Enterprise-level AI application construction method, computer equipment and medium

Abstract

The invention provides an enterprise-level AI application construction method, computer equipment and a medium. The method comprises the following steps: acquiring a plurality of to-be-trained models in computer equipment for constructing enterprise-level AI applications; displaying a first page comprising a plurality of to-be-trained models to determine a target to-be-trained model in the plurality of to-be-trained models; determining a plurality of candidate data sets matched with the target to-be-trained model in a plurality of to-be-used data sets in the computer equipment; displaying a second page comprising a plurality of candidate data sets to determine a target training set; and training the target to-be-trained model by using the target training set to obtain a target AI application. According to the method and the device, the to-be-trained models and the to-be-used data sets are summarized in the computer equipment, and after the target to-be-trained model is determined, the candidate data set is further determined based on the target to-be-trained model, so that the training of the target AI application is simpler and more efficient, and the construction efficiency of the AI application is improved.

Description

Technical Field

[0001] The present application relates to the field of AI application technology, and specifically to a method for constructing enterprise-level AI applications, computer equipment, and media. Background Art

[0002] When an enterprise needs to build an AI (Artificial Intelligence) application, relevant technical personnel often search one by one in multiple different data sources to manually collect models and training sets that meet the requirements, and then use the collected training sets to train the collected models to obtain the AI ​​applications required by the enterprise.

[0003] However, this manual process of building AI applications is too cumbersome and inconvenient, which will also make the construction efficiency of AI applications low. Summary of the invention

[0004] The embodiments of the present application provide a method, computer device, and medium for constructing an enterprise-level AI application, aiming to make the training of the target AI application simpler and more efficient, thereby improving the efficiency of constructing the AI ​​application.

[0005] In a first aspect, an embodiment of the present application provides a method for constructing an enterprise-level AI application, and the method for constructing an enterprise-level AI application includes:

[0006] Acquire a plurality of preset models to be trained in a computer device used to build the enterprise-level AI application;

[0007] Displaying a first page including a plurality of the models to be trained, so as to determine a target model to be trained among the plurality of the models to be trained based on a first instruction on the first page;

[0008] Determining, from among a plurality of preset data sets to be used in the computer device, a plurality of candidate data sets that match the target model to be trained;

[0009] displaying a second page including the plurality of candidate data sets, to determine a target training set among the plurality of candidate data sets based on a second instruction on the second page;

[0010] The target training set is used to train the target model to be trained to obtain a target AI application.

[0011] In some embodiments, the using the target training set to train the target to-be-trained model to obtain a target AI application includes:

[0012] Acquire configuration data of the target AI application, wherein the configuration data includes at least one of a target knowledge base, a target calling tool, and a target prompt word associated with the target AI application;

[0013] The target training set and the configuration data are used to train the target model to be trained to obtain the target AI application.

[0014] In some embodiments, obtaining the configuration data of the target AI application includes:

[0015] Get uploaded knowledge documents;

[0016] Slicing the knowledge document to obtain segmented data of the knowledge document;

[0017] Performing vectorization processing on the segmented data to obtain vector features of the segmented data;

[0018] The target knowledge base in the configuration data is generated based on the vector features.

[0019] In some embodiments, obtaining the configuration data of the target AI application includes:

[0020] Acquire a plurality of candidate calling tools preset in the computer device, wherein the candidate calling tools include at least one of a device preset tool, a device external tool, and a workflow tool;

[0021] A third page including the plurality of candidate calling tools is displayed to determine a target calling tool among the plurality of candidate calling tools.

[0022] In some embodiments, the workflow tool is obtained by the following steps:

[0023] Acquire a set data processing flow chart, wherein the data processing flow chart includes a plurality of data processing nodes, each of which is used to process data input by a previous data processing node of the data processing node;

[0024] The data processing flow chart is published in the computer device to obtain the workflow tool.

[0025] In some embodiments, the first page is obtained by the following steps:

[0026] Determine a plurality of first models and a plurality of second models among the plurality of models to be trained based on a model type of each of the models to be trained, wherein the model type includes at least one of a locally uploaded model and a device preset model;

[0027] Obtaining the model result accuracy and the model historical usage times of each of the first models;

[0028] Determining a model recommendation score for each of the first models based on the accuracy of the model results and the number of times the model has been used in history;

[0029] Sorting the plurality of first models based on the model recommendation score of each of the first models to obtain a first order;

[0030] sorting the plurality of said second models to obtain a second order;

[0031] The second sequence and the first sequence are concatenated to obtain a third sequence;

[0032] The first page is generated, wherein the first page includes a plurality of the models to be trained arranged in the third order.

[0033] In some embodiments, determining a plurality of candidate data sets that match the target to-be-trained model from a plurality of preset to-be-used data sets in the computer device includes:

[0034] Obtaining the enterprise domain of the enterprise to which the currently logged-in user of the computer device belongs;

[0035] Determine a plurality of first data sets related to the enterprise field among the plurality of data sets to be used;

[0036] Obtaining the training data file type supported by the target model to be trained;

[0037] Determining, from among the plurality of the first data sets, a plurality of second data sets that conform to the type of the training data file;

[0038] Each of the second data sets is used as a candidate data set, thereby obtaining a plurality of candidate data sets.

[0039] In some embodiments, the second page is obtained by the following steps:

[0040] Obtaining model parameter quantities in the target model to be trained;

[0041] Determining the expected number of training data that matches the model parameter quantity;

[0042] Determine the difference between the actual number of training data items in each of the candidate data sets and the expected number of training data items;

[0043] Based on the difference, sorting the plurality of candidate data sets to obtain a fourth order;

[0044] The second page is generated, wherein the second page includes the plurality of candidate data sets arranged in the fourth order.

[0045] In a second aspect, an embodiment of the present application provides a device for constructing an enterprise-level AI application, and the device for constructing an enterprise-level AI application includes:

[0046] A first acquisition module, used to acquire a plurality of preset models to be trained in a computer device used to build the enterprise-level AI application;

[0047] A first determination module, configured to display a first page including a plurality of the models to be trained, so as to determine a target model to be trained among the plurality of the models to be trained based on a first instruction on the first page;

[0048] A second determination module is used to determine a plurality of candidate data sets that match the target to-be-trained model from a plurality of preset to-be-used data sets in the computer device;

[0049] A third determination module is used to display a second page including the plurality of candidate data sets, so as to determine a target training set among the plurality of candidate data sets based on a second instruction on the second page;

[0050] The model training module is used to train the target model to be trained using the target training set to obtain the target AI application.

[0051] In a third aspect, an embodiment of the present application provides a computer device, comprising a processor and a memory, wherein the memory stores a computer program, and the computer program is configured to be executed by the processor to implement a method for constructing an enterprise-level AI application as described in any one of the above items.

[0052] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is configured to be executed by a processor to implement a method for constructing an enterprise-level AI application as described in any of the above items.

[0053] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program or instructions, which are executed by a processor to implement a method for constructing an enterprise-level AI application as described in any of the above items.

[0054] Beneficial effects of the embodiments of the present application:

[0055] In an embodiment of the present application, by aggregating multiple models to be trained and data sets to be used in a computer device for training, there is no need to manually collect models and training sets that meet the requirements, and after determining the target model to be trained, candidate data sets among multiple data sets to be used are further determined based on the target model to be trained, making the training of the target AI application simpler and more efficient, thereby improving the efficiency of building AI applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 This is a flowchart of an embodiment of a method for constructing an enterprise-level AI application provided by an embodiment of the present application;

[0058] Figure 2 It is a flowchart of another embodiment of a method for constructing an enterprise-level AI application provided by an embodiment of the present application;

[0059] Figure 3 It is a flowchart of another embodiment of a method for constructing an enterprise-level AI application provided by an embodiment of the present application;

[0060] Figure 4 This is a flowchart of another embodiment of a method for constructing an enterprise-level AI application provided by an embodiment of the present application;

[0061] Figure 5 This is a flowchart of another embodiment of a method for constructing an enterprise-level AI application provided by an embodiment of the present application;

[0062] Figure 6 It is a schematic diagram of the structure of an embodiment of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0064] In the description of this application, the meaning of "plurality" is two or more, unless otherwise expressly and specifically defined. In addition, in the description of this application, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0065] In order to improve the efficiency of building AI applications, the embodiments of the present application provide a method, computer device, and medium for building enterprise-level AI applications. By aggregating multiple models to be trained and data sets to be used in a computer device for training, there is no need to manually collect models and training sets that meet the requirements. After determining the target model to be trained, the candidate data sets in multiple data sets to be used are further determined based on the target model to be trained, so that the training of the target AI application is simpler and more efficient, thereby improving the efficiency of building AI applications. For specific solutions, please refer to the following specific description.

[0066] In the first aspect, the embodiments of the present application provide a method for constructing an enterprise-level AI application. Figure 1 , Figure 1 The following is a flow chart of an embodiment of a method for constructing an enterprise-level AI application. Figure 1 In the method for building an enterprise-level AI application, the method may include:

[0067] 101. Obtain a plurality of preset models to be trained in a computer device used to build the enterprise-level AI application.

[0068] In an embodiment of the present application, a one-stop construction platform for enterprise-level AI (Artificial Intelligence) applications is pre-deployed in a computer device used to build enterprise-level AI applications. The platform is oriented to enterprise users, and enterprise users can log in to the platform to build the AI ​​applications required by the enterprise.

[0069] The computer device is pre-installed with multiple models to be trained, which are generally neural network models, but can also be other machine learning models, which are not limited here. Enterprise users can directly select the required model in the one-stop enterprise-level AI application construction platform without having to search for the required model one by one in multiple different data sources, so as to simplify the steps of building AI applications.

[0070] 102. Display a first page including multiple models to be trained, and determine a target model to be trained among the multiple models to be trained based on a first instruction on the first page.

[0071] In an embodiment of the present application, a first page including multiple models to be trained is generated, and then the first page is displayed in a construction platform for enterprise-level AI applications, so that an enterprise user can manually trigger a first instruction in the first page to select a desired model, or automatically trigger a first instruction on the first page to select a specified model to be trained in the first page, such as the first model to be trained displayed in the first page. The selected model to be trained in the first page is the target model to be trained among the multiple models to be trained. For example, the first page may include profile information of multiple models to be trained, so that an enterprise user can manually trigger a first instruction based on the profile information to select a desired model.

[0072] 103. Determine a plurality of candidate data sets that match the target model to be trained from among the plurality of preset data sets to be used in the computer device.

[0073] In the embodiment of the present application, the computer device is also pre-installed with multiple data sets to be used, and the data sets to be used are used for model training. Enterprise users can directly select the required data sets to be used in the one-stop construction platform for enterprise-level AI applications, without having to search for the required data sets one by one in multiple different data sources, so as to simplify the steps of building AI applications.

[0074] Since there are many data sets to be used that are pre-set in the computer device, the enterprise user still needs to screen the required data sets. Therefore, in the embodiment of the present application, multiple candidate data sets that match the target model to be trained can be determined from the multiple data sets to be used to simplify the steps for the enterprise user to screen the data sets. The specific matching rules between the multiple candidate data sets and the target model to be trained are not limited here.

[0075] 104. Display a second page including multiple candidate data sets, and determine a target training set from the multiple candidate data sets based on a second instruction on the second page.

[0076] In an embodiment of the present application, a second page including multiple candidate data sets is generated, and then the second page is displayed in the construction platform for enterprise-level AI applications, so that the enterprise user can manually trigger the second instruction in the second page to select the desired data set, or automatically trigger the second instruction on the second page to select a specified data set in the second page, such as the first candidate data set displayed in the second page. The candidate data set selected in the second page is the target training set among the multiple candidate data sets. For example, the second page may include profile information of multiple candidate data sets, so that the enterprise user can manually trigger the second instruction based on the profile information to select the desired candidate data set.

[0077] 105. Use the target training set to train the target model to obtain the target AI application.

[0078] In an embodiment of the present application, the computer device also provides a model training function, that is, after determining the target model to be trained and the target training set, the target model to be trained can be automatically trained using the target training set in a software environment pre-installed in the computer device based on the hardware resources of the computer device, without the need for enterprise users to deploy the corresponding software environment and hardware resources on the user's local end, so as to further simplify the steps of building AI applications.

[0079] It can be seen that in the above embodiments of the present application, by aggregating multiple models to be trained and data sets to be used in a computer device for training, there is no need to manually collect models and training sets that meet the requirements, and after determining the target model to be trained, the candidate data sets in multiple data sets to be used are further determined based on the target model to be trained, making the training of the target AI application simpler and more efficient, thereby improving the efficiency of building AI applications. In addition, through the construction platform for enterprise-level AI applications, the full life cycle process management of the target AI application can be achieved, effectively solving the various challenges faced by enterprises in the implementation of the target AI application.

[0080] In some embodiments of the present application, the computer device also provides more diverse model functions to make the constructed AI application more adaptable to the needs of the enterprise. Figure 2 As shown, in Figure 1 Based on the illustrated embodiment, using the target training set to train the target model to be trained to obtain the target AI application may include:

[0081] 201. Obtain configuration data of a target AI application, where the configuration data includes at least one of a target knowledge base, a target calling tool, and a target prompt word associated with the target AI application.

[0082] In an embodiment of the present application, the configuration data of the target AI application refers to data used to expand or enhance the capabilities of the target AI application, and the configuration data may include at least one of a target knowledge base, a target calling tool, and a target prompt word associated with the target AI application. Taking the target knowledge base as an example, the target AI application can generate the results that the target AI application needs to output based on the professional knowledge stored in the target knowledge base, so that the results output by the target AI application are more accurate. Taking the target calling tool as an example, the target AI application can call the target calling tool so that the target calling tool collects or processes data in a corresponding manner, and then obtains the results that the target AI application needs to output, so that the results output by the target AI application are more accurate. Taking the target prompt word as an example, the target AI application can use the target prompt word as a constraint condition for the result output by the target AI application, so that the result output by the target AI application meets the requirements of the target prompt word, so that the result output by the target AI application is more accurate.

[0083] In some embodiments of the present application, taking the target knowledge base in the configuration data as an example, the enterprise user can directly select at least one knowledge base as the target knowledge base from the multiple knowledge bases preset in the computer device, or the enterprise user can customize the target knowledge base by uploading files to the computer device. Specifically, obtaining the configuration data of the target AI application can include: obtaining the uploaded knowledge document, which can be obtained by uploading the knowledge document to the computer device by the enterprise user, and the knowledge document can be a document in the form of an unstructured file, a table, a question and answer set, etc.; slicing the knowledge document to obtain segmented data of the knowledge document, and the slicing process can be implemented by tools such as a text segmenter; vectorizing the segmented data to obtain vector features of the segmented data to better reflect the features of the segmented data and facilitate computer processing and analysis; generating the target knowledge base in the configuration data based on the vector features, so that the target model to be trained and / or the target AI application can use the vector features to quickly find the required knowledge in the target knowledge base.

[0084] In some embodiments of the present application, taking the target calling tool in the configuration data as an example, the enterprise user can directly select at least one candidate calling tool as the target calling tool from the multiple candidate calling tools preset in the computer device. Specifically, obtaining the configuration data of the target AI application may include: obtaining multiple candidate calling tools preset in the computer device, wherein the candidate calling tools include at least one of the device preset tools, device external tools, and workflow tools, the device preset tools are tools provided by the computer device itself, the device external tools are tools in the form of plug-ins provided by external data sources of the computer device, and the workflow tools are tools arranged by enterprise users based on actual needs; displaying a third page including multiple candidate calling tools to determine the target calling tool among the multiple candidate calling tools.

[0085] In some embodiments of the present application, the specific generation method of the workflow tool is described. Specifically, the workflow tool is obtained by the following steps: obtaining a set data processing flow chart, the data processing flow chart includes multiple data processing nodes, each data processing node is used to process the data input by the previous data processing node of the data processing node, each data processing node is provided with a corresponding data processing action, the data processing action can be, for example, to classify the input text, and then input the text to the next data processing node corresponding to the classified category, and the data processing action can be, for example, based on the input text, to search in the specified data source, obtain the search results, and then input the search results to the next data processing node, wherein the number of data processing nodes, connection relationships and data processing actions in the data processing flow chart can be set by enterprise users based on actual needs; the data processing flow chart is published in a computer device to obtain a workflow tool, that is, the data processing process contained in the data processing flow chart is published as a data processing process in a workflow tool, so as to facilitate the workflow tool as a candidate calling tool. It can be seen that the embodiment of the present application realizes the custom generation of tools, thereby meeting the more diverse AI application needs of enterprise users.

[0086] 202. Train the target model to be trained using the target training set and configuration data to obtain the target AI application.

[0087] In an embodiment of the present application, since the configuration data of the target AI application can make the results output by the AI ​​application more accurate, in the process of training the target model to be trained using the target training set, the configuration data can be applied to the output results of the target model to be trained, so that the training of the target model to be trained is faster and more accurate.

[0088] It can be seen that in the above embodiments of the present application, by using the target knowledge base, target calling tool, target prompt words and other configuration data associated with the target AI application to train the target model to be trained, a more accurate target AI application can be obtained, that is, the target AI application is more in line with enterprise needs.

[0089] In some embodiments of the present application, Figure 3 As shown, in Figure 1 or Figure 2 Based on the illustrated embodiment, the first page can be obtained by the following steps:

[0090] 301. Based on the model type of each model to be trained, determine multiple first models and multiple second models from the multiple models to be trained, where the model type includes at least one of a locally uploaded model and a device preset model.

[0091] In the embodiment of the present application, the computer device can also allow enterprise users to upload models as models to be trained. The models uploaded by enterprise users are locally uploaded models. Therefore, in the computer device, the models to be trained can be classified to obtain the model types of the models to be trained. Among them, the model type of the first model is the device preset model, and the model type of the second model is the locally uploaded model.

[0092] 302. Obtain the model result accuracy and historical usage count of each first model.

[0093] In the embodiments of the present application, the accuracy of the model result refers to the accuracy of the result output by the first model, and the accuracy of the model result can be obtained by evaluating each first model in advance by a computer device. Among them, the evaluation method of the first model can be, for example, CMMLU (Chinese Massive Multitask Language Understanding), C-Eval (Chinese Basic Model Evaluation), AGIEval (General Ability Evaluation), etc.

[0094] In the embodiment of the present application, the number of historical usages of a model refers to the cumulative number of times the corresponding first model is used for model training on a computer device. The number of historical usages of a model can be obtained by statistics of the computer device.

[0095] 303. Determine a model recommendation score for each first model based on the accuracy of the model results and the number of times the model has been used in history.

[0096] In the embodiment of the present application, for each first model, the model recommendation score of the first model is positively correlated with the accuracy of the model result and the number of times the model has been used in history. That is, when the model result accuracy of the first model is higher, the model recommendation score of the first model is also higher. Similarly, when the model history usage number of the first model is higher, the model recommendation score of the first model is also higher.

[0097] In some embodiments of the present application, determining the model recommendation score of each first model based on the model result accuracy and the number of historical uses of the model may include: for each first model, taking a weighted sum of the model result accuracy and the number of historical uses of the model of the first model to obtain the model recommendation score of the first model.

[0098] 304. Sort the multiple first models based on the model recommendation score of each first model to obtain a first order.

[0099] In an embodiment of the present application, in a first order, a plurality of first models are sorted from high to low according to the model recommendation score, so that enterprise users can preferentially view the first models with higher model recommendation scores.

[0100] 305. Sort the multiple second models to obtain a second order.

[0101] In an embodiment of the present application, since the model type of the second model is a local upload model, and the local upload model may better meet the needs of enterprise users, multiple second models can be sorted independently of the first model to obtain a second order.

[0102] In some embodiments of the present application, sorting multiple second models to obtain a second order may include: obtaining the model result accuracy and the historical usage count of each second model; determining the model recommendation score of the second model based on the model result accuracy and the historical usage count of the model of the second model; and sorting multiple second models based on the model recommendation score of each second model to obtain a second order.

[0103] In some embodiments of the present application, since some locally uploaded models may not be able to obtain the accuracy of model results, multiple second models are sorted to obtain a second order, which can include: obtaining the upload time point and the historical usage times of each second model; determining the model recommendation score of the second model based on the upload time point and the historical usage times of the model; and sorting the multiple second models based on the model recommendation score of each second model to obtain a second order, so that the second model with an upload time point closer to the current time point and a greater number of historical usage times of the model is ranked higher in the second order.

[0104] 306. The second sequence and the first sequence are concatenated to obtain a third sequence.

[0105] In an embodiment of the present application, when the second order and the first order are spliced ​​together, the first order can be spliced ​​after the second order. In this way, in the third order obtained by splicing, the ranking of each second model can be higher than the ranking of each first model, so that enterprise users can give priority to viewing the second models uploaded locally.

[0106] 307. Generate a first page, where the first page includes multiple models to be trained arranged in a third order.

[0107] In an embodiment of the present application, a first page including multiple models to be trained arranged in a third order is generated, and then the first page is displayed in a construction platform for enterprise-level AI applications, so that enterprise users can find the models to be trained that they want to train more quickly in the first page.

[0108] In some embodiments of the present application, Figure 4 As shown, in Figures 1 to 3Based on any of the embodiments shown, determining multiple candidate data sets that match the target model to be trained from multiple preset data sets to be used in the computer device may include:

[0109] 401. Obtain the enterprise domain of the enterprise to which the current logged-in user of the computer device belongs.

[0110] In the embodiments of the present application, the enterprise field refers to the main business field of the enterprise, and the enterprise field may be, for example, the medical field, the automotive field, the sports field, etc. Therefore, the corresponding enterprise field may be pre-assigned to the enterprise to which each enterprise user belongs.

[0111] 402. Determine a plurality of first data sets related to the enterprise field from among a plurality of data sets to be used.

[0112] In an embodiment of the present application, a corresponding data set field is pre-assigned to each data set to be used. The data set field may be, for example, medical data, automobile data, sports data, etc. Therefore, multiple first data sets related to the enterprise field can be determined by matching the enterprise field with the data set fields of multiple data sets to be used. For example, when the enterprise field of the enterprise to which the currently logged-in user belongs is the medical field, the data set field of multiple first data sets is medical data. For another example, when the enterprise field of the enterprise to which the currently logged-in user belongs is the automobile field, the data set field of multiple first data sets is automobile data.

[0113] 403. Obtain the training data file type supported by the target model to be trained.

[0114] In the embodiments of the present application, due to the differences in model structure and model algorithm between different models, different models may be applicable to different objects. For example, one model is applicable to natural language data processing, while another model is applicable to computer vision data processing. Therefore, the training data file type supported by each model to be trained can be pre-configured to obtain the training data file type supported by the target model to be trained. Among them, the training data file type can be text data, image data, table data, structured data, unstructured data, etc., which are not limited here.

[0115] 404. Determine, from among the multiple first data sets, multiple second data sets that conform to the training data file type.

[0116] In an embodiment of the present application, the file type of each first data set is obtained, and the file type may be, for example, a text file, an image file, a table file, a structured file, an unstructured file, etc. Based on the match between the file type of each first data set and the above-mentioned training data file type, multiple second data sets that conform to the training data file type can be determined. For example, when the above-mentioned training data file type is structured data, the second data set that conforms to the training data file type is also a structured file. For another example, when the above-mentioned training data file type is unstructured data, the second data set that conforms to the training data file type is also an unstructured file.

[0117] 405. Take each second data set as a candidate data set, thereby obtaining multiple candidate data sets.

[0118] It can be seen that in the above embodiments of the present application, based on the enterprise field of the enterprise to which the currently logged-in user belongs and the training data file type supported by the target model to be trained, multiple candidate data sets are screened out from multiple data sets to be used, so that users can find the required data sets more quickly, thereby improving the efficiency of building AI applications.

[0119] In some embodiments of the present application, Figure 5 As shown, in Figures 1 to 4 Based on any of the embodiments shown, the second page can be obtained by the following steps:

[0120] 501. Obtain model parameter quantities in the target model to be trained.

[0121] In the embodiments of the present application, the model parameter quantity refers to the number of model parameters in the corresponding model, for example, a model to be trained includes 50,000 model parameters, and another model to be trained includes 100,000 model parameters. Therefore, the model parameter quantity in the target model to be trained can be obtained.

[0122] 502. Determine the expected number of training data that matches the model parameter quantity.

[0123] In an embodiment of the present application, since the larger the model parameter amount, the more complex the corresponding model is, and the more training data is required to train the model, the expected number of training data that matches the model parameter amount can be determined. For example, when the model parameter amount is 50,000, the corresponding expected number of training data is 200,000, that is, the training data required to train the model is 200,000, so that a relatively accurate target AI application can be obtained.

[0124] 503. Determine the difference between the actual number of training data items and the expected number of training data items in each candidate data set.

[0125] 504. Sort the multiple candidate data sets based on the difference values ​​to obtain a fourth order.

[0126] In an embodiment of the present application, in the step of sorting multiple candidate data sets based on the difference, the multiple candidate data sets can be sorted from small to large based on the absolute value of the difference to obtain the fourth order, that is, the candidate data set whose actual number of training data is approximately close to the expected number of training data is ranked higher in the fourth order. It can be understood that the candidate data set whose actual number of training data is approximately close to the expected number of training data is more likely to be the candidate data set intended for training the target model to be trained.

[0127] In some embodiments of the present application, before sorting multiple candidate data sets from small to large based on the absolute value of the difference, it can also include: detecting whether the absolute value of the difference is greater than a preset threshold; if the absolute value of the difference is greater than the preset threshold, it indicates that the corresponding candidate data set is not a candidate data set that is intended to be used for training the target model to be trained, so the candidate data set can be deleted, thereby filtering multiple candidate data sets; then sorting the multiple candidate data sets after filtering from small to large based on the absolute value of the difference to obtain the fourth order. It can be seen that the embodiments of the present application can enable enterprise users to find the candidate data sets they want to use for model training more quickly by filtering multiple candidate data sets.

[0128] 505. Generate a second page, where the second page includes multiple candidate data sets arranged in a fourth order.

[0129] In an embodiment of the present application, a second page including multiple candidate data sets arranged in a fourth order is generated, and then the second page is displayed in a construction platform for enterprise-level AI applications, so that enterprise users can find the candidate data sets they want to use for model training more quickly in the second page.

[0130] In the second aspect, based on the method for constructing an enterprise-level AI application in the above embodiment, an embodiment of the present application provides a device for constructing an enterprise-level AI application, and the device for constructing an enterprise-level AI application is used to execute the steps in any embodiment of the method for constructing an enterprise-level AI application. Specifically, the device for constructing an enterprise-level AI application may include:

[0131] A first acquisition module is used to acquire a plurality of preset models to be trained in a computer device used to build an enterprise-level AI application;

[0132] A first determination module, used to display a first page including a plurality of models to be trained, so as to determine a target model to be trained among the plurality of models to be trained based on a first instruction on the first page;

[0133] A second determination module is used to determine a plurality of candidate data sets that match the target to-be-trained model from among a plurality of preset to-be-used data sets in the computer device;

[0134] A third determination module, configured to display a second page including a plurality of candidate data sets, and to determine a target training set among the plurality of candidate data sets based on a second instruction on the second page;

[0135] The model training module is used to train the target model to be trained using the target training set to obtain the target AI application.

[0136] In a third aspect, an embodiment of the present application provides a computer device that integrates any of the construction devices for enterprise-level AI applications provided in the embodiments of the present application. The computer device includes a processor and a memory, and a computer program is stored in the memory. The computer program is configured to be executed by the processor to implement the construction method for enterprise-level AI applications described in any of the above embodiments, for example:

[0137] Acquire multiple preset models to be trained in a computer device for building an enterprise-level AI application; display a first page including multiple models to be trained to determine a target model to be trained among the multiple models to be trained based on a first instruction on the first page; determine multiple candidate data sets that match the target model to be trained among multiple preset data sets to be used in the computer device; display a second page including multiple candidate data sets to determine a target training set among the multiple candidate data sets based on a second instruction on the second page; use the target training set to train the target model to be trained to obtain a target AI application.

[0138] In a fourth aspect, the embodiments of the present application provide a computer device that integrates any of the construction devices for enterprise-level AI applications provided in the embodiments of the present application. Figure 6 As shown, it shows a schematic diagram of the structure of the computer device involved in the embodiment of the present application, specifically:

[0139] The computer device may include one or more processing core processors 601, one or more computer readable storage medium storage units 602, a power supply 603, an input unit 604 and other components. Those skilled in the art will understand that Figure 6 The computer device structure shown in the figure does not constitute a limitation on the computer device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Among them:

[0140] The processor 601 is the control center of the computer device. It uses various interfaces and lines to connect various parts of the entire computer device. By running or executing software programs and / or modules stored in the storage unit 602 and calling data stored in the storage unit 602, it executes various functions of the computer device and processes data, thereby monitoring the computer device as a whole. Optionally, the processor 601 may include one or more processing cores; preferably, the processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 601.

[0141] The storage unit 602 can be used to store software programs and modules. The processor 601 executes various functional applications and data processing by running the software programs and modules stored in the storage unit 602. The storage unit 602 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the storage unit 602 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the storage unit 602 may also include a memory controller to provide the processor 601 with access to the storage unit 602.

[0142] The computer device also includes a power supply 603 for supplying power to various components. Preferably, the power supply 603 can be logically connected to the processor 601 through a power management system, so as to manage charging, discharging, and power consumption through the power management system. The power supply 603 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0143] The computer device may further include an input unit 604, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0144] Although not shown, the computer device may further include a display unit, etc., which will not be described in detail herein. Specifically, in the embodiment of the present application, the processor 601 in the computer device will load the executable file corresponding to the process of one or more application programs into the storage unit 602 according to the following instructions, and the processor 601 will run the application program stored in the storage unit 602, thereby realizing various functions, such as:

[0145] Acquire multiple preset models to be trained in a computer device for building an enterprise-level AI application; display a first page including multiple models to be trained to determine a target model to be trained among the multiple models to be trained based on a first instruction on the first page; determine multiple candidate data sets that match the target model to be trained among multiple preset data sets to be used in the computer device; display a second page including multiple candidate data sets to determine a target training set among the multiple candidate data sets based on a second instruction on the second page; use the target training set to train the target model to be trained to obtain a target AI application.

[0146] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, which may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc. The computer-readable storage medium stores a computer program, and the computer program is configured to be executed by a processor to implement the method for constructing an enterprise-level AI application as described in any of the above items, for example:

[0147] Acquire multiple preset models to be trained in a computer device for building an enterprise-level AI application; display a first page including multiple models to be trained to determine a target model to be trained among the multiple models to be trained based on a first instruction on the first page; determine multiple candidate data sets that match the target model to be trained among multiple preset data sets to be used in the computer device; display a second page including multiple candidate data sets to determine a target training set among the multiple candidate data sets based on a second instruction on the second page; use the target training set to train the target model to be trained to obtain a target AI application.

[0148] In a sixth aspect, an embodiment of the present application provides a computer program product or a computer program, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes to implement the method for constructing an enterprise-level AI application as described in any one of the above items, for example:

[0149] Acquire multiple preset models to be trained in a computer device for building an enterprise-level AI application; display a first page including multiple models to be trained to determine a target model to be trained among the multiple models to be trained based on a first instruction on the first page; determine multiple candidate data sets that match the target model to be trained among multiple preset data sets to be used in the computer device; display a second page including multiple candidate data sets to determine a target training set among the multiple candidate data sets based on a second instruction on the second page; use the target training set to train the target model to be trained to obtain a target AI application.

[0150] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for constructing an enterprise-level AI application, characterized in that: The method for constructing an enterprise-level AI application includes: Acquire a plurality of preset models to be trained in a computer device used to build the enterprise-level AI application; Displaying a first page including a plurality of the models to be trained, so as to determine a target model to be trained among the plurality of the models to be trained based on a first instruction on the first page; Determining, from among a plurality of preset data sets to be used in the computer device, a plurality of candidate data sets that match the target model to be trained; displaying a second page including the plurality of candidate data sets, to determine a target training set among the plurality of candidate data sets based on a second instruction on the second page; The target training set is used to train the target model to be trained to obtain a target AI application.

2. The method for constructing an enterprise-level AI application according to claim 1, wherein: The using the target training set to train the target model to be trained to obtain a target AI application includes: Acquire configuration data of the target AI application, wherein the configuration data includes at least one of a target knowledge base, a target calling tool, and a target prompt word associated with the target AI application; The target training set and the configuration data are used to train the target model to be trained to obtain the target AI application.

3. The method for constructing an enterprise-level AI application according to claim 2, wherein: The obtaining of the configuration data of the target AI application includes: Get uploaded knowledge documents; Slicing the knowledge document to obtain segmented data of the knowledge document; Performing vectorization processing on the segmented data to obtain vector features of the segmented data; The target knowledge base in the configuration data is generated based on the vector features.

4. The method for constructing an enterprise-level AI application according to claim 2, wherein: The obtaining of the configuration data of the target AI application includes: Acquire a plurality of candidate calling tools preset in the computer device, wherein the candidate calling tools include at least one of a device preset tool, a device external tool, and a workflow tool; A third page including the plurality of candidate calling tools is displayed to determine a target calling tool among the plurality of candidate calling tools.

5. The method for constructing an enterprise-level AI application according to claim 4, wherein: The workflow tool is obtained by the following steps: Acquire a set data processing flow chart, wherein the data processing flow chart includes a plurality of data processing nodes, each of which is used to process data input by a previous data processing node of the data processing node; The data processing flow chart is published in the computer device to obtain the workflow tool.

6. The method for constructing an enterprise-level AI application according to claim 1, wherein: The first page is obtained by the following steps: Determine a plurality of first models and a plurality of second models among the plurality of models to be trained based on a model type of each of the models to be trained, wherein the model type includes at least one of a locally uploaded model and a device preset model; Obtaining the model result accuracy and the model historical usage times of each of the first models; Determining a model recommendation score for each of the first models based on the accuracy of the model results and the number of times the model has been used in history; Sorting the plurality of first models based on the model recommendation score of each of the first models to obtain a first order; sorting the plurality of said second models to obtain a second order; The second sequence and the first sequence are concatenated to obtain a third sequence; The first page is generated, wherein the first page includes a plurality of the models to be trained arranged in the third order.

7. The method for constructing an enterprise-level AI application according to claim 1, wherein: Determining a plurality of candidate data sets that match the target to-be-trained model from among the plurality of data sets to be used preset in the computer device comprises: Obtaining the enterprise domain of the enterprise to which the currently logged-in user of the computer device belongs; Determine a plurality of first data sets related to the enterprise field among the plurality of data sets to be used; Obtaining the training data file type supported by the target model to be trained; Determining, from among the plurality of the first data sets, a plurality of second data sets that conform to the type of the training data file; Each of the second data sets is used as a candidate data set, thereby obtaining a plurality of candidate data sets.

8. The method for constructing an enterprise-level AI application according to claim 1, wherein: The second page is obtained by following the steps: Obtaining model parameter quantities in the target model to be trained; Determining the expected number of training data that matches the model parameter quantity; Determine the difference between the actual number of training data items in each of the candidate data sets and the expected number of training data items; Based on the difference, sorting the plurality of candidate data sets to obtain a fourth order; The second page is generated, wherein the second page includes the plurality of candidate data sets arranged in the fourth order.

9. A computer device, characterized in that: The computer device includes a processor and a memory, wherein a computer program is stored in the memory, and the computer program is configured to be executed by the processor to implement the method for constructing an enterprise-level AI application according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is configured to be executed by a processor to implement the method for constructing an enterprise-level AI application according to any one of claims 1 to 8.

Images