Competition information generation method and device based on large model, equipment, storage medium and program product

Abstract

The invention relates to a competition information generation method and device based on a large model, equipment, a storage medium and a program product. The method comprises the following steps: receiving a model processing instruction sent by terminal equipment, wherein the model processing instruction comprises identification information of a target object for triggering the model processing instruction and at least one model operation indicated by the model processing instruction; at least one model operation is executed on a to-be-adjusted competition model corresponding to the identification information of the target object, an adjusted competition model is obtained, and the competition model is a machine learning model created by the target object under the subject category of the current competition; using the trained target prediction model to predict the adjusted match model to obtain prediction data of the adjusted match model; and according to the prediction data and the historical competition data of the target object, generating competition information of the target object in the current competition. By adopting the method, the generation speed of the competition information can be improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, storage medium, and program product for generating event information based on a large model. Background Technology

[0002] With the development of artificial intelligence, there are more and more modeling competitions for AI models. In these competitions, participants can generate new models and then fine-tune them to achieve better processing results and higher rankings.

[0003] In related technologies, during modeling competitions, after participants complete the model configuration, it is often necessary to manually test the performance of each model to obtain the competition information for each participant. Therefore, the current AI model modeling competitions have a relatively slow rate of information generation. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, equipment, storage medium, and program product for generating event information based on a large model, which can improve the speed of event information generation, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for generating event information based on a large model, including:

[0006] The device receives a model processing instruction sent by a terminal device, the model processing instruction including identification information of the target object that triggers the model processing instruction and at least one model operation indicated by the model processing instruction;

[0007] Perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object to obtain the adjusted event model, wherein the event model is a machine learning model created for the target object under the theme category of the current event;

[0008] Using the trained target prediction model, the adjusted event model is used to make predictions, and the prediction data of the adjusted event model is obtained. The prediction data represents the prediction performance of the model function specified for the current event.

[0009] Based on the predicted data and the target object's historical participation data, generate the target object's event information for the current event.

[0010] In one embodiment, the step of using a trained target prediction model to predict the adjusted event model and obtaining the prediction data of the adjusted event model includes:

[0011] Obtain the model parameters of the adjusted competition model, as well as the computing resource call data of the adjusted competition model;

[0012] The model parameters and the computing power resource call data are input into the trained target prediction model, and the prediction data output by the trained target prediction model is obtained.

[0013] The trained target prediction model is used to extract features from the model parameters and the computing resource call data, respectively, and to perform weighted merging of the features extracted from the model parameters and the computing resource call data, and to generate the prediction data using the merged features.

[0014] In one embodiment, after generating the target object's competition information for the current competition based on the predicted data and the target object's historical competition data, the process includes:

[0015] Obtain the ranking data of each object participating in the current event;

[0016] According to a preset sorting method, the event information of the target object in the current event is inserted into the event ranking data of each object in the current event to obtain the updated ranking data.

[0017] In one embodiment, inserting the event information of the target object into the event ranking data of each object in the current event includes:

[0018] Using either bubble sort or insertion sort, the event information of the target object in the current event is inserted into the event ranking data of each object in the current event.

[0019] In one embodiment, before generating the target object's competition information for the current competition based on the predicted data and the target object's historical competition data, the method further includes:

[0020] The predicted data and the historical competition data of the target object are preprocessed. The preprocessing operations include at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

[0021] In one embodiment, prior to receiving the model processing instructions sent by the receiving terminal device, the method includes:

[0022] Receive the entry request from the target object, the entry request including the entry information of the target object;

[0023] By identifying preset keywords in the target candidate's competition materials, the target candidate's historical performance data can be determined;

[0024] Based on the target object's historical performance data, grant the target object participation permission for the current competition. This participation permission allows the target object to send the model processing instructions through the terminal device.

[0025] Secondly, this application provides a large-scale model-based event information generation device, comprising:

[0026] A receiving module is used to receive a model processing instruction sent by a terminal device. The model processing instruction includes identification information of the target object that triggers the model processing instruction and at least one model operation indicated by the model processing instruction.

[0027] An execution module is used to perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object, so as to obtain an adjusted event model, wherein the event model is a machine learning model created for the target object under the theme category of the current event;

[0028] The prediction module is used to use a trained target prediction model to predict the adjusted event model and obtain the prediction data of the adjusted event model. The prediction data represents the prediction performance of the model function specified for the current event.

[0029] The generation module is used to generate the event information of the target object in the current event based on the predicted data and the historical participation data of the target object.

[0030] In one embodiment, the prediction module is further configured to acquire the model parameters of the adjusted competition model and the computing resource call data of the adjusted competition model; input the model parameters and the computing resource call data into the trained target prediction model, and acquire the prediction data output by the trained target prediction model.

[0031] The trained target prediction model is used to extract features from the model parameters and the computing resource call data, respectively, and to perform weighted merging of the features extracted from the model parameters and the computing resource call data, and to generate the prediction data using the merged features.

[0032] In one embodiment, the generation module is further configured to obtain the competition ranking data of each object participating in the current competition; and according to a preset sorting method, insert the competition information of the target object in the current competition into the competition ranking data of each object in the current competition to obtain updated ranking data.

[0033] In one embodiment, the generation module is further configured to use a sorting method such as bubble sort or insertion sort to insert the event information of the target object in the current event into the event ranking data of each object in the current event.

[0034] In one embodiment, the generation module is further configured to perform preprocessing operations on the predicted data and the historical competition data of the target object, the preprocessing operations including at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

[0035] In one embodiment, the receiving module is further configured to receive the participation request of the target object, the participation request including the participation information of the target object;

[0036] The execution module is used to determine the target object's historical performance data by identifying preset keywords in the target object's competition data; and to grant the target object participation permission for the current competition based on the target object's historical performance data. The participation permission is used to allow the target object to send the model processing instructions through the terminal device.

[0037] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for generating event information based on a large model as described in the first aspect.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the large-model-based event information generation method of the first aspect described above.

[0039] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the large-model-based event information generation method of the first aspect described above.

[0040] The aforementioned method, apparatus, device, storage medium, and program product for generating event information based on a large model receive model processing instructions sent by a terminal device. These instructions include the identifier information of the target object triggering the instructions and at least one model operation indicated by the instructions. The method executes at least one model operation on the event model to be adjusted corresponding to the identifier information of the target object, obtaining an adjusted event model. This adjusted event model is a machine learning model created for the target object under the current event's theme category. A trained target prediction model is used to predict the adjusted event model, obtaining prediction data that characterizes the predictive performance of the model function specified for the current event. Based on the prediction data and the target object's historical participation data, event information for the target object under the current event is generated. This method allows direct execution of model operations on the event model to be adjusted based on the model processing instructions, and direct prediction of the performance of the model function specified for the current event using a trained target prediction model, thereby obtaining event data. Since manual testing of each model is unnecessary, the speed of event information generation can be improved. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 An application environment diagram for a method for generating event information based on a large model, provided in an embodiment of this application;

[0043] Figure 2 A flowchart illustrating a method for generating event information based on a large model, provided in an embodiment of this application;

[0044] Figure 3 A flowchart illustrating another method for generating event information based on a large model, provided in an embodiment of this application;

[0045] Figure 4 A flowchart illustrating another method for generating event information based on a large model, provided in an embodiment of this application;

[0046] Figure 5 A structural block diagram of a large-model-based event information generation device provided in an embodiment of this application;

[0047] Figure 6 This is an internal structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0049] The event information generation method based on a large model provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal device 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers.

[0050] Terminal device 102 can send model processing instructions to the server. These instructions include the identifier information of the target object triggering the instruction and at least one model operation specified in the instruction. Server 104 performs at least one model operation on the event model to be adjusted corresponding to the target object's identifier information, obtaining an adjusted event model. This adjusted event model is a machine learning model created for the target object within the current event's theme category. Subsequently, server 104 uses a pre-trained target prediction model to predict the adjusted event model, obtaining prediction data that characterizes the predictive performance of the model function specified for the current event. Finally, server 104 generates event information for the target object within the current event based on the prediction data and the target object's historical participation data.

[0051] The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, and projection devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0052] In one exemplary embodiment, such as Figure 2 As shown, a method for generating event information based on a large model is provided, and this method is applied to... Figure 1 The following explanation uses the servers in the example, including S2101-S204:

[0053] S201. Receive a model processing instruction sent by a terminal device, wherein the model processing instruction includes identification information of the target object that triggers the model processing instruction and at least one model operation indicated by the model processing instruction.

[0054] In some embodiments, during a modeling competition, the target object can send model processing instructions to the server via a terminal device, thereby instructing the competition model to be adjusted to perform at least one model operation.

[0055] The target objects mentioned above can be users, teams, contestants, reviewers, etc. The identifiers of these target objects can be their ID number, name, nickname, account name, etc.

[0056] In some embodiments, after the target object starts the current event, it can first create a machine learning model under a theme category of the current event as the initial event model. Subsequently, the target object can continuously send model processing instructions to the server through the terminal device to perform at least one model operation on the event model, including model training, model parameter adjustment, etc., thereby continuously optimizing the event model.

[0057] The aforementioned current competition can be an artificial intelligence modeling competition, and the theme category of the aforementioned current competition can include image recognition, speech recognition, intelligent customer service, etc. This application embodiment does not make any further details on this.

[0058] In some embodiments, before receiving model processing instructions from the terminal device, the server may also receive a participation request from the target object, which includes the target object's participation information. Subsequently, the server can identify the target object's historical performance data by recognizing preset keywords in the target object's participation information. Finally, based on the target object's historical performance data, the server can grant the target object participation rights for the current competition, which allows the target object to send model processing instructions through the terminal device.

[0059] The aforementioned entry request can be an application submitted to the competition management system, whereby the server reviews and manages the target applicant to determine eligibility. The entry materials may include the target applicant's identity information, historical participation models, historical model ranking data, past participation experience, GitHub project links, and a statement of motivation for participation.

[0060] The aforementioned preset keywords may include grades, rankings, age, etc.

[0061] It should be understood that the review criteria in this application embodiment may include one or more criteria. For example, historical model ranking data may be used as the screening criterion, thereby using "ranking" as a preset keyword to filter the historical performance data of the target object from the object's participation materials. Subsequently, the historical performance data of the target object is compared with the performance screening value to determine whether the target object is allowed to participate.

[0062] For example, if the target's historical performance data is higher than or equal to the performance screening value (5th place), then the target is allowed to participate. Accordingly, the target can be granted participation rights for the current competition and is allowed to send model processing instructions through the terminal device. If the target's historical performance data is lower than the performance screening value (5th place), then the target is not allowed to participate. Accordingly, the target can not be granted participation rights for the current competition and is not allowed to send model processing instructions through the terminal device.

[0063] It should be noted that participation permissions for the current event can be assigned to a specific target. If the target is a user, the participation permissions granted for the current event will only apply to that user. If the target is a team, the participation permissions granted for the current event will only apply to that team.

[0064] In some embodiments, the server may also send a competition page to the terminal device so that the terminal device can display the competition page and trigger a participation application or model processing instruction through the controls on the competition page.

[0065] For example, the event page can also provide the target's participation status information so that the target can promptly determine whether they have been granted participation rights in the current event.

[0066] For example, the event list on the event page supports filtering and viewing based on event status information. The event list can display a filter control that allows users to select event status information such as "All," "Registering," "In Progress," or "Completed." Subsequently, when a target user selects the "Registering" status in the filter control, the event list will only display events currently accepting registrations, helping the target user quickly find new events they can participate in.

[0067] For example, the competition page may also include information such as competition introduction, competition problem description, related materials, and competition announcements. For instance, the competition introduction might include "This competition is for machine learning models that predict customer churn." The competition problem description might require building a model to predict whether customers will churn within the next 6 months and providing evaluation metrics, which could include the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). Related materials might include links to relevant academic papers and industry reports on churn analysis of the target audience. Competition announcements might include information such as the latest dataset updates and deadline changes.

[0068] S202. Perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object to obtain the adjusted event model.

[0069] In this step, after the server receives the model processing instruction sent by the terminal device, it can perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object, and obtain the adjusted event model.

[0070] In some embodiments, the event model is a machine learning model created for the target object under the theme category of the current event.

[0071] The aforementioned current competition can be an artificial intelligence modeling competition, and the theme category of the aforementioned current competition can include image recognition, speech recognition, intelligent customer service, etc. This application embodiment does not make any further details on this.

[0072] In some embodiments, the above-mentioned model processing instructions may instruct the server to perform at least one model operation on the competition model. The model operation may include model training and model parameter adjustment. This application embodiment does not limit this.

[0073] S203. Using the trained target prediction model, predict the adjusted event model to obtain the prediction data of the adjusted event model.

[0074] In this step, after the server performs at least one model operation on the event model to be adjusted corresponding to the identification information of the target object, and obtains the adjusted event model, the trained target prediction model can be used to predict the adjusted event model to obtain the prediction data of the adjusted event model.

[0075] The prediction data represents the predictive performance of the model function specified for the current event.

[0076] For example, model functions may include image recognition, speech recognition, and customer process prediction accuracy, while the performance of model functions may include the model's recognition accuracy and the intensity of computing resource allocation.

[0077] For example, after the target entity adjusts the competition model, the adjusted model can be input into a pre-trained target prediction model. This pre-trained target prediction model then predicts the performance of the participant's adjusted competition model. The target prediction model can access historical model data and historical prediction data to predict the target entity's adjusted competition model, recording the prediction results each time. Finally, after the target entity has repeatedly used the pre-trained target prediction model to predict the adjusted competition model, it can submit the finalized competition model for the competition.

[0078] In some embodiments, the server can first obtain the model parameters of the adjusted competition model, as well as the computing resource allocation data of the adjusted competition model. Subsequently, the server can input the model parameters and computing resource allocation data into the trained target prediction model and obtain the prediction data output by the trained target prediction model.

[0079] The trained target prediction model is used to extract features from the model parameters and computing resource call data respectively. The features extracted from the model parameters and computing resource call data are weighted and merged, and the merged features are used to generate prediction data.

[0080] For example, the target prediction model mentioned above may include linear regression model, logistic regression model, decision tree model, random forest model, support vector machine model or neural network model, etc.

[0081] In some embodiments, before generating the target object's event information for the current event based on the predicted data and the target object's historical participation data, the server may also perform preprocessing operations on the predicted data and the target object's historical participation data.

[0082] The preprocessing operations include at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

[0083] For example, the server can preprocess the predicted data and historical competition data of the target object. Data cleaning and data format conversion can be implemented using Python data processing libraries. Missing value handling refers to identifying and processing missing values ​​in the predicted data and historical competition data of the target object, which can be achieved using methods such as interpolation, deletion, or imputation. Outlier handling refers to detecting and processing outliers in the predicted data and historical competition data of the target object, which can be identified and processed using statistical methods, machine learning methods, or domain knowledge, or visual processing methods can also be used.

[0084] S204. Based on the predicted data and the target object's historical participation data, generate the target object's event information for the current event.

[0085] In this step, the server uses the trained target prediction model to predict the adjusted event model. After obtaining the prediction data of the adjusted event model, it can generate the event information of the target object in the current event based on the prediction data and the historical participation data of the target object.

[0086] The target's historical participation data may include model data of historical participants, historical competition data, and historical participant data.

[0087] For example, the server can extract relevant data from the target object's historical participation data based on the type of prediction data. Then, the server merges the prediction data and the extracted relevant data to obtain the target object's event information in the current event.

[0088] In some embodiments, after the server generates the event information of the target object in the current event, it can first obtain the event ranking data of each object participating in the current event. Subsequently, the server can insert the event information of the target object in the current event into the event ranking data of each object in the current event according to a preset sorting method, thereby obtaining updated ranking data.

[0089] The aforementioned preset sorting method may include the order of event rankings, the order of start times, etc.

[0090] For example, the server can use bubble sort or insertion sort to insert the event information of the target object into the event ranking data of each object in the current event.

[0091] In some embodiments, the server may also send the updated ranking data to the terminal device so that the terminal device can display the updated ranking data.

[0092] In the example, the target can view the updated ranking data displayed on the terminal device and continue to optimize the competition model based on the rankings and other data of this competition in order to participate in the next competition. The ranking may be based on different rules and indicators, such as scores, time, points, etc., and this application embodiment does not limit this. When ranking, factors such as tied rankings and the complexity of ranking rules can also be taken into account to ensure the accuracy and fairness of the ranking results.

[0093] For example, ranking data can be displayed to users through the front-end user interface of a terminal device. Results announcements may include ranking tables, statistics of various results, award settings, and other information. When displaying results, factors such as interface design and user experience should be considered so that users can clearly and intuitively view the competition rankings and results. Common front-end development technologies include Hypertext Markup Language (HTML), Cascading Style Sheets (CSS), and JavaScript, and front-end frameworks can be used to accelerate the development process.

[0094] For example, if the current competition's theme is image classification, the terminal device can build a neural network model and continuously train and optimize its parameters through model processing instructions. This allows the target prediction model to predict the neural network model's performance on the test set, such as accuracy and F1 score. After adjusting the neural network model, the target user can select the best model version on the terminal device and click the "Submit" button to submit it to the judging system on the competition platform on the server. Subsequently, the target user can view their ranking changes in real time on the dashboard and understand the gap between themselves and other participants. This integrated working environment allows the target user to focus on model optimization rather than switching between various tools and platforms, thereby improving participation efficiency and experience.

[0095] It should be understood that the competition information generation method based on large models provided in this application embodiment, through the competition platform created by the server, can also provide entry points for interactive modeling, model evaluation, submission of competition results, and viewing of real-time rankings. In addition, it can also perform related information queries and data display. The interactive modeling function allows the target object to develop models directly on the platform, reducing the need to switch between different tools. The model evaluation function can provide instant feedback, helping to quickly iterate and improve the model. The result submission function simplifies the competition process, allowing the target object to easily update the latest results. The real-time ranking display increases the sense of competition and participation, encouraging the target object to continuously improve their models.

[0096] The event information generation method based on a large model provided in this application receives a model processing instruction sent by a terminal device. The model processing instruction includes the identification information of the target object that triggered the model processing instruction and at least one model operation indicated by the model processing instruction. At least one model operation is performed on the event model to be adjusted corresponding to the identification information of the target object to obtain an adjusted event model. The event model is a machine learning model created for the target object under the theme category of the current event. A trained target prediction model is used to predict the adjusted event model to obtain prediction data of the adjusted event model. The prediction data characterizes the predictive performance of the model function specified for the current event. Based on the prediction data and the historical participation data of the target object, event information of the target object under the current event is generated. This method can directly execute model operations on the event model to be adjusted based on the model processing instruction, and directly predict the performance of the model function specified for the current event through the trained target prediction model, thereby obtaining event data. Since manual testing of each model is not required, the speed of event information generation can be improved.

[0097] The following explains how to make predictions using the adjusted event model. Figure 3 A flowchart illustrating another method for generating event information based on a large model, as provided in this application embodiment, is shown below. Figure 3 As shown, this method for generating event information based on a large model includes S301-S305:

[0098] S301, Receive model processing instructions sent by the terminal device.

[0099] The model processing instruction includes the identification information of the target object that triggered the model processing instruction and at least one model operation indicated by the model processing instruction.

[0100] S302. Perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object to obtain the adjusted event model.

[0101] The competition model is a machine learning model created for the target objects under the current competition's theme category.

[0102] S303. Obtain the model parameters of the adjusted competition model, as well as the computing resource call data of the adjusted competition model.

[0103] S304. Input the model parameters and computing resource call data into the trained target prediction model, and obtain the prediction data output by the trained target prediction model.

[0104] The trained target prediction model is used to extract features from the model parameters and computing resource call data respectively. The features extracted from the model parameters and computing resource call data are weighted and merged, and the merged features are used to generate prediction data.

[0105] S305. Based on the predicted data and the target object's historical participation data, generate the target object's event information for the current event.

[0106] In this application, the target object and the server can interact to directly create and adjust the competition model on the server platform. After the initial completion of the competition model, the target object can use the target prediction model to perform model evaluation and obtain the model's prediction data on the test set. Subsequently, based on the performance of the model function specified for the current competition as reflected in the prediction data, the target object can continue to adjust the competition model. After each adjustment, the result submission function can be used to update the competition model. Throughout the competition, the target object can view the competition information at any time, understand its current ranking position, and decide whether further optimization of the competition model is needed.

[0107] Figure 4 A flowchart illustrating another method for generating event information based on a large model, as provided in this application embodiment, is shown below. Figure 4 As shown, this method for generating event information based on a large model includes S401-S411:

[0108] S401. Receive the entry request from the target object, which includes the entry information of the target object.

[0109] S402. By identifying preset keywords in the target participant's competition materials, determine the target participant's historical performance data.

[0110] S403. Based on the target object's historical performance data, grant the target object the right to participate in the current competition. The right to participate is used to allow the target object to send model processing instructions through the terminal device.

[0111] S404: Receive model processing instructions sent by the terminal device.

[0112] The model processing instruction includes the identification information of the target object that triggered the model processing instruction and at least one model operation indicated by the model processing instruction.

[0113] S405. Perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object to obtain the adjusted event model.

[0114] The competition model is a machine learning model created for the target objects under the current competition's theme category.

[0115] S406. Obtain the model parameters of the adjusted competition model, as well as the computing resource call data of the adjusted competition model.

[0116] S407. Input the model parameters and computing resource call data into the trained target prediction model, and obtain the prediction data output by the trained target prediction model.

[0117] The trained target prediction model is used to extract features from the model parameters and computing resource call data respectively. The features extracted from the model parameters and computing resource call data are weighted and merged, and the merged features are used to generate prediction data.

[0118] S408. Perform preprocessing operations on the predicted data and the historical competition data of the target object.

[0119] The preprocessing operations include at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

[0120] S409. Based on the predicted data and the target object's historical participation data, generate the target object's event information for the current event.

[0121] S410: Obtain the ranking data of each object participating in the current event.

[0122] S411. Based on the preset sorting method, use bubble sort or insertion sort to insert the event information of the target object in the current event into the event ranking data of each object in the current event, and obtain the updated ranking data.

[0123] The competition information generation method based on large models provided in this application solves problems of low resource utilization, high cost, security risks, and collaborative efficiency through heterogeneous computing power scheduling, training optimization, security enhancement, and real-time interaction, thus promoting the evolution of artificial intelligence competitions from experimental exploration to industrial application. Furthermore, it can predict contestant rankings through a predictive model and display them to participants. Participants can also adjust the parameters of their competition models to predict their own rankings and use this information to adjust their own model parameters and data, significantly improving the information management process of the competition and enhancing the experience for both participants and administrators.

[0124] The event information generation method based on a large model provided in this application receives a model processing instruction sent by a terminal device. The model processing instruction includes the identification information of the target object that triggered the model processing instruction and at least one model operation indicated by the model processing instruction. At least one model operation is performed on the event model to be adjusted corresponding to the identification information of the target object to obtain an adjusted event model. The event model is a machine learning model created for the target object under the theme category of the current event. A trained target prediction model is used to predict the adjusted event model to obtain prediction data of the adjusted event model. The prediction data characterizes the predictive performance of the model function specified for the current event. Based on the prediction data and the historical participation data of the target object, event information of the target object under the current event is generated. This method can directly execute model operations on the event model to be adjusted based on the model processing instruction, and directly predict the performance of the model function specified for the current event through the trained target prediction model, thereby obtaining event data. Since manual testing of each model is not required, the speed of event information generation can be improved.

[0125] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0126] Based on the same inventive concept, this application also provides a large-model-based event information generation device for implementing the above-mentioned large-model-based event information generation method. The solution provided by this device is similar to the implementation described in the above method. Therefore, the specific limitations of one or more large-model-based event information generation device embodiments provided below can be found in the limitations of the large-model-based event information generation method described above, and will not be repeated here.

[0127] In one exemplary embodiment, such as Figure 5 As shown, a large-model-based event information generation device 500 is provided, including: a receiving module 501, an execution module 502, a prediction module 503, and a generation module 504, wherein:

[0128] The receiving module 501 is used to receive model processing instructions sent by the terminal device. The model processing instructions include identification information of the target object that triggers the model processing instructions and at least one model operation indicated by the model processing instructions.

[0129] The execution module 502 is used to perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object, so as to obtain the adjusted event model. The event model is a machine learning model created for the target object under the theme category of the current event.

[0130] The prediction module 503 is used to make predictions on the adjusted event model using the trained target prediction model, and obtain the prediction data of the adjusted event model. The prediction data represents the prediction performance of the model function specified for the current event.

[0131] The generation module 504 is used to generate the event information of the target object in the current event based on the prediction data and the historical participation data of the target object.

[0132] In one embodiment, the prediction module 503 is further configured to acquire the model parameters of the adjusted competition model and the computing resource call data of the adjusted competition model; input the model parameters and computing resource call data into the trained target prediction model, and acquire the prediction data output by the trained target prediction model.

[0133] The trained target prediction model is used to extract features from the model parameters and computing resource call data respectively. The features extracted from the model parameters and computing resource call data are weighted and merged, and the merged features are used to generate prediction data.

[0134] In one embodiment, the generation module 504 is further configured to obtain the ranking data of each object participating in the current event; and according to a preset sorting method, insert the event information of the target object in the current event into the ranking data of each object in the current event to obtain the updated ranking data.

[0135] In one embodiment, the generation module 504 is further configured to use a sorting method such as bubble sort or insertion sort to insert the event information of the target object in the current event into the event ranking data of each object in the current event.

[0136] In one embodiment, the generation module 504 is further configured to perform preprocessing operations on the prediction data and the historical competition data of the target object. The preprocessing operations include at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

[0137] In one embodiment, the receiving module 501 is further configured to receive a participation request from the target object, the participation request including the target object's participation information.

[0138] The execution module 502 is used to identify the target object's historical performance data by recognizing preset keywords in the target object's competition data; based on the target object's historical performance data, it grants the target object the right to participate in the current competition, which allows the target object to send model processing instructions through the terminal device.

[0139] Each module in the aforementioned large-scale model-based event information generation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0140] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 6 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When the computer program is executed by the processor, it implements a method for generating event information based on a large model.

[0141] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0142] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for generating event information based on a large model.

[0143] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for generating event information based on a large model.

[0144] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described method for generating event information based on a large model.

[0145] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this 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. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0146] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0147] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for generating event information based on a large model, characterized in that, The method includes: The device receives a model processing instruction sent by a terminal device, the model processing instruction including identification information of the target object that triggers the model processing instruction and at least one model operation indicated by the model processing instruction; Perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object to obtain the adjusted event model, wherein the event model is a machine learning model created for the target object under the theme category of the current event; Using the trained target prediction model, the adjusted event model is used to make predictions, and the prediction data of the adjusted event model is obtained. The prediction data represents the prediction performance of the model function specified for the current event. Based on the predicted data and the target object's historical participation data, generate the target object's event information for the current event.

2. The method according to claim 1, characterized in that, The step of using the trained target prediction model to predict the adjusted event model and obtaining the prediction data of the adjusted event model includes: Obtain the model parameters of the adjusted competition model, as well as the computing resource call data of the adjusted competition model; The model parameters and the computing power resource call data are input into the trained target prediction model, and the prediction data output by the trained target prediction model is obtained. The trained target prediction model is used to extract features from the model parameters and the computing resource call data, respectively, and to perform weighted merging of the features extracted from the model parameters and the computing resource call data, and to generate the prediction data using the merged features.

3. The method according to claim 1, characterized in that, After generating the target object's competition information for the current competition based on the predicted data and the target object's historical competition data, the process includes: Obtain the ranking data of each object participating in the current event; According to a preset sorting method, the event information of the target object in the current event is inserted into the event ranking data of each object in the current event to obtain the updated ranking data.

4. The method according to claim 3, characterized in that, The step of inserting the event information of the target object into the event ranking data of each object in the current event includes: Using either bubble sort or insertion sort, the event information of the target object in the current event is inserted into the event ranking data of each object in the current event.

5. The method according to claim 1, characterized in that, Before generating the target object's competition information for the current competition based on the predicted data and the target object's historical competition data, the method further includes: The predicted data and the historical competition data of the target object are preprocessed. The preprocessing operations include at least one of the following: data cleaning, data format conversion, handling of missing data values, and identification of outliers.

6. The method according to any one of claims 1-5, characterized in that, Before receiving the model processing instructions sent by the receiving terminal device, the method includes: Receive the entry request from the target object, the entry request including the entry information of the target object; By identifying preset keywords in the target candidate's competition materials, the target candidate's historical performance data can be determined; Based on the target object's historical performance data, grant the target object participation permission for the current competition. This participation permission allows the target object to send the model processing instructions through the terminal device.

7. A device for generating event information based on a large model, characterized in that, The device includes: A receiving module is used to receive a model processing instruction sent by a terminal device. The model processing instruction includes identification information of the target object that triggers the model processing instruction and at least one model operation indicated by the model processing instruction. An execution module is used to perform at least one model operation on the event model to be adjusted corresponding to the identification information of the target object, so as to obtain an adjusted event model, wherein the event model is a machine learning model created for the target object under the theme category of the current event; The prediction module is used to use a trained target prediction model to predict the adjusted event model and obtain the prediction data of the adjusted event model. The prediction data represents the prediction performance of the model function specified for the current event. The generation module is used to generate the event information of the target object in the current event based on the predicted data and the historical participation data of the target object.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements 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, it implements the steps of the method according to any one of claims 1 to 6.

Images