Model processing method and device, storage medium, electronic equipment and program product

Abstract

The invention relates to a model processing method and device, a storage medium, electronic equipment and a program product, and the method comprises the steps: displaying a model processing platform page, the model processing platform page is used for displaying different function modules, and the different function modules comprise a model evaluation module used for evaluating the training effect of a trained model; and in response to a triggering operation on the model evaluation module, at least displaying an evaluation result used for indicating whether to continue to train the to-be-evaluated model or not. According to the scheme, the model evaluation module for evaluating the training effect of the trained model can be displayed on the page of the model processing platform, so that the training effect of the trained model can be evaluated by triggering the model evaluation module, and the evaluation result is obtained. Besides, the evaluation result is used for indicating whether to continue to train the to-be-evaluated model, so that the model can be further processed or optimized according to the evaluation result, and the model processing effect is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent technology, and in particular, to a model processing method, device, storage medium, electronic device and program product. Background Art

[0002] With the development of intelligent technology, machine learning models have emerged to handle tasks in different fields, such as text generation models, translation models, and image generation models. However, before processing tasks based on machine learning models, the machine learning models need to be processed first. Summary of the invention

[0003] The purpose of the present disclosure is to provide a model processing method, device, storage medium, electronic device and program product to solve the above technical problems.

[0004] In order to achieve the above-mentioned object, the present disclosure provides a model processing method in a first aspect, the model processing method comprising: Displaying a model processing platform page, wherein the model processing platform page is used to display different functional modules, wherein the different functional modules include a model evaluation module, and the model evaluation module is used to evaluate the training effect of the trained model; In response to a triggering operation on the model evaluation module, at least an evaluation result of the model to be evaluated is displayed, and the evaluation result is used to indicate whether to continue training the model to be evaluated.

[0005] Optionally, in response to the triggering operation on the model evaluation module, at least the evaluation result of the model to be evaluated is displayed, including: In response to a triggering operation on the model evaluation module, displaying a model evaluation page; In response to a configuration operation on a model for training effect evaluation in the model evaluation page, a model to be evaluated is obtained; in response to a configuration operation on an evaluation parameter of the model to be evaluated in the model evaluation page, a target evaluation parameter is obtained; In response to a triggering operation on an evaluation control in the model evaluation page for triggering model evaluation, determining an evaluation result of the model to be evaluated according to the target evaluation parameter and a preset evaluation data set; The evaluation result is displayed.

[0006] Optionally, the different functional modules further include a model training module, and the model training module is used to train the model. The method further includes: In response to a triggering operation on the model training module, displaying a model training page for training the model; After the training of the model is completed on the model training page, in response to a triggering operation of an import control in the model training page for importing the trained model into the model evaluation module, the model file and the prediction script of the model are packaged into a tar package, and the tar package is imported into the model evaluation module to display the tar package on the model evaluation page, wherein the model file is used to record the model information of the model, and the prediction script is used to load the model file and predict the evaluation data set based on the model; The step of obtaining the model to be evaluated in response to the configuration operation of the model for evaluating the training effect in the model evaluation page includes: In response to a selection operation on a target tar package in the model evaluation page, a model corresponding to the target tar package is determined as a model to be evaluated.

[0007] Optionally, when the to-be-evaluated model includes multiple models, the response to the triggering operation on the model evaluation module at least displays the evaluation result of the to-be-evaluated model, including: The evaluation result of each of the models to be evaluated, the comparison results between the multiple models to be evaluated, and the guidance information for guiding the user to optimize the model structure and / or model parameters of the models to be evaluated are displayed.

[0008] Optionally, the different functional modules further include a model training module, and the model training module is used to train the model. The method further includes: When the evaluation result is used to indicate to continue training the model to be evaluated, in response to a triggering operation on the model training module, displaying a model training page; In response to an editing operation on the parameters of the model to be evaluated in the model training page, a target parameter of the model to be evaluated is obtained, and / or, in response to an editing operation on the original file of the model to be evaluated in the model training page, a target original file is obtained, wherein the original file is used to record a data set used to train the model to be evaluated; In response to a triggering operation on a training control for training a model in the model training page, the model to be evaluated is trained according to the target parameters and / or the target original file to obtain a trained model to be evaluated.

[0009] Optionally, the different functional modules further include an effect analysis module, and the effect analysis module is used to analyze the model effect of the model deployed in the target environment. The method further includes: In response to a triggering operation on the effect analysis module, an effect analysis page is displayed, wherein the effect analysis page displays a plurality of deployed models; In response to a selection operation on the deployed model in the effect analysis page, analyzing the model effect of the deployed model corresponding to the selection operation to obtain an analysis result, wherein the analysis result is used to indicate whether to train and / or adjust model parameters of the deployed model corresponding to the selection operation; The analysis results are displayed.

[0010] A second aspect of the present disclosure provides a model processing device, the model processing device comprising: A first display module, used to display a model processing platform page, wherein the model processing platform page is used to display different functional modules, wherein the different functional modules include a model evaluation module, and the model evaluation module is used to evaluate the training effect of the trained model; The second display module is used to display at least the evaluation result of the model to be evaluated in response to the triggering operation of the model evaluation module, and the evaluation result is used to indicate whether to continue training the model to be evaluated.

[0011] A third aspect of the present disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of any one of the methods described in the first aspect.

[0012] A fourth aspect of the present disclosure provides an electronic device, including: a storage device having a computer program stored thereon; A processing device is used to execute the computer program in the storage device to implement the steps of any one of the methods in the first aspect.

[0013] A fifth aspect of the present disclosure provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect.

[0014] Through the above technical solution, a model evaluation module for evaluating the training effect of a trained model can be displayed on the model processing platform page, thereby, the training effect of the trained model can be evaluated by triggering the model evaluation module to obtain an evaluation result. In addition, since the evaluation result is used to indicate whether to continue training the model to be evaluated, the model can be further processed or optimized according to the evaluation result to improve the model processing effect.

[0015] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings: Figure 1 is a flow chart of a model processing method according to an exemplary embodiment of the present disclosure; Figure 2 is a schematic diagram showing a method of performing model training on a model training page according to an exemplary embodiment of the present disclosure; Figure 3 is a schematic diagram showing a method of performing model evaluation on a model evaluation page according to an exemplary embodiment of the present disclosure; Figure 4 is a schematic diagram showing an effect analysis on an effect analysis page according to an exemplary embodiment of the present disclosure; Figure 5 is a flowchart of another model processing method according to an exemplary embodiment of the present disclosure; Figure 6 is a structural block diagram of a model processing device according to an exemplary embodiment of the present disclosure; Figure 7 It is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0017] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0018] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the device is located and with the authorization given by the owner of the corresponding device.

[0019] As mentioned in the background technology, with the development of intelligent technology, machine learning models for processing tasks in different fields have emerged, such as text generation models, translation models, and image generation models. However, before processing tasks based on machine learning models, the machine learning models need to be processed first.

[0020] In the related technology, the processing of machine learning models generally includes two major links: model training and model deployment. Model training includes the collection of data sets, the preprocessing of data sets, and the training of machine learning models based on the preprocessed data sets. However, these links are often performed independently on different systems or platforms, resulting in a lack of consistency in the processing flow, which not only reduces the efficiency of model processing, but also makes it difficult to maximize the use of computing and data resources due to the dispersion of resources, thereby increasing the cost of model processing.

[0021] In order to overcome the above technical problems, the related technology proposes an integrated processing platform, which can cover basic links such as data collection, preprocessing, model training and model deployment. However, since the integrated processing platform can only provide some basic links for processing the model, it is still insufficient when the model needs to be processed and optimized in depth. For example, in the model competition scenario, it is impossible to evaluate the training effect of the model based on the integrated processing platform.

[0022] In view of this, the embodiments of the present disclosure provide a model processing method, device, storage medium, electronic device and program product to solve the above technical problems.

[0023] The embodiments of the present disclosure are further explained below with reference to the accompanying drawings.

[0024] Figure 1 is a flowchart of a model processing method according to an exemplary embodiment of the present disclosure, referring to Figure 1 , the model processing method may include the following steps: S101: Displaying a model processing platform page, where the model processing platform page is used to display different functional modules, including a model evaluation module, which is used to evaluate the training effect of a trained model.

[0025] In this embodiment, the functional modules displayed on the model processing platform page can be set according to actual conditions, and the embodiment of the present disclosure does not impose any restrictions on this. For example, in addition to the model evaluation module, the functional modules displayed on the model processing platform page may also include a data import module, a computing resource management and scheduling module, a model training module, a model reasoning module, and a model deployment module. Among them, the data import module can be used to upload data (such as data marked with labels) to the model processing platform and store it in a specified location in the model processing platform as training data or evaluation data. At the same time, the data stored in the specified location in the model processing platform provides read-only permissions to the user, so that the user can copy the data set file to the working root directory required by the user for subsequent processing operations. Among them, the data import module can support data upload in multiple formats, such as video, picture or text, and the embodiment of the present disclosure does not impose any restrictions on this. The computing resource management and scheduling module can be used by users to dynamically configure computing resources such as central processing units (CPUs), graphics processing units (GPUs), and memory according to actual task requirements, and can also monitor cluster resource usage in real time for load balancing. The model training module can configure training materials (such as training sets and validation sets), model structures, and training parameters (such as learning rate, batch size, and number of iterations), and then train the model corresponding to the configured model structure based on the configured training materials and training parameters. At the same time, in order to improve the flexibility and efficiency of model training, the model training module can also support a variety of deep learning frameworks (for example, mainstream frameworks such as TensorFlow and PyTorch) and training algorithms in possible ways. The model inference module can be used to deploy the model trained by the model training module to an inference environment, such as a server or mobile device, so that new data can be identified and analyzed based on the trained model to provide real-time and accurate prediction results. The model deployment module can save the model trained by the model training module in an executable format, and then deploy the model to the actual environment, so that the model can perform real-time prediction and analysis on the newly input data.

[0026] In this embodiment, the trained model may be a model trained through a model processing platform, or a model trained through other external platforms, and the present disclosure embodiment does not impose any restrictions on this.

[0027] When the trained model is a model obtained by training through a model processing platform, the functional modules displayed on the model processing platform page may also include a model training module for training the model. Thus, the model can be trained through the model training module to obtain a trained model, and then the trained model can be imported into the model evaluation module, so that the training effect of the trained model can be evaluated through the model evaluation module.

[0028] For example, the model processing platform page displays a model training module. When the user clicks the model training module, the model training page is displayed, such as Figure 2 As shown, the model training page may display a model structure configuration control, a training parameter configuration control, a training material configuration control, a training control, and an import control. By clicking the model structure configuration control, the model structure configuration page is displayed. The model structure configuration page displays multiple model structures. By clicking the target model structure, the model to be trained is obtained. By entering the training parameter value in the training parameter configuration control (i.e., the input box corresponding to the training parameter), the training parameters of the model to be trained are obtained. By entering the working root directory and the ratio of the training set and the validation set in the training material configuration control (i.e., the input box corresponding to the training material), the training sample is obtained. After that, by clicking the training control, the configured model to be trained is trained based on the configured training parameters and the configured training material, and after the training is completed, the training result is displayed, and the training result is used to indicate whether the training is successful. If the training result indicates that the model training fails, the configured training parameters and / or training samples can be adjusted by triggering the training parameter configuration control and / or the training material configuration control, and then the model is retrained by triggering the training control. If the training result indicates that the model training is successful, you can trigger the import control to package the model file (usually refers to the file that records the model information of the model, which may include model structure or model parameters, etc.) and the prediction script (usually refers to a program or script file used to load the trained model and use this model to predict new input data) into a tar package, and import the tar package into the model evaluation module to display the tar package on the model evaluation page.

[0029] That is to say, in a possible manner, different functional modules may further include a model training module, and the model training module is used to train the model. Accordingly, the model processing method may further include: In response to a trigger operation on a model training module, a model training page for training the model is displayed; after the training of the model is completed on the model training page, in response to a trigger operation on an import control in the model training page for importing the trained model into a model evaluation module, the model file and the prediction script of the model are packaged into a tar package, and the tar package is imported into the model evaluation module to display the tar package on the model evaluation page, wherein the model file is used to record the model information of the model, and the prediction script is used to load the model file and predict the evaluation data set based on the model.

[0030] If the trained model is a model trained through other external platforms, the model evaluation module page may also display an import control, which can be triggered to import the model trained through other external platforms into the model evaluation module.

[0031] S102: In response to a triggering operation on the model evaluation module, at least an evaluation result of the model to be evaluated is displayed, where the evaluation result is used to indicate whether to continue training the model to be evaluated.

[0032] In a possible manner, in response to a triggering operation on the model evaluation module, at least displaying the evaluation result of the model to be evaluated may include: In response to the triggering operation on the model evaluation module, the model evaluation page is displayed; in response to the configuration operation on the model used for training effect evaluation in the model evaluation page, the model to be evaluated is obtained, and in response to the configuration operation on the evaluation parameters of the model to be evaluated in the model evaluation page, the target evaluation parameters are obtained; in response to the triggering operation on the evaluation control used to trigger model evaluation in the model evaluation page, the evaluation result of the model to be evaluated is determined according to the target evaluation parameters and the preset evaluation data set; the evaluation result is displayed.

[0033] For example, the model processing platform page displays a model evaluation module. When the user clicks the model evaluation module, the model evaluation page is displayed, such as Figure 3 As shown. The model evaluation page displays an evaluation control, the upper side of the model evaluation page displays the tar package imported from the model training module, and the lower side of the model evaluation page displays the evaluation parameter configuration page. Therefore, when training the model, the user can first click on the target tar package to obtain the model to be evaluated. After that, the evaluation indicators and evaluation ratios can be configured on the evaluation parameter configuration page to obtain the target evaluation parameters. Finally, by clicking the evaluation control, the evaluation result of the model to be evaluated is displayed.

[0034] In this embodiment, the model to be evaluated may be one or more. When there are multiple models to be evaluated, for each model to be evaluated, a set of evaluation parameters may be configured for the model to be evaluated, or a set of evaluation parameters may be configured for multiple models to be evaluated, so that multiple models to be evaluated share the same set of evaluation parameters.

[0035] In this embodiment, the content of the evaluation result can be set according to the actual situation, and the embodiment of the present disclosure does not impose any restrictions on this. For example, the content of the result to be evaluated can be "the evaluation passed, no need to continue to train the model to be evaluated", or it can be "the evaluation failed, and the model to be evaluated needs to continue to be trained". It can also be the evaluation score of the model to be evaluated, so that it can be determined whether to continue to train the model to be evaluated based on the size of the evaluation score.

[0036] For example, when there is only one model to be evaluated, the evaluation result may include an effect score. Thus, after obtaining the model to be evaluated, the evaluation indicators and evaluation ratios, etc., may be configured on the evaluation parameter configuration page to obtain the target evaluation parameters. Finally, by clicking the evaluation control, the effect score of the model to be evaluated is displayed. When there are multiple models to be evaluated, the evaluation result may include an effect score and a performance score. Thus, after obtaining multiple models to be evaluated, the evaluation indicators and evaluation ratios, etc. of each model to be evaluated may be configured on the evaluation parameter configuration page to obtain the target evaluation parameters. Finally, by clicking the evaluation control, the effect score and performance score of the model to be evaluated are displayed.

[0037] The evaluation index and the evaluation ratio may be determined according to actual conditions, and the embodiments of the present disclosure do not impose any limitation on this.

[0038] For example, when there is only one model to be evaluated, and the model to be evaluated is a single target detection model, the evaluation indicators can be set to precision and recall, and the evaluation ratio of precision and recall can be set to 6:4. Therefore, when the evaluation control is clicked, the preset evaluation data set can be called through the underlying code of the model evaluation module, the result is predicted based on the model, and the effect score is obtained based on the prediction result, the configured evaluation indicators and the evaluation ratio, that is:

[0039]

[0040]

[0041] in, Represents the effect score of the single target detection model, (True Positive) indicates that the model describes the positive samples as the number of positive samples, (False Positive) indicates the number of negative samples described by the model as positive samples. (False Negative) indicates the number of positive samples described by the model as negative samples.

[0042] For example, when there is only one model to be evaluated, and the model to be evaluated is a multi-target detection model, the evaluation indicators can be set to precision and recall, and the evaluation ratio of precision and recall can be set to 6:4. Therefore, when the evaluation control is clicked, the preset evaluation data set can be called through the underlying code of the model evaluation module, the result is predicted based on the model, and the effect score is obtained based on the prediction result, the configured evaluation indicators and the evaluation ratio, that is:

[0043]

[0044] in, Represents the effect score of the multi-target detection model, Represents the effect score of each category in the multi-target detection model.

[0045] For example, when there are multiple models to be evaluated, the evaluation indicators can be set to precision, recall, the mean running time of each model, and the variance of the running time of each model, and the evaluation ratio of precision and recall can be set to 6:4, the median running time of multiple models can be used as the benchmark, and the benchmark score can be set to 60 points. Therefore, when the evaluation control is clicked, the effect score of each model can be obtained through the above method, and the performance score of each model can be obtained through the following formula: If the model running time is less than the mean, the performance score of the model is:

[0046] If the model running time is greater than the mean, the performance score of the model is:

[0047] in, represents the mean running time, represents the running time variance, Indicates The running time of the model, Then, based on the performance score and effect score, the total score of the model is obtained, namely:

[0048] in, Indicates the algorithm effect score, Indicates the algorithm performance score.

[0049] As mentioned above, the model to be evaluated can be one or more. When there are multiple models to be evaluated, in order to facilitate users to timely understand the differences between the models and the optimization direction, the comparison results and guidance information between the models can also be displayed, and the guidance information is used to guide the user to optimize the model structure and / or model parameters of the model to be evaluated. That is to say, in a possible way, when there are multiple models to be evaluated, in response to the triggering operation of the model evaluation module, at least the evaluation results of the model to be evaluated are displayed, which may include: The evaluation result of each model to be evaluated, the comparison results between multiple models to be evaluated, and guidance information for guiding the user to optimize the model structure and / or model parameters of the model to be evaluated are displayed.

[0050] After obtaining the evaluation result indicating that the model to be evaluated should not be further trained, the model to be evaluated can be imported into the model deployment module, and the model to be evaluated can be deployed into the actual environment through the model deployment module. After obtaining the guidance information and / or the evaluation result indicating that the model to be evaluated should be continued to be trained, if the model to be evaluated is a model obtained by training the model training module, the model can continue to be trained through the model training module. If the model to be evaluated is a model obtained by training other external platforms, the model can continue to be trained through other external platforms, or the model can be imported into the model training module, and the model can be trained through the model training module. That is to say, in a possible manner, the model processing method can also include: When the evaluation result is used to indicate that the model to be evaluated should continue to be trained, in response to a trigger operation on the model training module, a model training page is displayed; in response to an editing operation on the training parameters of the model to be evaluated in the model training page, target training parameters of the model to be evaluated are obtained, and / or, in response to an editing operation on the original file of the model to be evaluated in the model training page, a target original file is obtained, and the original file is used to record the data set used to train the model to be evaluated; in response to a trigger operation on a training control for training the model in the model training page, the model to be evaluated is trained according to the target training parameters and / or the target original file to obtain a trained model to be evaluated.

[0051] In this embodiment, the related operations of training the model on the model training page can refer to the above-mentioned related descriptions and will not be repeated here.

[0052] Through the above technical solution, a model evaluation module for evaluating the training effect of a trained model can be displayed on the model processing platform page, thereby, the training effect of the trained model can be evaluated by triggering the model evaluation module to obtain an evaluation result. In addition, since the evaluation result is used to indicate whether to continue training the model to be evaluated, the model can be further processed or optimized according to the evaluation result to improve the model processing effect.

[0053] In a possible manner, different functional modules may further include an effect analysis module, which is used to analyze the model effect of the model deployed in the target environment. Accordingly, the model processing method may further include: In response to a trigger operation on the effect analysis module, an effect analysis page is displayed, and the effect analysis page displays multiple deployed models; in response to a selection operation on a deployed model in the effect analysis page, the model effect of the deployed model corresponding to the selection operation is analyzed to obtain an analysis result, and the analysis result is used to indicate whether to train and / or adjust model parameters of the deployed model corresponding to the selection operation; the analysis result is displayed.

[0054] For example, the model processing platform page displays an effect analysis module. When the user clicks the effect analysis module, the effect analysis page is displayed, such as Figure 4 As shown, the effect analysis page displays the models and analysis controls that have been deployed to the actual environment. By clicking the target model, the model to be analyzed is obtained. By clicking the analysis control, the performance data of the target model in the actual application, such as accuracy and response time, is obtained. Then, based on the performance data, analysis is performed to obtain analysis results indicating whether to train the target model and / or adjust the model parameters, and the analysis results are displayed.

[0055] In this embodiment, if the analysis result is used to indicate the training and / or adjustment of model parameters of the deployed model corresponding to the selection operation, the model can be retrained by triggering the model training module, or by adjusting the model parameters and performing corresponding training until an evaluation result is obtained indicating that the model to be evaluated should not continue to be trained.

[0056] Through the above method, after the model is deployed in the actual environment, the performance data of the model in the actual application can be collected through the effect analysis module, and the model can be continuously optimized and iterated based on the performance data to ensure that the model always meets business needs.

[0057] To facilitate understanding of the model processing method disclosed herein, a possible implementation of this solution is described below: For example, in a model competition scenario, the competition organizer can import the training data into a designated location in the model processing platform in advance through the data import module. During the competition, each contestant can first configure the computing resources required in the model training process in the computing resource management and scheduling module according to the competition requirements of the competition organizer. After that, by triggering the model training module, the model training page is displayed, and the model is trained in the model training page through the aforementioned description process. After the model training is completed, the trained model can be deployed to the reasoning environment through the model reasoning module, for example, deployed locally, and the new data is identified and analyzed by executing the local evaluation instructions to obtain the verification result. If the verification fails, the model training module can continue to be trained. If the verification passes, the model can be imported into the model evaluation module. After that, the competition organizer can trigger the model evaluation module to display the model evaluation page, and configure the model and the evaluation parameters on the model evaluation page through the aforementioned description process to obtain each model evaluation result, thereby determining the competition result according to each model evaluation result.

[0058] To facilitate understanding of the model processing method disclosed herein, another possible implementation of this solution is described below: For example, Figure 5As shown, in the model development scenario, the model developer can import the training data into the specified location in the model processing platform through the data import module. After that, the model developer can configure the computing resources required in the model training process in the computing resource management and scheduling module according to the model requirements. Then, by triggering the model training module, the model training page is displayed, and the model is trained in the model training page through the aforementioned description process. After the model training is completed, the trained model can be deployed to the reasoning environment through the model inference module, for example, deployed locally, and the new data is identified and analyzed by executing the local evaluation instructions to obtain the verification result. If the verification fails, the model training module can continue to be trained; if the verification passes, the model can be imported into the model evaluation module. After that, the model developer can trigger the model evaluation module to display the model evaluation page, and configure the model and the evaluation parameters on the model evaluation page through the aforementioned description process to obtain the model evaluation result. If the evaluation result indicates to continue training the model, the model training module can be re-triggered to execute the model training process. If the evaluation results indicate that the model should not be trained further, the model can be deployed to the actual environment through the model deployment module, and the performance data of the model in actual applications can be collected through the effect analysis module, and the performance data can be analyzed to obtain the analysis results. If the analysis results are used to indicate that the target model should be trained and / or the model parameters should be adjusted, the training data can be re-imported into the specified location in the model processing platform through the data import module, and the above development process can be repeated to achieve continuous optimization and iteration of the model to ensure that the model always meets business needs.

[0059] Based on the same concept, the present disclosure provides a model processing device, such as Figure 6 As shown, the model processing device 600 may include: A first display module 601 is used to display a model processing platform page, wherein the model processing platform page is used to display different functional modules, wherein the different functional modules include a model evaluation module, and the model evaluation module is used to evaluate the training effect of the trained model; The second display module 602 is used to display at least the evaluation result of the model to be evaluated in response to the triggering operation of the model evaluation module, and the evaluation result is used to indicate whether to continue training the model to be evaluated.

[0060] Through the above-mentioned model processing device 600, a model evaluation module for evaluating the training effect of the trained model can be displayed on the model processing platform page, thereby, the training effect of the trained model can be evaluated by triggering the model evaluation module to obtain an evaluation result. In addition, since the evaluation result is used to indicate whether to continue training the model to be evaluated, the model can be further processed or optimized according to the evaluation result to improve the model processing effect.

[0061] In a possible manner, the second display module 602 may include: A first display submodule, configured to display a model evaluation page in response to a triggering operation on the model evaluation module; A configuration submodule, configured to obtain a model to be evaluated in response to a configuration operation on a model for training effect evaluation in the model evaluation page, and to obtain target evaluation parameters in response to a configuration operation on evaluation parameters of the model to be evaluated in the model evaluation page; A determination submodule, configured to determine an evaluation result of the model to be evaluated according to the target evaluation parameter and a preset evaluation data set in response to a triggering operation on an evaluation control for triggering model evaluation in the model evaluation page; The second display submodule is used to display the evaluation results.

[0062] In a possible manner, the different functional modules further include a model training module, and the model training module is used to train the model. Accordingly, the model processing device may further include: A third display module, configured to display a model training page for training a model in response to a triggering operation on the model training module; An import module, for, after completing the training of the model on the model training page, in response to a triggering operation of an import control on the model training page for importing the trained model into the model evaluation module, packaging the model file and the prediction script of the model into a tarball, and importing the tarball into the model evaluation module, so as to display the tarball on the model evaluation page, wherein the model file is used to record the model information of the model, and the prediction script is used to load the model file and predict the evaluation data set based on the model; Accordingly, the configuration submodule may be used to: in response to a selection operation on a target tar package in the model evaluation page, determine the model corresponding to the target tar package as the model to be evaluated.

[0063] In a possible manner, when there are multiple models to be evaluated, the second display module 602 can be used to: display the evaluation results of each model to be evaluated, the comparison results between multiple models to be evaluated, and guidance information for guiding the user to optimize the model structure and / or model parameters of the model to be evaluated.

[0064] In a possible manner, the different functional modules further include a model training module, and the model training module is used to train the model. Accordingly, the model processing device may further include: A fourth display module, configured to display a model training page in response to a triggering operation on the model training module when the evaluation result is used to indicate that the model to be evaluated should continue to be trained; A configuration module, configured to obtain target parameters of the model to be evaluated in response to an editing operation on parameters of the model to be evaluated in the model training page, and / or, in response to an editing operation on an original file of the model to be evaluated in the model training page, obtain a target original file, wherein the original file is used to record a data set used to train the model to be evaluated; A training module is used to respond to the triggering operation of the training control for training the model in the model training page, train the model to be evaluated according to the target parameters and / or the target original file, and obtain the trained model to be evaluated.

[0065] In a possible manner, the different functional modules further include an effect analysis module, which is used to analyze the model effect of the model deployed in the target environment. Accordingly, the model processing device may further include: A fourth display module, configured to display an effect analysis page in response to a trigger operation on the effect analysis module, wherein the effect analysis page displays a plurality of deployed models; An analysis module, configured to analyze the model effect of the deployed model corresponding to the selection operation in the effect analysis page in response to the selection operation of the deployed model, and obtain an analysis result, wherein the analysis result is used to indicate whether to train and / or adjust model parameters of the deployed model corresponding to the selection operation; The fifth display module is used to display the analysis results.

[0066] Regarding the model processing device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0067] Based on the same concept, an embodiment of the present disclosure further provides a computer-readable medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned model processing methods are implemented.

[0068] Based on the same concept, the embodiment of the present disclosure further provides an electronic device, including: a storage device having a computer program stored thereon; A processing device is used to execute the computer program in the storage device to perform the steps of any of the above-mentioned model processing methods.

[0069] Based on the same concept, an embodiment of the present disclosure further provides a computer program product, including a computer program, which implements the steps of any of the above-mentioned model processing methods when executed by a processor.

[0070] Figure 7 FIG. 7 is a block diagram of an electronic device 700 according to an exemplary embodiment. Figure 7 As shown, the electronic device 700 may include: a processor 701 , a memory 702 . The electronic device 700 may also include one or more of a multimedia component 703 , an input / output (I / O) interface 704 , and a communication component 705 .

[0071] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the above-mentioned model processing method. The memory 702 is used to store various types of data to support the operation of the electronic device 700. For example, these data may include instructions for any application or method used to operate on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 702 or sent through the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules, and the other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, etc., or a combination of one or more of them, is not limited here. Therefore, the corresponding communication component 705 may include: Wi-Fi module, Bluetooth module, NFC module, etc.

[0072] In an exemplary embodiment, the electronic device 700 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned model processing method.

[0073] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned model processing method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 702 including program instructions, and the above-mentioned program instructions can be executed by the processor 701 of the electronic device 700 to complete the above-mentioned model processing method.

[0074] In another exemplary embodiment, a computer program product is also provided. The computer program product includes a computer program executable by a programmable device. The computer program has a code portion for executing the above-mentioned model processing method when executed by the programmable device.

[0075] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0076] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0077] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A model processing method, characterized in that: The model processing method comprises: Displaying a model processing platform page, wherein the model processing platform page is used to display different functional modules, wherein the different functional modules include a model evaluation module, and the model evaluation module is used to evaluate the training effect of the trained model; In response to a triggering operation on the model evaluation module, at least an evaluation result of the model to be evaluated is displayed, and the evaluation result is used to indicate whether to continue training the model to be evaluated.

2. The model processing method according to claim 1, characterized in that: The step of at least displaying the evaluation result of the model to be evaluated in response to the triggering operation on the model evaluation module includes: In response to a triggering operation on the model evaluation module, displaying a model evaluation page; In response to a configuration operation on a model for training effect evaluation in the model evaluation page, a model to be evaluated is obtained; in response to a configuration operation on an evaluation parameter of the model to be evaluated in the model evaluation page, a target evaluation parameter is obtained; In response to a triggering operation on an evaluation control in the model evaluation page for triggering model evaluation, determining an evaluation result of the model to be evaluated according to the target evaluation parameter and a preset evaluation data set; The evaluation result is displayed.

3. The model processing method according to claim 2, characterized in that: The different functional modules further include a model training module, and the model training module is used to train the model. The method further includes: In response to a triggering operation on the model training module, displaying a model training page for training the model; After the training of the model is completed on the model training page, in response to a triggering operation of an import control in the model training page for importing the trained model into the model evaluation module, the model file and the prediction script of the model are packaged into a tar package, and the tar package is imported into the model evaluation module to display the tar package on the model evaluation page, wherein the model file is used to record the model information of the model, and the prediction script is used to load the model file and predict the evaluation data set based on the model; The step of obtaining the model to be evaluated in response to the configuration operation of the model for evaluating the training effect in the model evaluation page includes: In response to a selection operation on a target tar package in the model evaluation page, a model corresponding to the target tar package is determined as a model to be evaluated.

4. The model processing method according to any one of claims 1 to 3, characterized in that: When the to-be-evaluated model includes multiple ones, the response to the triggering operation of the model evaluation module at least displays the evaluation result of the to-be-evaluated model, including: The evaluation result of each of the models to be evaluated, the comparison results between the multiple models to be evaluated, and the guidance information for guiding the user to optimize the model structure and / or model parameters of the models to be evaluated are displayed.

5. The model processing method according to any one of claims 1 to 3, characterized in that: The different functional modules further include a model training module, and the model training module is used to train the model. The method further includes: When the evaluation result is used to indicate to continue training the model to be evaluated, in response to a triggering operation on the model training module, displaying a model training page; In response to an editing operation on the parameters of the model to be evaluated in the model training page, a target parameter of the model to be evaluated is obtained, and / or, in response to an editing operation on the original file of the model to be evaluated in the model training page, a target original file is obtained, wherein the original file is used to record a data set used to train the model to be evaluated; In response to a triggering operation on a training control for training a model in the model training page, the model to be evaluated is trained according to the target parameters and / or the target original file to obtain a trained model to be evaluated.

6. The model processing method according to any one of claims 1 to 3, characterized in that: The different functional modules further include an effect analysis module, and the effect analysis module is used to analyze the model effect of the model deployed in the target environment. The method further includes: In response to a triggering operation on the effect analysis module, an effect analysis page is displayed, wherein the effect analysis page displays a plurality of deployed models; In response to a selection operation on the deployed model in the effect analysis page, analyzing the model effect of the deployed model corresponding to the selection operation to obtain an analysis result, wherein the analysis result is used to indicate whether to train and / or adjust model parameters of the deployed model corresponding to the selection operation; The analysis results are displayed.

7. A model processing device, characterized in that: The model processing device comprises: A first display module, used to display a model processing platform page, wherein the model processing platform page is used to display different functional modules, wherein the different functional modules include a model evaluation module, and the model evaluation module is used to evaluate the training effect of the trained model; The second display module is used to display at least the evaluation result of the model to be evaluated in response to the triggering operation of the model evaluation module, and the evaluation result is used to indicate whether to continue training the model to be evaluated.

8. A non-transitory 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 6 are implemented.

9. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, 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 6 are implemented.

Images