Model evaluation method and device, computer equipment, storage medium and program product

Abstract

The invention relates to a model evaluation method and device, computer equipment, a storage medium and a program product. The method comprises the steps of testing a target model indicated by an evaluation control according to a test data set in response to a trigger operation on the evaluation control in a model evaluation interface to obtain a test result, displaying an annotation interface in the model evaluation interface, and then obtaining an annotation result in response to an annotation operation on the annotation interface, and performing training effect evaluation on the target model according to the test result and the labeling result to obtain an evaluation result. The method for evaluating the to-be-evaluated model based on the man-machine interaction mode is provided, efficient evaluation of the model can be achieved through two responses, and the evaluation efficiency and the evaluation flexibility of the model are greatly improved.

Description

Technical Field

[0001] The present application relates to the field of large model technology, and in particular to a model evaluation method, apparatus, computer equipment, storage medium and program product. Background Art

[0002] With the rapid development of artificial intelligence models, more and more intelligent models are being developed and applied in various fields, such as image recognition models, target detection models, target classification models, etc. Among them, in the development process of intelligent models, it is usually necessary to train the initial architecture model to realize the application of corresponding functions. In order to improve the quality and ability of the model, it is very meaningful to evaluate the effect of the trained intelligent model during the model training process.

[0003] In the related art, the method for evaluating the effect of the trained model is usually to test the trained model using a test data set and evaluate the model training effect according to the test results. However, the above method has the problem of low evaluation efficiency. Summary of the invention

[0004] Based on this, it is necessary to provide a model evaluation method, device, computer equipment, storage medium and program product that can improve the efficiency of model evaluation in response to the above technical problems.

[0005] In a first aspect, the present application provides a model evaluation method, comprising:

[0006] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0007] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0008] The above method provides a method for evaluating a model to be evaluated based on a human-computer interaction mode. Through two responses, efficient evaluation of the model can be achieved, that is, first responding to the evaluation control triggered by the user on the model evaluation interface, determining the target model from multiple models to be evaluated on the model evaluation interface, and testing the target model according to the test data set corresponding to the target model to obtain the test result, and then responding to the user's annotation operation on the annotation interface to evaluate the training effect of the target model based on the test result and the annotation result. Compared with the existing method of evaluating based on the test result after model training, the above method only requires the user to trigger the relevant information of the model to be evaluated on the model evaluation interface to realize the evaluation of the model, which greatly improves the evaluation efficiency and evaluation flexibility of the model.

[0009] In one embodiment, the method for obtaining the target model includes:

[0010] In response to a triggering operation on a training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to the training data set, and the target model and training process information are obtained and saved;

[0011] Alternatively, in response to a selection operation on a first target option in a model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as a target model.

[0012] The embodiments of the present application provide two methods for acquiring target models, both of which achieve rapid response to user instructions through human-computer interaction, greatly improve the efficiency of acquiring the target model, and further improve the evaluation efficiency of subsequent models.

[0013] In one embodiment, the method for obtaining the above test data set includes:

[0014] Search the dataset corresponding to the type of the target model from the dataset library to obtain the test dataset;

[0015] Alternatively, in response to a selection operation on a second target option in a data set import area in the model evaluation interface, the data set selected by the selection operation is used as a test data set.

[0016] The embodiments of the present application provide two methods for acquiring test data sets, both of which achieve rapid response to user instructions through human-computer interaction, greatly improve the efficiency of acquiring test data sets, and further improve the evaluation efficiency of subsequent models.

[0017] In one embodiment, the training effect of the target model is evaluated based on the test results and the annotation results to obtain the evaluation results, including:

[0018] Compare the test results and the annotation results to obtain a comparison chart;

[0019] The training effect of the target model is evaluated according to the comparison chart, and the evaluation results and comparison chart are obtained and displayed in the model evaluation interface.

[0020] The method for evaluating the model training effect provided in the embodiments of the present application evaluates the training effect of the model based on the comparison results between the test results and the annotation results. Compared with the method of evaluating the model training effect through test data, the above method does not require more test data to evaluate the model to be evaluated, which greatly improves the evaluation efficiency of the model.

[0021] In one embodiment, the method further comprises:

[0022] If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors why the training effect does not meet the expected requirements.

[0023] The model evaluation method provided in the embodiment of the present application provides a method for analyzing the training process information to determine the factors why the training effect does not meet the preset requirements when the model training effect does not meet the expectations, thereby providing a certain training basis for subsequent re-training of the model.

[0024] In one embodiment, the method further comprises:

[0025] Get the training process information of the target model;

[0026] Based on the training process information of the target model, the target model and the corresponding evaluation results, an evaluation report is generated and displayed in the model evaluation interface.

[0027] The model evaluation method provided in the embodiment of the present application can intuitively display the model evaluation results and the model training process on the model evaluation interface, so that the user can intuitively obtain the model evaluation results and the model training process information, thereby providing a certain basis for the subsequent re-evaluation of the model.

[0028] In one embodiment, the method further comprises:

[0029] In response to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of the generation control, the following steps are performed:

[0030] Obtain an initial natural language model;

[0031] Fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model;

[0032] A model processing interface corresponding to the selection operation is generated using the fine-tuned natural language model, and the model processing interface corresponding to the selection operation is switched from the function display interface; the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0033] The model evaluation method provided in the embodiment of the present application generates a model processing interface including a model evaluation interface by fine-tuning the natural language model, thereby realizing the function of automatically processing the model and improving the processing efficiency of the model.

[0034] In a second aspect, the present application also provides a model evaluation device, comprising:

[0035] The first response module is used to respond to the triggering operation of the evaluation control in the model evaluation interface, test the target model indicated by the evaluation control according to the test data set, obtain the test result, and display the annotation interface in the model evaluation interface;

[0036] The second response module is used to respond to the annotation operation on the annotation interface, obtain the annotation result, and evaluate the training effect of the target model according to the test result and the annotation result to obtain the evaluation result.

[0037] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0038] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0039] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0040] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0041] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0042] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0043] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0044] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0045] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0046] The above-mentioned model evaluation method, device, computer equipment, storage medium and program product, the above-mentioned method provides a method for evaluating the model to be evaluated based on the human-computer interaction mode, and the efficient evaluation of the model can be achieved through two responses, that is, first responding to the evaluation control triggered by the user on the model evaluation interface, determining the target model from the multiple models to be evaluated on the model evaluation interface, and testing the target model according to the test data set corresponding to the target model to obtain the test results, and then responding to the user's annotation operation on the annotation interface to evaluate the training effect of the target model based on the test results and the annotation results. Compared with the existing method of evaluating based on the test results after model training, the above-mentioned method only requires the user to trigger the relevant information of the model to be evaluated on the model evaluation interface to realize the evaluation of the model, which greatly improves the evaluation efficiency and evaluation flexibility of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions 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 related drawings can be obtained based on these drawings without paying creative work.

[0048] FIG1 is a diagram of an application environment of a model evaluation method in one embodiment;

[0049] Figure 2 A schematic diagram of a flow chart of a model evaluation method in one embodiment;

[0050] Figure 3 A schematic diagram of a model evaluation interface in one embodiment;

[0051] Figure 4 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0052] Figure 5 A schematic diagram showing a comparison between an evaluation result and an annotation result in one embodiment;

[0053] Figure 6 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0054] Figure 7 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0055] Figure 8 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0056] Fig. 9 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0057] Fig.10 A schematic diagram of a flow chart of a model evaluation method in another embodiment;

[0058] Fig.11 is a structural block diagram of a model evaluation device in one embodiment;

[0059] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0061] With the rapid development of artificial intelligence models, more and more intelligent models are being developed and applied in various fields, such as image recognition models, target detection models, target classification models, etc. Among them, in the development process of intelligent models, it is usually necessary to train the initial architecture model to realize the application of corresponding functions. In order to improve the quality and ability of the model, it is very meaningful to evaluate the effect of the trained intelligent model during the model training process.

[0062] In the related art, the method for evaluating the effect of a trained model is usually to test the trained model using a test data set and evaluate the model training effect based on the test results. However, the above method has the problem of low evaluation efficiency. This application aims to solve this problem.

[0063] After the above introduction to the background technology of the model evaluation method provided by the embodiment of the present application, the implementation environment involved in the model evaluation method provided by the embodiment of the present application will be briefly described below. The model evaluation method provided by the embodiment of the present application can be applied in the implementation environment as shown in Figure 1. The implementation environment includes a server 104, and the server 104 can be implemented with an independent server 104 or a server cluster composed of multiple servers 104, and the data storage system 102 can store the data that the server 104 needs to process. The data storage system 102 can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the data storage system 102 can collect and pre-store various test data sets, and the server 104 can obtain various test data sets from the data storage system 102, and test the model corresponding to the test data set according to the test data set to obtain the test result, and evaluate the target model according to the annotation result and the test result to obtain the evaluation result.

[0064] In other possible implementations, the model evaluation method provided in the embodiment of the present application can also be applied to terminals, which can be, but are not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc.

[0065] After introducing the application scenarios of the model evaluation method provided in the embodiments of the present application, the following focuses on introducing the model evaluation method described in the present application.

[0066] In one embodiment, Figure 2 As shown, a model evaluation method is provided, which is described by taking the method applied to the server in FIG1 as an example, and includes the following steps:

[0067] S201. In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a labeling interface is displayed in the model evaluation interface.

[0068] The model evaluation interface refers to the display interface of the model evaluation software, which includes evaluation controls, training controls, and acquisition controls, as well as annotation interfaces, evaluation result display interfaces, and test result display interfaces. Figure 3 As shown, 01 represents the model evaluation interface, 10 represents the model display area of ​​the model evaluation interface, 20 represents the data set import area of ​​the model evaluation interface, 30 represents the display control, 40 represents the annotation control, 50 represents the training control and 60 represents the evaluation control, 101~10n represent the model 1 to be evaluated to be evaluated to be n, and 201~20n represent the test data set 1 to test data set n, respectively.

[0069] Among them, the evaluation control is used to instruct the model evaluation software to evaluate the corresponding model to be evaluated based on the target model indicated by the evaluation control and the test data set corresponding to the target model. The model evaluation interface may include multiple evaluation controls, each evaluation control corresponds to a different model to be evaluated. After the user triggers any evaluation control in the model evaluation interface, the target model corresponding to the evaluation control is selected, and the target model can be tested based on the test data set corresponding to the target model in the subsequent process to obtain the test results.

[0070] Among them, the training control is used to instruct the model evaluation software to train the model to be trained indicated by the training control; the acquisition control is used to obtain a test data set matching the type of the model to be evaluated from the database based on the type of the model to be evaluated.

[0071] Among them, the annotation interface is used to instruct the user to annotate the data type and other information at the corresponding position of the data displayed on the annotation interface; the evaluation result display interface is used to display the evaluation results of the model to be evaluated; the test result display interface is used to display the test results corresponding to the model to be evaluated, and the test results corresponding to the model to be evaluated are obtained by testing the target model indicated by the evaluation control based on the test data set.

[0072] Among them, the annotation interface includes a test data display column in the test data set and a type annotation column corresponding to the test data. For example, when the user displays the test data on the annotation interface, the user can annotate the type information corresponding to the test data in the type annotation column corresponding to the test data.

[0073] In the embodiment of the present application, when it is necessary to evaluate the performance of the model after training (for example, prediction performance, classification performance, and recognition performance, etc.), it can be implemented based on the model evaluation software in the embodiment of the present application. The user can first start the model evaluation software, and trigger the evaluation control corresponding to the model to be evaluated in the model evaluation interface of the model evaluation software, so that the server responds to the user's triggering operation of the corresponding evaluation control of the model to be evaluated in the model evaluation interface, and at the same time, obtains the test data set corresponding to the model to be evaluated, and tests the target model indicated by the evaluation control according to the test data set corresponding to the model to be evaluated, so as to obtain the test results, and displays the annotation interface on the test result display interface of the model evaluation interface and on the model evaluation interface.

[0074] S202. In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0075] The marking operation may be relevant information of the test data input by the user on the marking interface. For example, the marking operation may be inputting information such as the data type corresponding to the test data at a corresponding position of the test data on the marking interface.

[0076] In an embodiment of the present application, after the annotation interface is displayed in the above-mentioned model evaluation interface, the user can manually annotate the data type and other information of the corresponding test data based on the test data displayed in the annotation interface, so that the server responds to the user's annotation operation on the annotation interface, and obtains the annotation result corresponding to the annotation operation, and compares the annotation result with the test result to obtain a comparison result, so as to evaluate the training effect of the target model based on the comparison result, thereby obtaining an evaluation result.

[0077] The model evaluation method provided in the embodiment of the present application first responds to the triggering operation of the evaluation control in the model evaluation interface, tests the target model indicated by the evaluation control according to the test data set, obtains the test result, and displays the annotation interface in the model evaluation interface, then responds to the annotation operation on the annotation interface, obtains the annotation result, and evaluates the training effect of the target model according to the test result and the annotation result to obtain the evaluation result. The above method provides a method for evaluating the model to be evaluated based on the human-computer interaction mode. Through two responses, the efficient evaluation of the model can be achieved, that is, first responding to the evaluation control triggered by the user on the model evaluation interface, determining the target model from multiple models to be evaluated on the model evaluation interface, and testing the target model according to the test data set corresponding to the target model to obtain the test result, and then responding to the user's annotation operation on the annotation interface to evaluate the training effect of the target model based on the test result and the annotation result. Compared with the existing method of evaluating based on the test result after model training, the above method only requires the user to trigger the relevant information of the model to be evaluated on the model evaluation interface to achieve the evaluation of the model, which greatly improves the evaluation efficiency and evaluation flexibility of the model.

[0078] In one embodiment, Figure 2 Based on the illustrated embodiment, the process of acquiring the target model can be described. The above target model acquisition method includes the following two methods:

[0079] First, in response to the triggering operation of the training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to the training data set, and the target model and training process information are obtained and saved.

[0080] Among them, the model evaluation interface also includes a training control, and the training control is used to instruct the model evaluation software to train the model to be trained indicated by the training control.

[0081] Optionally, the model evaluation interface may include multiple training controls, each training control corresponds to a different model to be trained, and the user can trigger the corresponding training control to enable the server to train the model to be trained corresponding to the training control.

[0082] Optionally, the model evaluation interface may also include multiple identifiers of models to be trained and a training control, where each identifier of the model to be trained corresponds to a different model to be trained. When it is necessary to train the target model to be trained, the user can first trigger the identifier corresponding to the target model to be trained, and then trigger the training control to enable the server to train the target model to be trained.

[0083] Among them, the training data set is the data set corresponding to the model to be trained. Different models to be trained correspond to different training data sets. The training data set can be pre-stored in the database. When the user triggers the training control in the model evaluation interface, the server can call the training data set that matches the model to be trained corresponding to the training control from the training data set.

[0084] The training process information refers to the information generated during the model training process, including parameter settings during the training process, training data distribution, trained model name, model training time and other information.

[0085] In an embodiment of the present application, when it is necessary to obtain the target model, the user can trigger the training control in the model evaluation interface to make the server respond to the model to be trained corresponding to the training control, and obtain the training data set corresponding to the model to be trained from the database, and then train the model to be trained according to the training data set, and then obtain the target model after the model to be trained and the training process information during the model training process, and save the completed target model and training process information.

[0086] Second, in response to a selection operation on the first target option in the model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as the target model.

[0087] The model display area is used to display multiple model options to be evaluated, each model option to be evaluated corresponds to a different model to be evaluated. The first target option is any one of the multiple model options to be evaluated.

[0088] In the embodiment of the present application, when the target model needs to be acquired, the user can trigger any model option to be evaluated (ie, the first target option) in the model display area, and use the model to be evaluated corresponding to the first target option as the target model.

[0089] It should be noted that the models to be evaluated corresponding to the models to be evaluated options displayed in the model display area are all trained models. Optionally, the model training method described above can be used to complete the model training, obtain the trained model, and store the trained model in the model database, so that the trained model can be directly obtained from the model database later.

[0090] The embodiments of the present application provide two methods for acquiring target models, both of which achieve rapid response to user instructions through human-computer interaction, greatly improve the efficiency of acquiring the target model, and further improve the evaluation efficiency of subsequent models.

[0091] In one embodiment, Figure 2Based on the illustrated embodiment, the process of acquiring the test data set can be described. The method for acquiring the test data set includes:

[0092] First, search the dataset corresponding to the type of the target model from the dataset library to obtain the test dataset.

[0093] The data set library pre-stores test data sets corresponding to multiple models to be evaluated. Optionally, test data sets corresponding to each model to be evaluated can be collected in advance, and the multiple test data sets can be stored in the data set library according to the type of the test data sets.

[0094] In an embodiment of the present application, when it is necessary to evaluate the target model, it is necessary to obtain a test data set corresponding to the target model. After the above-mentioned response to the trigger operation of the evaluation control, the type of the corresponding target model can be determined according to the evaluation control, and then the data set corresponding to the type of the target model can be searched from the data set library according to the type of the target model, thereby obtaining the test data set.

[0095] Second, in response to a selection operation on a second target option in a data set import area in a model evaluation interface, the data set selected by the selection operation is used as a test data set.

[0096] The model evaluation interface may also display multiple data set options, each corresponding to a different type of test data set. The second target option is any one of the multiple data set options.

[0097] In the embodiment of the present application, when it is necessary to obtain a test data set, the user can trigger any data set option (ie, the second target option) in the model display area and use the data set corresponding to the second target option as the test data set.

[0098] The embodiments of the present application provide two methods for acquiring test data sets, both of which achieve rapid response to user instructions through human-computer interaction, greatly improve the efficiency of acquiring test data sets, and further improve the evaluation efficiency of subsequent models.

[0099] In one embodiment, Figure 2 Based on the embodiment shown, the process of obtaining the evaluation result can be described as follows: Figure 4 As shown, the above S202 "evaluating the training effect of the target model according to the test results and the annotation results to obtain the evaluation results" includes:

[0100] S301, compare the test results and the marked results to obtain a comparison chart.

[0101] The comparison chart represents the similarity between the test result and the annotation result. For example, the comparison chart can be a similarity set between the test result and the annotation result, or a line chart consisting of the difference between the test result and the standard value with the annotation result as the standard value (see Figure 5 ), through this line chart, we can intuitively see the similarity between the test results and the annotation results.

[0102] See also Figure 5 , the annotation results and test results of 6 data were collected respectively, and the annotation results of these 6 data were set to the standard value 1 (see the black solid points in the figure). Then the test result is the similarity value between the test result and the annotation result (see the black hollow points in the figure).

[0103] In an embodiment of the present application, after obtaining the test results obtained by testing the target model and the annotation results obtained by annotation, the test results and the annotation results can be compared to obtain the similarity between the test results and the annotation results, and a comparison chart can be generated based on the similarity between the test results and the annotation results.

[0104] S302: Evaluate the training effect of the target model according to the comparison chart, and obtain and display the evaluation results and comparison chart in the model evaluation interface.

[0105] In the embodiment of the present application, after the comparison chart is obtained as described above, the training effect of the target model can be evaluated according to the comparison chart to obtain the evaluation effect, and the evaluation result and the comparison chart can be displayed in the model evaluation interface.

[0106] Optionally, the training effect of the target model can be evaluated based on the similarity between the test results and the annotation results in the comparison chart to obtain an evaluation effect. For example, if the similarity between the test results and the annotation results in the comparison chart is greater than a preset similarity threshold, it can be considered that the training effect of the target model is good. For another example, if the similarity between the test results and the annotation results in the comparison chart is not greater than the preset similarity threshold, it can be considered that the training effect of the target model is poor. It should be noted that the embodiment of the present application does not limit the way in which the training effect of the model is evaluated.

[0107] Optionally, after obtaining the evaluation results of the target model as described above, the evaluation results and comparison charts may be displayed in the model evaluation interface.

[0108] The method for evaluating the model training effect provided in the embodiments of the present application evaluates the training effect of the model based on the comparison results between the test results and the annotation results. Compared with the method of evaluating the model training effect through test data, the above method does not require more test data to evaluate the model to be evaluated, which greatly improves the evaluation efficiency of the model.

[0109] In one embodiment, Figure 2 Based on the embodiment shown, Figure 6 As shown, the above method also includes:

[0110] S203. If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors that cause the training effect to not meet the expected requirements.

[0111] The expected requirement may be a preset similarity threshold.

[0112] The training process information refers to the information generated during the model training process, including parameter settings during the training process, training data distribution, trained model name, model training time and other information.

[0113] In an embodiment of the present application, after the evaluation result is determined as above, if the evaluation result indicates that the training effect of the target model does not meet the preset similarity threshold, the training process information corresponding to the target model can be obtained from the server, and the training process information can be statistically analyzed to obtain the factors why the training effect does not meet the expected requirements.

[0114] Optionally, when it is necessary to obtain the training process information of the target model, the user can trigger the training control in the model evaluation interface to make the server respond to the model to be trained corresponding to the training control, and obtain the training data set corresponding to the model to be trained from the database, and then train the model to be trained according to the training data set, so as to obtain the target model after training the model to be trained and the training process information during the model training process, and save the trained target model and training process information so that the training process information of the model can be directly obtained when the training process information of the model is needed later.

[0115] Optionally, a statistical analysis may be performed on the parameter settings, training data distribution, trained model name, model training time and other information in the training process information to obtain factors that cause the training effect to fail to meet the expected requirements. For example, a statistical analysis may be performed on the parameter settings, and by comparing the fluctuations of the parameter settings, when the fluctuations of the parameter settings are greater than a preset fluctuation threshold, it is determined that the factor that causes the training effect to fail to meet the expected requirements is that the parameter settings are unreasonable. For another example, a statistical analysis may be performed on the distribution of training data, and by comparing the distribution of the training data, when the distribution of the training data is greater than a preset distribution threshold, it is determined that the factor that causes the training effect to fail to meet the expected requirements is that the distribution of the training data is unreasonable. For another example, a statistical analysis may be performed on the training time of the model, and by comparing the training time of the model, when the training time is less than a preset training time threshold, it is determined that the factor that causes the training effect to fail to meet the expected requirements is that the training time of the model is short.

[0116] The model evaluation method provided in the embodiment of the present application provides a method for analyzing the training process information to determine the factors why the training effect does not meet the preset requirements when the model training effect does not meet the expectations, thereby providing a certain training basis for subsequent re-training of the model.

[0117] In one embodiment, Figure 2 Based on the embodiment shown, Figure 7 As shown, the above method also includes:

[0118] S204: Obtain training process information of the target model.

[0119] The training process information refers to the information generated during the model training process, including parameter settings during the training process, training data distribution, trained model name, model training time and other information.

[0120] Optionally, when it is necessary to obtain the training process information of the target model, the user can trigger the training control in the model evaluation interface to make the server respond to the model to be trained corresponding to the training control, and obtain the training data set corresponding to the model to be trained from the database, and then train the model to be trained according to the training data set, so as to obtain the target model after training the model to be trained and the training process information during the model training process, and according to the type of model, save the trained target model and training process information to the information storage database, so that the training process information of the model can be directly obtained when the training process information of the model is needed later.

[0121] In an embodiment of the present application, when it is necessary to obtain the training process information of the target model, the training process information of the target model can be searched and obtained in the information storage database according to the type of the target model.

[0122] S205. Generate an evaluation report based on the training process information of the target model, the target model and the corresponding evaluation results, and display it in the model evaluation interface.

[0123] The evaluation report includes the training process information of each model and the corresponding evaluation results of each model. The training process information of a model and its corresponding evaluation results can generate the evaluation report corresponding to the model, and each model corresponds to its own evaluation report.

[0124] In an embodiment of the present application, after the training process information of the target model, the target model and the evaluation results corresponding to the target model are obtained as mentioned above, an evaluation report of the target model can be generated according to the training process information corresponding to the target model and the evaluation results corresponding to the target model, and the evaluation report of the target model can be displayed on the model evaluation interface.

[0125] The model evaluation method provided in the embodiment of the present application can intuitively display the model evaluation results and the model training process on the model evaluation interface, so that the user can intuitively obtain the model evaluation results and the model training process information, thereby providing a certain basis for the subsequent re-evaluation of the model.

[0126] In one embodiment, Figure 2 Based on the embodiment shown, Figure 8 As shown, the above method also includes:

[0127] S206: Responding to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of generating a control.

[0128] Among them, the function display interface displays multiple function options for model processing applications, for example, the functions include training function, evaluation function, adaptation function, etc. Each function corresponds to a different function option, and the function corresponding to the third target option is the adaptation function, which is used to adapt the initial natural language model to other models, so that other models can apply the automatic function of the initial natural language model to process the business of the local model.

[0129] Among them, the generation control is used to generate a corresponding model processing interface according to the fine-tuned natural language model.

[0130] In an embodiment of the present application, the user can first start the model processing application. At this time, the server can display multiple function selection items on the function display interface of the model processing application, so that the user can select a third target option from the multiple function selection items according to his or her own needs, and trigger the selection operation of the third target option to display the adaptation interface corresponding to the third target option on the function display interface, so that the user can trigger the generation of controls on the adaptation interface to switch the display of the model processing interface corresponding to the selection operation from the function display interface.

[0131] See also Fig. 9 After the server executes the above step S207, the following steps are executed:

[0132] S401: Obtain an initial natural language model.

[0133] Among them, the initial natural language model is the Large Language Model (LLM).

[0134] In an embodiment of the present application, the server may first obtain an initial natural language model.

[0135] S402: fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model.

[0136] In the embodiment of the present application, after acquiring the initial natural language model as described above, the server may fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model.

[0137] Optionally, you can also first obtain model training data according to the model application function of the option corresponding to the selection operation, so as to train the initial natural language model according to the model training data to obtain a trained natural language model; it should be noted that the trained natural language model has the same function as the model application function.

[0138] S403: Generate a model processing interface corresponding to the selection operation using the fine-tuned natural language model, and switch the display of the model processing interface corresponding to the selection operation from the function display interface.

[0139] Among them, the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0140] In an embodiment of the present application, after the fine-tuned natural language model is generated as described above, the server can use the fine-tuned natural language model to generate a model processing interface corresponding to the selection operation, and switch the display of the model processing interface corresponding to the selection operation from the function display interface, so that the user can select the relevant controls of the model evaluation interface from the model processing interface, and the user can switch the interface to the model evaluation interface by triggering the relevant controls of the model evaluation interface.

[0141] The model evaluation method provided in the embodiment of the present application generates a model processing interface including a model evaluation interface by fine-tuning the natural language model, thereby realizing the function of automatically processing the model and improving the processing efficiency of the model.

[0142] In one embodiment, Fig.10 As shown, a model evaluation method is also provided, including:

[0143] S10, obtaining an initial natural language model;

[0144] S11, fine-tuning the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model;

[0145] S12, using the fine-tuned natural language model to generate a model processing interface corresponding to the selection operation, and switching the display of the model processing interface corresponding to the selection operation from the function display interface;

[0146] S13, in response to the triggering operation of the evaluation control in the model evaluation interface, testing the target model indicated by the evaluation control according to the test data set, obtaining the test result, and displaying the annotation interface in the model evaluation interface;

[0147] S14, in response to the annotation operation on the annotation interface, obtaining the annotation result, and comparing the test result with the annotation result to obtain a comparison chart;

[0148] S15. Evaluate the training effect of the target model according to the comparison chart, and obtain and display the evaluation results and comparison chart in the model evaluation interface.

[0149] The above method provides a method for evaluating a model to be evaluated based on a human-computer interaction mode. Through two responses, efficient evaluation of the model can be achieved, that is, first responding to the evaluation control triggered by the user on the model evaluation interface, determining the target model from multiple models to be evaluated on the model evaluation interface, and testing the target model according to the test data set corresponding to the target model to obtain the test result, and then responding to the user's annotation operation on the annotation interface to evaluate the training effect of the target model based on the test result and the annotation result. Compared with the existing method of evaluating based on the test result after model training, the above method only requires the user to trigger the relevant information of the model to be evaluated on the model evaluation interface to realize the evaluation of the model, which greatly improves the evaluation efficiency and evaluation flexibility of the model.

[0150] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0151] Based on the same inventive concept, the embodiment of the present application also provides a model evaluation device for implementing the model evaluation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more model evaluation device embodiments provided below can refer to the limitations of the model evaluation method above, and will not be repeated here.

[0152] In an exemplary embodiment, Fig.11As shown, a model evaluation device is provided, comprising: a first response module 10 and a second response module 11, wherein:

[0153] The first response module 10 is used to respond to the triggering operation of the evaluation control in the model evaluation interface, test the target model indicated by the evaluation control according to the test data set, obtain the test result, and display the annotation interface in the model evaluation interface.

[0154] The second response module 11 is used to respond to the annotation operation on the annotation interface, obtain the annotation result, and evaluate the training effect of the target model according to the test result and the annotation result to obtain the evaluation result.

[0155] In an exemplary embodiment, the first response module 10 includes a first response unit, or a second response unit, wherein:

[0156] A first response unit, specifically used to respond to a triggering operation on a training control in the model evaluation interface, train the model to be trained indicated by the training control according to the training data set, and obtain and save the target model and training process information;

[0157] The second response unit is specifically configured to respond to a selection operation on a first target option in a model display area in the model evaluation interface, and use the model to be evaluated selected by the selection operation as a target model.

[0158] In an exemplary embodiment, the first response module 10 further includes an acquisition unit, or a third response unit, wherein:

[0159] An acquisition unit, specifically used to search a data set corresponding to the type of the target model from a data set library to obtain a test data set;

[0160] The third response unit is specifically used to respond to the selection operation of the second target option in the data set import area in the model evaluation interface, and use the data set selected by the selection operation as the test data set.

[0161] In an exemplary embodiment, the second response module 11 includes a comparison unit and an evaluation unit, wherein:

[0162] A comparison unit, specifically used to compare the test result and the marked result to obtain a comparison chart;

[0163] The evaluation unit is specifically used to evaluate the training effect of the target model according to the comparison chart, and obtain and display the evaluation results and comparison chart in the model evaluation interface.

[0164] In an exemplary embodiment, the above-mentioned device also includes: a first acquisition module, which is used to obtain the training process information of the target model when the evaluation result indicates that the training effect of the target model does not meet the expected requirements, and perform statistical analysis on the training process information to obtain the factors why the training effect does not meet the expected requirements.

[0165] In an exemplary embodiment, the above device further includes: a second acquisition module and a generation module, wherein:

[0166] The second acquisition module is used to obtain the training process information of the target model;

[0167] The generation module is used to generate and display the evaluation report in the model evaluation interface based on the training process information of the target model, the target model and the corresponding evaluation results.

[0168] In an exemplary embodiment, the apparatus further includes: a third response module, which is used to respond to the selection operation of the third target option in the function display interface of the model processing application and the trigger operation of the generation control, including: a third acquisition unit, a fine-tuning unit and a switching unit, wherein:

[0169] A third acquisition unit, specifically used to acquire an initial natural language model;

[0170] A fine-tuning unit, specifically used to fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation, to obtain a fine-tuned natural language model;

[0171] The switching unit is specifically used to generate a model processing interface corresponding to the selection operation using a fine-tuned natural language model, and switch the display of the model processing interface corresponding to the selection operation from the function display interface; the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0172] Each module in the above-mentioned model evaluation device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0173] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.12As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a model evaluation method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0174] Those skilled in the art will understand that Fig.12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0175] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0176] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0177] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0178] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0179] In response to a triggering operation on a training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to the training data set, and the target model and training process information are obtained and saved;

[0180] Alternatively, in response to a selection operation on a first target option in a model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as a target model.

[0181] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0182] Search the dataset corresponding to the type of the target model from the dataset library to obtain the test dataset;

[0183] Alternatively, in response to a selection operation on a second target option in a data set import area in the model evaluation interface, the data set selected by the selection operation is used as a test data set.

[0184] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0185] Compare the test results and the annotation results to obtain a comparison chart;

[0186] The training effect of the target model is evaluated according to the comparison chart, and the evaluation results and comparison chart are obtained and displayed in the model evaluation interface.

[0187] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0188] If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors why the training effect does not meet the expected requirements.

[0189] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0190] Get the training process information of the target model;

[0191] Based on the training process information of the target model, the target model and the corresponding evaluation results, an evaluation report is generated and displayed in the model evaluation interface.

[0192] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0193] In response to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of the generation control, the following steps are performed:

[0194] Obtain an initial natural language model;

[0195] Fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model;

[0196] A model processing interface corresponding to the selection operation is generated using the fine-tuned natural language model, and the model processing interface corresponding to the selection operation is switched from the function display interface; the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0197] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0198] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0199] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0200] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0201] In response to a triggering operation on a training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to the training data set, and the target model and training process information are obtained and saved;

[0202] Alternatively, in response to a selection operation on a first target option in a model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as a target model.

[0203] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0204] Search the dataset corresponding to the type of the target model from the dataset library to obtain the test dataset;

[0205] Alternatively, in response to a selection operation on a second target option in a data set import area in the model evaluation interface, the data set selected by the selection operation is used as a test data set.

[0206] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0207] Compare the test results and the annotation results to obtain a comparison chart;

[0208] The training effect of the target model is evaluated according to the comparison chart, and the evaluation results and comparison chart are obtained and displayed in the model evaluation interface.

[0209] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0210] If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors why the training effect does not meet the expected requirements.

[0211] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0212] Get the training process information of the target model;

[0213] Based on the training process information of the target model, the target model and the corresponding evaluation results, an evaluation report is generated and displayed in the model evaluation interface.

[0214] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0215] In response to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of the generation control, the following steps are performed:

[0216] Obtain an initial natural language model;

[0217] Fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model;

[0218] A model processing interface corresponding to the selection operation is generated using the fine-tuned natural language model, and the model processing interface corresponding to the selection operation is switched from the function display interface; the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0219] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0220] In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set, a test result is obtained, and a marking interface is displayed in the model evaluation interface;

[0221] In response to the annotation operation on the annotation interface, the annotation result is obtained, and the training effect of the target model is evaluated according to the test result and the annotation result to obtain the evaluation result.

[0222] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0223] In response to a triggering operation on a training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to the training data set, and the target model and training process information are obtained and saved;

[0224] Alternatively, in response to a selection operation on a first target option in a model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as a target model.

[0225] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0226] Search the dataset corresponding to the type of the target model from the dataset library to obtain the test dataset;

[0227] Alternatively, in response to a selection operation on a second target option in a data set import area in the model evaluation interface, the data set selected by the selection operation is used as a test data set.

[0228] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0229] Compare the test results and the annotation results to obtain a comparison chart;

[0230] The training effect of the target model is evaluated according to the comparison chart, and the evaluation results and comparison chart are obtained and displayed in the model evaluation interface.

[0231] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0232] If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors why the training effect does not meet the expected requirements.

[0233] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0234] Get the training process information of the target model;

[0235] Based on the training process information of the target model, the target model and the corresponding evaluation results, an evaluation report is generated and displayed in the model evaluation interface.

[0236] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0237] In response to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of the generation control, the following steps are performed:

[0238] Obtain an initial natural language model;

[0239] Fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model;

[0240] A model processing interface corresponding to the selection operation is generated using the fine-tuned natural language model, and the model processing interface corresponding to the selection operation is switched from the function display interface; the function display interface includes options and generation controls corresponding to multiple model application functions; the model processing interface includes a model evaluation interface.

[0241] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0242] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0243] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A model evaluation method, characterized in that: The method comprises: In response to a triggering operation on an evaluation control in a model evaluation interface, a target model indicated by the evaluation control is tested according to a test data set to obtain a test result, and a marking interface is displayed in the model evaluation interface; In response to a labeling operation on the labeling interface, a labeling result is obtained, and a training effect evaluation is performed on the target model according to the test result and the labeling result to obtain an evaluation result.

2. The method according to claim 1, characterized in that The method for obtaining the target model includes: In response to a triggering operation on a training control in the model evaluation interface, the model to be trained indicated by the training control is trained according to a training data set, and the target model and training process information are obtained and saved; Alternatively, in response to a selection operation on a first target option in a model display area in the model evaluation interface, the model to be evaluated selected by the selection operation is used as the target model.

3. The method according to claim 1, characterized in that The method for obtaining the test data set includes: Searching for a data set corresponding to the type of the target model from a data set library to obtain the test data set; Alternatively, in response to a selection operation on a second target option in a data set import area in the model evaluation interface, the data set selected by the selection operation is used as the test data set.

4. The method according to any one of claims 1 to 3, characterized in that: The evaluating the training effect of the target model according to the test result and the annotation result to obtain the evaluation result includes: Comparing the test result with the marked result to obtain a comparison chart; The training effect of the target model is evaluated according to the comparison chart, and the evaluation result and the comparison chart are obtained and displayed in the model evaluation interface.

5. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: If the evaluation result indicates that the training effect of the target model does not meet the expected requirements, the training process information of the target model is obtained, and the training process information is statistically analyzed to obtain the factors that cause the training effect to not meet the expected requirements.

6. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Obtaining training process information of the target model; Based on the training process information of the target model, the target model and the corresponding evaluation results, an evaluation report is generated and displayed in the model evaluation interface.

7. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: In response to the selection operation of the third target option in the function display interface of the model processing application and the triggering operation of the generation control, the following steps are performed: Obtain an initial natural language model; Fine-tune the initial natural language model according to the model application function of the option corresponding to the selection operation to obtain a fine-tuned natural language model; The model processing interface corresponding to the selection operation is generated using the fine-tuned natural language model, and the model processing interface corresponding to the selection operation is switched to be displayed from the function display interface; the function display interface includes options corresponding to multiple model application functions and the generation control; the model processing interface includes the model evaluation interface.

8. A model evaluation device, characterized in that: The device comprises: A first response module is used to respond to a triggering operation of an evaluation control in a model evaluation interface, test a target model indicated by the evaluation control according to a test data set, obtain a test result, and display a marking interface in the model evaluation interface; The second response module is used to respond to the annotation operation on the annotation interface, obtain the annotation result, and evaluate the training effect of the target model according to the test result and the annotation result to obtain the evaluation result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images