Method and device for checking normativity of model design, storage medium and terminal equipment

Abstract

The embodiment of the invention discloses a normativity checking method and device for model design, a storage medium and terminal equipment, and relates to the field of software design. The method comprises the steps of obtaining a model design specification document of a research and development project, and generating a specification knowledge base according to the model design specification document; the model design specification document comprises a plurality of model check rules; constructing a training set according to the standard knowledge base; training the large model by adopting pre-training and fine tuning modes based on the training set; the trained large model is utilized, a to-be-checked model in a research and development project is processed, then a check report is output, and the check report comprises problem description information and modification suggestion information through the steps of building a standard knowledge base, generating a training set, training the large model in a pre-training and fine tuning mode, automatically outputting the check report and the like. The defects of efficiency, consistency and flexibility of a traditional model design normative checking mode are effectively overcome.

Description

Technical Field

[0001] The present application relates to the field of software design, and in particular to a method, device, storage medium and terminal device for checking the standardization of model design. Background Art

[0002] In the field of software development, especially in the development of complex application systems, model design, as the core link in the detailed design stage, plays a vital role in ensuring the rationality of the system architecture, data integrity and function realization. Model design not only involves the definition of the system's data structure, but also covers the processing flow of business logic, and is one of the key factors for the success of software development projects.

[0003] Traditionally, the normative checking of model design mainly relies on two methods: manual review and rule engine checking. However, both methods have significant limitations and shortcomings.

[0004] First of all, manual review is a traditional and direct means of review. Its process relies on the professional knowledge and experience of designers or reviewers, and the model and its attribute design are meticulously checked by comparing with the design specification manual. Although manual review can play the advantages of professional judgment in some aspects, its low efficiency and long time consumption cannot be ignored. As the scale of the project expands, the complexity of model design increases dramatically, and the workload of manual review also increases exponentially, resulting in a longer review cycle and difficulty in meeting the needs of rapid iterative development. In addition, manual review is prone to human omissions, and the inspection results between different reviewers may be different, and consistency is difficult to ensure, which increases the potential risk of the system.

[0005] On the other hand, rule engine checking, as an automated checking method, aims to perform fast and efficient normative checking of models through a preset rule engine. However, the implementation and maintenance of rule engine checking also face many challenges. The configuration process of rules is complex and cumbersome, requiring professionals with deep professional knowledge and practical experience to operate and maintain. At the same time, with the continuous updating and improvement of design specifications, the update of rule engines is not flexible enough, and it is difficult to quickly adapt to the requirements of new design specifications, resulting in the accuracy and effectiveness of the inspection results being limited. In addition, for unstructured design specifications, such as design principles, best practices, etc., rule engines are often difficult to effectively identify and judge, and thus cannot fully cover all aspects of model design.

[0006] In summary, the traditional model design normative checking method has obvious deficiencies in efficiency, consistency and flexibility, and it is difficult to meet the requirements of modern software development projects for efficient, accurate and adaptive normative checking. Therefore, it is particularly important to develop a new method and technology that can overcome the above limitations and improve the efficiency and accuracy of model design normative checking. Summary of the invention

[0007] The embodiments of the present application provide a method, device, storage medium and terminal device for checking the standardization of model design, which can solve the problems of low efficiency and low accuracy of model standardization checking in the prior art. The technical solution is as follows:

[0008] In a first aspect, an embodiment of the present application provides a method for checking the standardization of a model design, the method comprising:

[0009] Obtaining a model design specification document of a research and development project, and generating a specification knowledge base according to the model design specification document; the model design specification document includes a plurality of model checking rules;

[0010] Constructing a training set according to the specification knowledge base;

[0011] Based on the training set, a large model is trained by pre-training and fine-tuning;

[0012] The trained large model is used to process the model to be checked in the R&D project and then an inspection report is output, wherein the inspection report includes: problem description information and modification suggestion information.

[0013] In a second aspect, an embodiment of the present application provides a model design normative checking device, the device comprising:

[0014] An acquisition unit, used to acquire a model design specification document of a research and development project, and to generate a specification knowledge base according to the model design specification document; the model design specification document includes a plurality of model checking rules;

[0015] A construction unit, used for constructing a training set according to the specification knowledge base;

[0016] A training unit, used for training the large model based on the training set by using pre-training and fine-tuning methods;

[0017] The inspection unit is used to use the trained large model to process the model to be inspected in the R&D project and then output an inspection report, wherein the inspection report includes: problem description information and modification suggestion information.

[0018] In a third aspect, an embodiment of the present application provides a computer storage medium, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0019] In a fourth aspect, an embodiment of the present application provides a terminal device, which may include: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0020] The beneficial effects brought about by the technical solutions provided by some embodiments of the present application include at least:

[0021] By obtaining the model design specification documents of the R&D project and converting them into a specification knowledge base, the model checking rules are stored in a systematic and structured manner. This not only improves the queryability and reusability of the rules, but also provides a solid foundation for subsequent automated inspections.

[0022] Building a training set based on the specification knowledge base enables the large model to learn the core elements and inspection logic of the model design specification. The training set generation process ensures that the model can be exposed to various possible model design scenarios during the training process, improving the generalization ability of the model.

[0023] Through pre-training, the large model can learn a lot of prior knowledge and improve its understanding of the model design specifications. Subsequently, through fine-tuning for specific R&D projects, the model can further adapt to the specific needs of the project and improve the accuracy and pertinence of the inspection. This training method not only takes advantage of large-scale data, but also ensures the effectiveness of the model in practical applications.

[0024] By using the trained large model to process the models to be inspected in the R&D project, an inspection report can be automatically generated. The problem description information and modification suggestion information in the report directly point to the non-compliant parts of the model design, helping designers to quickly locate the problem and repair it. This not only significantly improves the inspection efficiency, but also reduces the risk of human omissions.

[0025] Through automated inspection, consistency between different inspectors is ensured, avoiding potential risks caused by human differences. At the same time, as design specifications are continuously updated and improved, large models can be adapted to new requirements through simple retraining or fine-tuning, improving the flexibility and adaptability of inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 It is a schematic diagram of the network architecture provided by the embodiment of the present application;

[0028] Figure 2 It is a flowchart of a method for checking the standardization of a model design provided in an embodiment of the present application;

[0029] Figure 3 is a schematic diagram of a control point management interface provided in an embodiment of the present application;

[0030] Figure 4 is a schematic diagram of a model management interface provided in an embodiment of the present application;

[0031] Figure 5 It is a structural schematic diagram of a model-designed normative inspection device provided by the present application;

[0032] Figure 6 It is a structural schematic diagram of a terminal device provided by this application. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0034] It should be noted that the model design standardization checking method provided in the present application is generally executed by a terminal device, and correspondingly, the model design standardization checking device is generally arranged in the terminal device.

[0035] Figure 1 An exemplary system architecture of a model design normative checking method or a model design normative checking device that can be applied to the present application is shown.

[0036] like Figure 1 As shown, the system architecture may include: a terminal device 101 and a server 102. The terminal device 101 and the server 102 may communicate via a network, and the network is used as a medium for providing communication links between the above-mentioned units. The network may include various types of wired communication links or wireless communication links, for example: a wired communication link includes an optical fiber, a twisted pair or a coaxial cable, etc., and a wireless communication link includes a Bluetooth communication link, a wireless fidelity (WIreless-FIdelity, Wi-Fi) communication link or a microwave communication link, etc.

[0037] The server 102 stores the model design specification document of the R&D project, and the terminal device reads the document from the server 102, and then uses the document to train a large model for model design standardization inspection.

[0038] It should be noted that the terminal device 101 and the server 102 can be hardware or software. When the terminal device 101 and the server 102 are hardware, they can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the terminal device 101 and the server 102 are software, they can be implemented as multiple software or software modules (for example, to provide distributed services), or as a single software or software module, which is not specifically limited here.

[0039] Various communication client applications can be installed on the terminal device of the present application, such as: video recording applications, video playback applications, voice interaction applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0040] The terminal device can be hardware or software. When the terminal device is hardware, it can be various terminal devices with a display screen, including but not limited to smart phones, tablet computers, laptops, and desktop computers. When the terminal device is software, it can be installed in the terminal devices listed above. It can be implemented as multiple software or software modules (for example: used to provide distributed services), or it can be implemented as a single software or software module, which is not specifically limited here.

[0041] When the terminal device is hardware, a display device and a camera may be installed thereon. The display device may be any device capable of realizing a display function, and the camera is used to collect video streams; for example, the display device may be a cathode ray tube display (CR), a light-emitting diode display (LED), an electronic ink screen, a liquid crystal display (LCD), a plasma display panel (PDP), etc. The user may use the display device on the terminal device to view displayed text, pictures, videos and other information.

[0042] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be used as required.

[0043] The following will be combined with the attached Figure 2 , the model design normative inspection method provided in the embodiment of the present application is described in detail. Among them, the model design normative inspection device in the embodiment of the present application can be Figure 1 The terminal device shown.

[0044] See also Figure 2, which is a flow chart of a method for checking the standardization of model design provided in the present application embodiment. Figure 2 As shown, the method of the embodiment of the present application may include the following steps:

[0045] S201. Obtain a model design specification document of a research and development project, and generate a specification knowledge base based on the model design specification document; the model design specification document includes a plurality of model checking rules.

[0046] Among them, the terminal device loads the model design specification document of the R&D project through a file transfer protocol (such as FTP, SFTP), a cloud storage service (such as AWS S3, Alibaba Cloud OSS) or directly from the local file system. This document usually exists in PDF, Word, Markdown or a configuration file in a specific format (such as JSON, YAML), and contains various specifications for model design, such as naming rules, input and output formats, algorithm complexity restrictions, etc. The terminal device has built-in or calls an external parsing library (such as pdfminer for PDF parsing, python-docx for Word parsing, PyYAML for YAML parsing, etc.) to convert the document content into structured data and extract multiple model checking rules. Each rule contains information such as rule description, trigger conditions (such as specific code patterns, parameter ranges), and impact levels (such as warnings, errors). The parsed structured data is stored in the local database or memory structure of the terminal device to form a standard knowledge base. The database or memory structure must support efficient retrieval to quickly match rules during subsequent training and use.

[0047] S202. Construct a training set based on the specification knowledge base.

[0048] The terminal device first parses and understands the content in the specification knowledge base. The specification knowledge base is generated based on the model design specification document and contains multiple model checking rules, each of which defines a pattern that should be followed or avoided in model design.

[0049] Definition of positive samples: A positive sample refers to a model code snippet or a complete model that complies with all model checking rules. Creating positive samples: To generate positive samples, the terminal device can: Extract snippets from existing, verified, and compliant model codes. Use automated tools or scripts to generate compliant model code examples based on the rules in the specification knowledge base. Manually write compliant model code as a supplement. Each positive sample should be accompanied by a label indicating that it complies with the specification (for example, the label is "compliant").

[0050] Definition of negative samples: A negative sample refers to a model code snippet or a complete model that violates at least one model checking rule. Create negative samples: To generate negative samples, the terminal device can: Find model code snippets that are known to violate specifications from historical projects. Deliberately modify certain parts of the positive sample to violate specific model checking rules. Use mutation testing techniques to make slight changes to the positive sample to generate new code variants that violate the specifications. Each negative sample should be accompanied by one or more labels indicating the specific rules it violates (for example, the labels are "naming violation", "too high complexity", etc.).

[0051] For automatically generated or collected code samples, further manual review and annotation may be required to ensure their accuracy and representativeness. The annotation process includes confirming that the positive samples do meet all specifications and determining the specific rules that the negative samples violate. Integrate all positive and negative samples into a training set. The training set should contain enough samples to cover all rules in the specification knowledge base and ensure the generalization ability of the model. The format of the training set should be suitable for the input requirements of the large model, and may include information such as code text, labels, and violation locations. Before the training set is input into the large model, some preprocessing may be required, such as code formatting, removing comments, word segmentation, encoding conversion, etc. The preprocessing step should ensure that the code data in the training set is consistent in format and content to improve the training efficiency and accuracy of the model.

[0052] S203. Based on the training set, the large model is trained by pre-training and fine-tuning.

[0053] Among them, a pre-trained model (such as BERT, GPT series) is selected as the starting point. These models have been trained on a large amount of text data and have strong language understanding and generation capabilities. Use a general programming code data set (such as open source project code on GitHub) to further train the pre-trained model to enhance its understanding of programming language and code structure. Load the pre-trained model to the terminal device (or cloud training platform) and fine-tune it using the training set constructed in step S202. During the fine-tuning process, adjust the model parameters to enable it to better recognize specific rules in the model design specification document. Use cross-validation, learning rate adjustment, early stopping and other strategies to optimize the training process and avoid overfitting.

[0054] S204. Use the trained large model to process the model to be checked in the R&D project and output an inspection report, which includes: problem description information and modification suggestion information.

[0055] Among them, the trained large model is loaded into the memory of the terminal device to prepare for inference. The model code to be checked in the R&D project is received, which can be a single file, a code snippet or the entire project directory. The code is preprocessed, such as removing comments and formatting the code, to meet the model input requirements. The preprocessed code is input into the loaded large model, and the model analyzes the code line by line or block by block to determine whether there is any violation of the specification. After identifying the violation point, the problem description information and modification suggestion information are generated according to the information in the specification knowledge base. All identified violation points and related information are organized into an inspection report, and the report content includes problem description, violation location, modification suggestions, etc. The report can be output in text, HTML, PDF and other formats for easy viewing and understanding by R&D personnel. R&D personnel modify the model code according to the inspection report, and can resubmit the modified code to the terminal device for inspection to verify the modification effect. The terminal device can collect feedback data for continuous optimization and iteration of the model.

[0056] In some embodiments of the present application, it also includes:

[0057] Based on the user's editing instruction, perform operations of adding a model checking rule, modifying a model checking model, or deleting a model checking rule on the model design specification document;

[0058] Retrain large models using the updated model design specification document.

[0059] The terminal device provides an interface that allows the user to enter or select an editing instruction. This can be a text input box (for entering a new rule description), a drop-down menu (for selecting a rule to be modified or deleted), and corresponding buttons (such as "Add", "Modify", "Delete"). After the user submits the editing instruction, the terminal device first parses the instruction to determine the type of operation the user wants to perform (add, modify, delete) and related rule information (such as rule description, rule ID, etc.).

[0060] The terminal device loads the current model design specification document and parses its structure for subsequent modification operations. Add new rules: Add the new rules entered by the user to the appropriate location in the document, ensuring that the rule number, description, trigger conditions and other information are complete and correct. Modify rules: Find the corresponding rule based on the rule ID provided by the user, and then update its description, trigger conditions or impact level and other information. Delete rules: Find and delete the corresponding rule based on the rule ID provided by the user, while ensuring that the rule numbers in the document remain continuous or reordered. Document saving: Save the updated model design specification document to the local storage or cloud storage of the terminal device for subsequent use.

[0061] Parse the updated document: Use the same parsing method as in step S201 to convert the updated model design specification document into structured data. Update the specification knowledge base: Based on the parsed structured data, update the specification knowledge base in the terminal device to ensure that it contains the latest model checking rules. Data collection and annotation: Since the specification knowledge base has been updated, it may be necessary to re-collect or annotate some training data to ensure that they comply with the new rules. This can include updating existing positive and negative samples, and adding new samples to cover the new rules. Training set construction: Use the updated specification knowledge base and the re-collected / annotated data to build a new training set. Load the previously trained large model as a starting point to speed up the training process. Using the new training set, fine-tune or retrain the loaded model to ensure that it can recognize and apply the updated model checking rules. During the training process, evaluate the performance of the model regularly to ensure that its recognition ability of the new rules reaches the expected accuracy. Save the trained new model to the local storage or cloud storage of the terminal device for subsequent use.

[0062] Furthermore, the user executes the model rule editing process through the interface including:

[0063] After loading the model design specification document, the control point information table is displayed on the control point management interface. The control point information table includes a table header and multiple data rows. Each data row corresponds to a model checking rule. The table header includes the following fields: control category, control point category, control point subcategory, control point (indicating the specific content of the model checking rule), control point code, control type, deactivated status, release identifier, whether solidification is required, whether the tool supports it, and creator;

[0064] The operation of adding a new data row, modifying a data row, or deleting a data row is performed based on the user's interactive instruction on the control point information table.

[0065] For details, see Figure 3 As shown, the terminal device loads the model design specification document from the specified location. Use an appropriate parser (such as an XML parser, a JSON parser, etc., depending on the document format) to parse the document content into structured data. Extract the control point information from the parsed structured data, including control category, control point category, control point subcategory, control point, control point code, control type, deactivated status, release identifier, whether solidification is required, whether the tool supports it, and creator and other fields.

[0066] Terminal equipment construction Figure 3The control point management interface shown is used to display the control point information table. Design a table header in the UI interface, including all the fields mentioned above as column headers. Based on the extracted control point information, dynamically generate multiple data rows and fill each data row into the table, ensuring that the value of each field correctly corresponds to the corresponding column. Add interactive elements (such as buttons, drop-down menus, etc.) next to or above the table to allow users to perform operations such as adding, modifying, and deleting.

[0067] The terminal device listens for user interactions on the control point information table, such as clicking a button, selecting a drop-down menu item, etc. When the user performs an interactive operation, the terminal device parses the user's instructions to determine the type of operation the user wants to perform (add, modify, delete) and the relevant control point information. Based on the new control point information entered by the user, a new data row is created and added to the end of the table. At the same time, the new information is saved to the corresponding part of the model design specification document (if the document is editable). The data row specified by the user is found and its field values ​​are updated to reflect the user's modification. At the same time, the corresponding information in the model design specification document is updated. The data row specified by the user is deleted from the table and the corresponding control point information is removed from the model design specification document. After performing the data row operation, the terminal device updates the control point information table in real time to reflect the latest control point information. If the user performs a modification or deletion operation and the model design specification document is editable, the terminal device saves the changes to the document and updates the storage location of the document (such as the local file system, cloud storage, etc.).

[0068] Furthermore, Figure 3 The control point management interface shown also includes: a filtering element, which is an input box. Based on the filtering conditions input by the user in the filtering element, the corresponding data rows are queried in the control point information table. The filtering conditions include: the same or multiple field values ​​of control category, control point category, control point subcategory, control point and current status.

[0069] Furthermore, in the process of training the large model, focus on the model design field, covering the professional terms, common concepts, technical methods, etc. in this field. Consider the application scenarios of the large model, such as natural language processing, computer vision, etc., to ensure that the vocabulary is closely related to the actual application of the model. Collect vocabulary from authoritative professional books, academic papers, technical documents and other sources. Filter out vocabulary that is too obscure or irrelevant to the application scenario of the large model to ensure the practicality and pertinence of the vocabulary. Sort the filtered vocabulary according to certain rules (such as alphabetical order, category, etc.). Provide accurate definitions and explanations for each vocabulary so that the large model can refer to it when understanding it. Use word embedding technology (such as Word2Vec, GloVe, etc.) to convert the vocabulary in the vocabulary into low-dimensional vector representations. These vectors can capture the semantic relationship and similarity between words, which helps the large model better understand the meaning of the words. In the training process of the large model, the vocabulary embedding is used as additional input information. By adjusting the parameters of the model, the model can learn the semantic features and usage of the words in the vocabulary. When the large model processes natural language tasks, the vocabulary information in the vocabulary is used to enhance the model's understanding of the input text. By identifying specialized terms and concepts in text, the model can more accurately understand the intent and meaning of the text. When generating natural language output, the model can refer to the vocabulary and definitions in the vocabulary to generate more accurate and professional answers or explanations. This helps improve the quality and credibility of the output and enhances users' trust in the big model.

[0070] Further, in some embodiments of the present application, the use of the trained large model to process the model to be checked in the research and development project and then output an inspection report includes:

[0071] After loading the model to be checked, the model data table is displayed in the model management interface;

[0072] Based on a check instruction triggered by a user on the model data table, inputting the data in the model data table into the trained large model;

[0073] A model checking window pops up on the model management interface, and non-compliant field values ​​are marked in the model data table; the model checking window includes: problem description information and modification suggestion information.

[0074] Specifically, the terminal device loads the model file to be checked from a specified storage location (such as a local file system, a database, a cloud storage, etc.). The model file is parsed to extract the structure and data information of the model. Figure 4 Build a data table on the model management interface shown in the figure. The data table should include all relevant data fields of the model, such as model name, version, parameters, input / output format, etc. Each field corresponds to a column, and each model instance corresponds to a row.

[0075] exist Figure 4 The model management interface shown provides interactive elements (such as buttons, menu items, etc.) to allow users to trigger model check instructions. Users can select one or more model instances for inspection, or batch check the entire model data table. The terminal device receives the inspection instruction triggered by the user, parses the instruction content, and determines the model instance and data field that the user wants to check.

[0076] According to the model instance and data field selected by the user, the corresponding data is extracted from the model data table. Preprocess the data, such as format conversion and data cleaning, to ensure that it meets the input requirements of the big model. Input the preprocessed data into the trained big model. The big model infers the input data, determines whether the data is compliant, and generates corresponding inspection results. According to the inspection results of the big model, Figure 4 A model check window pops up on the model management interface shown. The check window includes two parts: problem description information and modification suggestion information. The problem description information is used to list all non-compliant field values ​​and their corresponding error types or problem descriptions. The modification suggestion information provides modification suggestions or solutions for each non-compliant field value.

[0077] In the model data table, all non-compliant field values ​​are marked. The marking method can be to change the color of the field value, add a highlight border, display an error icon next to the field, etc. The marking should be obvious enough so that users can quickly identify all non-compliant field values. Users can view the problem description information and modification suggestion information in the model check window. Users can modify the non-compliant field values ​​according to the suggestions and re-trigger the check instruction to verify whether the modification is effective. The terminal device can collect user feedback on the model check function, such as the accuracy of the check results, the practicality of the modification suggestions, etc. Based on user feedback, the model check function is optimized and improved to improve its accuracy and user experience.

[0078] Through the above steps, the terminal device can load the model to be checked and display the model data table on the model management interface. When the user triggers the check instruction, the terminal device inputs the data in the model data table into the trained large model, pops up the model check window on the model management interface, and marks the non-compliant field values ​​in the model data table. This helps users quickly identify and fix non-compliant issues in the model.

[0079] Furthermore, the data in the model data table is editable, and based on the user's interactive instructions to the model data table, the model data table can be deleted, modified, or newly added with data rows.

[0080] Specifically, the terminal device provides interactive elements for the model data table on the model management interface, such as buttons, drop-down menus, or quick operation icons, allowing users to trigger instructions for deleting data rows, modifying data rows, or adding new data rows. These interactive elements should clearly mark their functions so that users can easily identify and use them. When the user clicks or selects an interactive element, the terminal device receives the instruction triggered by the user. The terminal device parses the instruction content to determine the type of operation the user wants to perform (delete, modify, add) and related data row information (such as row number, unique identifier, etc.). Based on the row number or unique identifier provided by the user, the terminal device locates the data row to be deleted in the model data table. The terminal device removes the data row from the data table and updates the displayed content of the data table. If the model data table is associated with a back-end database, the terminal device also needs to send a deletion request to the back-end database to ensure that the data in the database is also deleted synchronously.

[0081] Modify data row: Based on the row number or unique identifier provided by the user, the terminal device locates the data row to be modified in the model data table. The terminal device pops up a data editing window or provides an in-place editing function to allow the user to modify the content of the data row. After the user completes the modification, the terminal device verifies the legitimacy and accuracy of the modified content. If the verification passes, the terminal device updates the data row content in the data table and (if applicable) sends an update request to the backend database.

[0082] Adding a new data row: The terminal device provides a data input window or form to allow the user to enter the content of the new data row. After the user completes the input, the terminal device verifies the legality and accuracy of the input content. If the verification passes, the terminal device adds a new data row to the model data table and fills it with the content entered by the user. At the same time, the terminal device (if applicable) sends an insert request to the backend database to save the information of the new data row.

[0083] After deleting, modifying or adding a new data row, the terminal device updates the display content of the model data table in real time to reflect the latest data status. If the operation involves the backend database, the terminal device should wait until the database operation is completed before updating the data table.

[0084] The embodiments of the present application specifically include the following beneficial effects:

[0085] By obtaining the model design specification documents of the R&D project and converting them into a specification knowledge base, the model checking rules are stored in a systematic and structured manner. This not only improves the queryability and reusability of the rules, but also provides a solid foundation for subsequent automated inspections.

[0086] Building a training set based on the specification knowledge base enables the large model to learn the core elements and inspection logic of the model design specification. The training set generation process ensures that the model can be exposed to various possible model design scenarios during the training process, improving the generalization ability of the model.

[0087] Through pre-training, the large model can learn a lot of prior knowledge and improve its understanding of the model design specifications. Subsequently, through fine-tuning for specific R&D projects, the model can further adapt to the specific needs of the project and improve the accuracy and pertinence of the inspection. This training method not only takes advantage of large-scale data, but also ensures the effectiveness of the model in practical applications.

[0088] By using the trained large model to process the models to be inspected in the R&D project, an inspection report can be automatically generated. The problem description information and modification suggestion information in the report directly point to the non-compliant parts of the model design, helping designers to quickly locate the problem and repair it. This not only significantly improves the inspection efficiency, but also reduces the risk of human omissions.

[0089] Through automated inspection, consistency between different inspectors is ensured, avoiding potential risks caused by human differences. At the same time, as design specifications are continuously updated and improved, large models can be adapted to new requirements through simple retraining or fine-tuning, improving the flexibility and adaptability of inspection.

[0090] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0091] See also Figure 5 , which shows a schematic diagram of the structure of a model design normative inspection device provided by an exemplary embodiment of the present application, hereinafter referred to as device 5. The device 5 can be implemented as all or part of a terminal device through software, hardware, or a combination of both. The device 5 includes: an acquisition unit 501, a construction unit 502, a training unit 503, and a checking unit 504.

[0092] The acquisition unit 501 is used to acquire a model design specification document of a research and development project, and generate a specification knowledge base according to the model design specification document; the model design specification document includes a plurality of model checking rules;

[0093] A construction unit 502, configured to construct a training set according to the specification knowledge base;

[0094] A training unit 503 is used to train the large model based on the training set by using pre-training and fine-tuning methods;

[0095] The inspection unit 504 is used to use the trained large model to process the model to be inspected in the R&D project and then output an inspection report, wherein the inspection report includes: problem description information and modification suggestion information.

[0096] In one or more possible embodiments, it further includes:

[0097] An updating unit, configured to perform operations of adding a new model checking rule, modifying a model checking model, or deleting a model checking rule on the model design specification document based on an editing instruction of a user;

[0098] Retrain large models using the updated model design specification document.

[0099] In one or more possible embodiments, adding a new model checking rule, modifying a model checking model, or deleting a model checking rule to the model design specification document based on the user's editing instruction includes:

[0100] After the model design specification document is loaded, a control point information table is displayed on the control point management interface. The control point information table includes a table header and multiple data rows, each data row corresponds to a model checking rule, and the table header includes the following fields: control category, control point category, control point subcategory, control point, control point code, control type, deactivated status, release identifier, whether solidification is required, whether the tool supports it, and creator;

[0101] The operation of adding a new data row, modifying a data row, or deleting a data row is performed based on the user's interactive instruction on the control point information table.

[0102] In one or more possible embodiments, the control point management interface further includes: a screening element; and further includes:

[0103] A filtering unit is used to query corresponding data rows in the control point information table based on the filtering conditions input by the user in the filtering element, and the filtering conditions include: one or more of the field values ​​of control category, control point category, control point subcategory, control point and current status.

[0104] In one or more possible embodiments, the vocabulary of the model design domain is used to enhance the understanding ability of large models.

[0105] In one or more possible embodiments, the using of the trained large model to process the model to be checked in the R&D project and then outputting an inspection report includes:

[0106] After loading the model to be checked, the model data table is displayed in the model management interface;

[0107] Based on a check instruction triggered by a user on the model data table, inputting the data in the model data table into the trained large model;

[0108] A model checking window pops up on the model management interface, and non-compliant field values ​​are marked in the model data table; the model checking window includes: problem description information and modification suggestion information.

[0109] In one or more possible embodiments, it further includes:

[0110] The editing unit is used to delete data rows, modify data rows or add data rows to the model data table based on the user's interactive instructions to the model data table.

[0111] It should be noted that the device 5 provided in the above embodiment only uses the division of the above functional modules as an example when executing the model design normative inspection method. In actual applications, the above functional distribution can be completed by different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above functions. In addition, the model design normative inspection device provided in the above embodiment and the model design normative inspection method embodiment belong to the same concept. The implementation process is detailed in the method embodiment, which will not be repeated here.

[0112] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0113] The present application also provides a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded and executed by a processor as described above. Figure 2 The method steps of the embodiment shown in the figure can be found in the specific implementation process. Figure 2 The specific description of the illustrated embodiment will not be repeated here.

[0114] The present application also provides a computer program product, which stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the model design normative checking method described in the above embodiments.

[0115] See also Figure 6 , which is a schematic diagram of the structure of a terminal device provided in an embodiment of the present application. Figure 6 As shown, the terminal device 600 may include: at least one processor 601 , at least one network interface 604 , a user interface 603 , a memory 605 , and at least one communication bus 602 .

[0116] The communication bus 602 is used to realize the connection and communication between these components.

[0117] The user interface 603 may include a display screen (Display) and a camera (Camera), and the optional user interface 603 may also include a standard wired interface and a wireless interface.

[0118] The network interface 604 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0119] Among them, the processor 601 may include one or more processing cores. The processor 601 uses various interfaces and lines to connect various parts within the entire terminal device 600, and executes various functions and processes data of the terminal device 600 by running or executing instructions, programs, code sets or instruction sets stored in the memory 605, and calling data stored in the memory 605. Optionally, the processor 601 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 601 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 601, and it can be implemented separately through a chip.

[0120] Among them, the memory 605 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 605 includes a non-transitory computer-readable storage medium. The memory 605 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 605 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store data involved in the above-mentioned method embodiments, etc. The memory 605 may optionally be at least one storage device located away from the aforementioned processor 601. As Figure 6 As shown, the memory 605 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program.

[0121] exist Figure 6 In the terminal device 600 shown in FIG. 1 , the user interface 603 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 601 can be used to call the application stored in the memory 605 and specifically execute the following steps: Figure 2 The specific process can be referred to Figure 2 As shown, no further description is given here.

[0122] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0123] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A method for checking the standardization of model design, characterized in that: include: Obtaining a model design specification document of a research and development project, and generating a specification knowledge base based on the model design specification document; The model design specification document includes a plurality of model checking rules; Constructing a training set according to the specification knowledge base; Based on the training set, a large model is trained by pre-training and fine-tuning; The trained large model is used to process the model to be checked in the R&D project and then output an inspection report, which includes: problem description information and modification suggestion information.

2. The method according to claim 1, characterized in that Also includes: Based on the user's editing instruction, perform operations of adding a model checking rule, modifying a model checking model, or deleting a model checking rule on the model design specification document; Retrain large models using the updated model design specification document.

3. The method according to claim 2, characterized in that Adding a new model checking rule, modifying a model checking model, or deleting a model checking rule to the model design specification document based on the user's editing instruction includes: After the model design specification document is loaded, a control point information table is displayed on the control point management interface. The control point information table includes a table header and multiple data rows, each data row corresponds to a model checking rule, and the table header includes the following fields: control category, control point category, control point subcategory, control point, control point code, control type, deactivated status, release identifier, whether solidification is required, whether the tool supports it, and creator; The operation of adding a new data row, modifying a data row, or deleting a data row is performed based on the user's interactive instruction on the control point information table.

4. The method according to claim 3, characterized in that The control point management interface also includes: a screening element; The method further comprises: Based on the filtering conditions input by the user in the filtering element, the corresponding data rows are queried in the control point information table, and the filtering conditions include: the same or multiple field values ​​of control category, control point category, control point subcategory, control point and current status.

5. The method according to claim 1 or 2 or 3 or 4, characterized in that: Enhance the understanding of large models by leveraging vocabulary and knowledge from the domain in which the model was designed.

6. The method according to claim 5, characterized in that The method of using the trained large model to process the model to be inspected in the R&D project and then outputting an inspection report includes: After loading the model to be checked, the model data table is displayed in the model management interface; Based on a check instruction triggered by a user on the model data table, inputting the data in the model data table into the trained large model; A model checking window pops up on the model management interface, and non-compliant field values ​​are marked in the model data table; the model checking window includes: problem description information and modification suggestion information.

7. The method according to claim 6, characterized in that Also includes: Based on the user's interactive instructions on the model data table, the model data table deletes data rows, modifies data rows, or adds data rows.

8. A device for checking the standardization of model design, characterized in that: include: An acquisition unit, used to acquire a model design specification document of a research and development project, and to generate a specification knowledge base according to the model design specification document; The model design specification document includes a plurality of model checking rules; A construction unit, used for constructing a training set according to the specification knowledge base; A training unit, used for training the large model by pre-training and fine-tuning based on the training set; The inspection unit is used to use the trained large model to process the model to be inspected in the R&D project and then output an inspection report, wherein the inspection report includes: problem description information and modification suggestion information.

9. A computer storage medium, characterized in that The computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method steps according to any one of claims 1 to 7.

10. A terminal device, characterized in that: include: A processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps as claimed in any one of claims 1 to 7.

Images