configuration file my_file in the model package has a configuration item of env_file: abc.env, at this time, the my_file can be parsed, the env_file configuration item is identified, and a rule for checking the abc.env file is automatically newly added in the currently loaded JSON-Schema, without defining a fixed rule for checking the abc.env file in advance, thereby realizing a dynamic checking model. Note that abc.env is a user-defined file, and the model specification does not know the existence of the document in advance, and the document may be preset according to requirements, for example, abc.env, def.prop, and the like.
Specifically, automatic specification expansion can be performed according to the configuration information, and the target model specification is updated by using the expanded model specification.
In one implementation, according to the configuration information and a preset specification extension mapping table, an extension specification corresponding to the configuration information is determined; the preset specification expansion mapping table comprises mapping relation between configuration information and expansion specification.
The configuration information may include information such as an attribute, a parameter, and a configuration of the operating environment of the first model.
The preset specification expansion mapping table is a specification expansion mapping table preset according to actual conditions. The preset specification expansion mapping table comprises mapping relation between configuration information and expansion specification.
The extended specification may be determined by parsing the original model specification in a manner corresponding to the parsing model package. Specifically, the extension specification may be JSON mode.
Specifically, according to the identified configuration information of the first model, a preset specification expansion mapping table can be queried to obtain an expansion specification corresponding to the configuration information.
Then, the extended specification is added to the target model specification, generating an updated target model specification.
Specifically, the extended specification may be added to the target model specification, thereby generating an updated target model specification.
Further, the extended specification may be managed using a preset specification table. The preset specification list can be visualized, so that a user can edit the preset specification list in a visualized manner.
Further, a preset specification extension mapping table may be visualized. And the user can edit the preset specification expansion mapping table in a visual mode.
Specifically, by adopting a preset specification extension mapping table, the extension specification can be automatically increased according to the identified configuration information, so that the target model specification is updated. The method can automatically adapt to the attribute, parameter and running environment configuration of different models, and can flexibly perform specification expansion.
For example, the A model is a Python model, there may be Python version requirements in the runtime environment configuration, the B model is a Java model, and there may be Java version requirements in the runtime environment. For another example, a classification model and a clustering model, the super parameters of the two models are different, and automatic adjustment is needed according to the configuration file.
Specifically, the model structure corresponding to the first model can be verified by using the updated target model specification, specifically, JSON Schema is used to verify whether the JSON string expressing the machine learning model structure meets the constraint of the JSON Schema defining the model specification. And obtaining a verification result of the first model. The verification result lists the contents of the model structure which do not accord with the target model specification.
The verification result may then be fed back to the terminal to enable the terminal to adjust the first model with the verification result.
Specifically, the scheme uses the JSON Schema to define the model specification, so that the complex model specification is converted into the structured data, and the user can understand and maintain conveniently. A JSON Schema checker is used to check whether the model structure (JSON string) complies with the model specification (JSON Schema). Meanwhile, in the verification process, the attribute value of the model configuration file is analyzed, and the rule is automatically added by utilizing the characteristics of the JSON Schema, so that the model specification of the verification is dynamically adjusted. The complex model verification is converted into simple, flexible and dynamic JSON string verification.
And unified management and convenient maintenance of model specifications of different versions of different machine learning platforms are realized through visualization, version management, specification mapping and other modes.
FIG. 3 is a schematic diagram of an implementation process of model verification according to an exemplary embodiment of the present disclosure.
As shown in fig. 3, 1 indicates that a user can enter/edit model specifications (JSON Schema) at a visual interface. 2 represents the mapping relation between the user and the machine learning platform and the framework, wherein the mapping relation between the user and the machine learning platform can be input/edited in a visual interface. 3 indicates that the acquired model can be input to a model parser through an API/SDK (i.e., a preset application interface), and the model parser parses the acquired model to obtain a model structure (JSON string). 4 indicates that the API/SDK may pass the machine learning platform and framework information corresponding to the model verifier. 5 indicates that the model checker can obtain the model structure (JSON string) from the model parser. 6 shows that the model verifier can obtain model specifications (JSON Schema) from the version management and specification mapping module according to the machine learning platform and framework information. Then, the model calibrator can utilize the model specification to calibrate the model to obtain a calibration result. 7 indicates that the model verifier can communicate the verification result to the user/caller.
In fig. 3, version management may be used to implement real-time editing and version management of JSON Schema defined model specifications. The standard mapping can be used for establishing model standard and mapping relation between each machine learning platform and frame, and supporting dynamic adjustment of users. The visualization can be used for providing visual support for version management and specification images, and is convenient for users to operate and manage. The model parser may be used to read a machine learning model, mapping the directory structure, file format, file content and attribute values of the model into JSON strings. The model checker can read corresponding model specifications according to a machine learning platform to which the model belongs and a used framework, check a JSON character string representing a model structure by utilizing a JSON Schema, and simultaneously support the analysis of attribute values of a model configuration file in the checking process, automatically add rules by utilizing the characteristics of the JSON Schema and dynamically adjust the JSON Schema of the current checking. The API/SDK may be used to provide the model checking capability of the present invention to the application layer and external systems.
Fig. 4 is a block diagram of an implementation apparatus of model checking shown in an exemplary embodiment of the present disclosure.
As shown in fig. 4, an implementation apparatus 400 for model verification provided by the present disclosure includes:
an obtaining unit 410, configured to obtain an application call request sent by a terminal, where the application call request includes a model package of a first model to be verified, and a first machine learning platform and a first frame corresponding to the first model;
the model parsing unit 420 is configured to parse the model package of the first model in a preset manner to obtain a model structure corresponding to the first model;
a specification mapping unit 430, configured to determine, according to a preset specification mapping table, a target model specification corresponding to the first machine learning platform and the first framework; the preset standard mapping table comprises a machine learning platform, a frame and a mapping relation of model standards; the model specification is determined by analyzing the original model specification;
the model checking unit 440 is configured to check a model structure corresponding to the first model by using the target model specification, and obtain a check result of the first model; and feeding back the verification result to the terminal.
The model checking unit 440 is specifically configured to identify configuration information in a model package of the first model, and update the target model specification according to the configuration information;
and verifying the model structure corresponding to the first model by using the updated target model specification.
The model checking unit 440 is specifically configured to determine an extension specification corresponding to the configuration information according to the configuration information and a preset specification extension mapping table; the preset specification expansion mapping table comprises mapping relation between configuration information and expansion specification;
and adding the extension specification to the target model specification to generate an updated target model specification.
The model parsing unit 420 is specifically configured to convert a directory structure, a file format, file contents, and attribute values included in a model package of the first model into a data exchange format string, and determine the data exchange format string as a model structure corresponding to the first model.
The implementation apparatus 400 for model verification provided in the present disclosure further includes:
a visualization unit 450, configured to visually display a preset specification mapping table;
and responding to a mapping editing instruction of a user, and editing a preset standard mapping table.
The version management unit 460 is configured to manage the model specification using a preset version management table.
The visualization unit 450 is further configured to visually display a preset version management table;
and responding to a specification editing instruction of a user, and editing a preset version management table.
Fig. 5 is a block diagram of a server shown in an exemplary embodiment of the present disclosure.
As shown in fig. 5, the server provided in this embodiment includes:
a memory 501;
a processor 502; and
a computer program;
wherein a computer program is stored in the memory 501 and configured to be executed by the processor 502 to implement a method of implementing any of the model checks described above.
The present embodiment also provides a computer-readable storage medium having stored thereon a computer program that is executed by a processor to implement a method of implementing any of the model checks described above.
The present embodiment also provides a computer program product, including a computer program, which when executed by a processor, implements a method for implementing any of the above model checking.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the method embodiments described above may be performed by hardware associated with program instructions. The foregoing program may be stored in a computer readable storage medium. The program, when executed, performs steps including the method embodiments described above; and the aforementioned storage medium includes: various media that can store program code, such as ROM, RAM, magnetic or optical disks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (10)
1. The implementation method of model verification is characterized by comprising the following steps:
acquiring an application calling request sent by a terminal, wherein the application calling request comprises a model package of a first model to be checked, and a first machine learning platform and a first frame corresponding to the first model;
analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model;
determining target model specifications corresponding to the first machine learning platform and the first framework according to a preset specification mapping table; the preset specification mapping table comprises a machine learning platform, a framework and a mapping relation of model specifications; the model specification is determined by analyzing an original model specification;
verifying a model structure corresponding to the first model by using the target model specification, and obtaining a verification result of the first model; and feeding back the verification result to the terminal.
2. The method of claim 1, wherein verifying the model structure corresponding to the first model using the target model specification comprises:
identifying configuration information in a model package of the first model, and updating the target model specification according to the configuration information;
and checking a model structure corresponding to the first model by using the updated target model specification.
3. The method of claim 2, wherein said updating said object model specification based on said configuration information comprises:
determining an expansion specification corresponding to the configuration information according to the configuration information and a preset specification expansion mapping table; the preset specification expansion mapping table comprises mapping relation between configuration information and expansion specification;
and adding the extension specification to a target model specification to generate an updated target model specification.
4. The method of claim 1, wherein the analyzing the model package of the first model in a preset manner to obtain the model structure corresponding to the first model includes:
and converting the directory structure, the file format, the file content and the attribute value contained in the model package of the first model into a data exchange format character string, and determining the data exchange format character string as a model structure corresponding to the first model.
5. The method of any one of claims 1-4, further comprising:
visually displaying the preset specification mapping table;
and responding to a mapping editing instruction of a user, and editing the preset standard mapping table.
6. The method of any one of claims 1-4, further comprising:
managing model specifications by using a preset version management table;
visually displaying the preset version management table;
and responding to a specification editing instruction of a user, and editing the preset version management table.
7. A device for implementing model verification, comprising:
the system comprises an acquisition unit, a verification unit and a verification unit, wherein the acquisition unit is used for acquiring an application calling request sent by a terminal, and the application calling request comprises a model package of a first model to be verified, a first machine learning platform and a first frame corresponding to the first model;
the model analysis unit is used for analyzing the model package of the first model in a preset mode to obtain a model structure corresponding to the first model;
a specification mapping unit, configured to determine, according to a preset specification mapping table, a target model specification corresponding to the first machine learning platform and the first framework; the preset specification mapping table comprises a machine learning platform, a framework and a mapping relation of model specifications; the model specification is determined by analyzing an original model specification;
the model verification unit is used for verifying the model structure corresponding to the first model by utilizing the target model specification and obtaining a verification result of the first model; and feeding back the verification result to the terminal.
8. A server comprising a memory and a processor; wherein,,
the memory is used for storing a computer program;
the processor being configured to read a computer program stored in the memory and to perform the method according to any of the preceding claims 1-6 according to the computer program in the memory.
9. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor implement the method of any of the preceding claims 1-6.
10. A computer program product comprising a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310125263.4A CN116070716A (en) | 2023-02-07 | 2023-02-07 | Model verification realization method and device and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310125263.4A CN116070716A (en) | 2023-02-07 | 2023-02-07 | Model verification realization method and device and server |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116070716A true CN116070716A (en) | 2023-05-05 |
Family
ID=86183575
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310125263.4A Pending CN116070716A (en) | 2023-02-07 | 2023-02-07 | Model verification realization method and device and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116070716A (en) |
-
2023
- 2023-02-07 CN CN202310125263.4A patent/CN116070716A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11163906B2 (en) | Adaptive redaction and data releasability systems using dynamic parameters and user defined rule sets | |
| US10782961B2 (en) | Analyzing components related to a software application in a software development environment | |
| US20180210883A1 (en) | System for converting natural language questions into sql-semantic queries based on a dimensional model | |
| US9256582B2 (en) | Conversion of a presentation to Darwin Information Typing Architecture (DITA) | |
| US20100121883A1 (en) | Reporting language filtering and mapping to dimensional concepts | |
| US20210012103A1 (en) | Systems and methods for information extraction from text documents with spatial context | |
| CN111722839A (en) | Code generation method and device, electronic equipment and storage medium | |
| CN111581920A (en) | Document conversion method, device, equipment and computer storage medium | |
| CN108762743A (en) | Data table operation code generation method and device | |
| CN111784108B (en) | Modeling method and device of main data management platform | |
| US11763095B2 (en) | Creating apps from natural language descriptions | |
| CN110474836A (en) | A kind of email template generation method, calculates equipment and medium at device | |
| US20210124752A1 (en) | System for Data Collection, Aggregation, Storage, Verification and Analytics with User Interface | |
| CN115062117A (en) | Method for automatically generating and classifying documents based on natural language processing technology | |
| CN112100984A (en) | Data conversion method and system from EBOM to SBOM | |
| WO2023151397A1 (en) | Application program deployment method and apparatus, device, and medium | |
| US20220215034A1 (en) | Electronic apparatus and controlling method thereof | |
| CN116070716A (en) | Model verification realization method and device and server | |
| CN113448852A (en) | Test case obtaining method and device, electronic equipment and storage medium | |
| US9086891B2 (en) | Centralized system to transform a language neutral string content into a format of a desired platform | |
| CN116860227B (en) | Data development system and method based on big data ETL script arrangement | |
| US20220334829A1 (en) | Custom abap cloud enabler | |
| CN115934682A (en) | Database migration method and system | |
| CN117891531A (en) | System parameter configuration method, system, medium and electronic equipment for SAAS software | |
| US20190026289A1 (en) | Storage control program, apparatus, and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |