Data processing system and method and electronic equipment

Abstract

The invention discloses a data processing system and method and electronic equipment. The data processing system is used for executing a target task. The data processing system comprises a plurality of application programs, and different application programs are used for executing different subtasks. Each application program has the capability of utilizing the target model, and the number of trainable parameters of the target model is greater than 10 billion; a logic for executing the target task is determined among the multiple application programs in a session mode.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a data processing system, method and electronic device. Background Art

[0002] At present, when using multiple models to perform tasks, it is necessary to set a fixed process for the multiple models according to the current application scenario, so that the multiple models can complete the tasks in the application scenario according to the fixed process. Once the multiple models do not follow the fixed process, they cannot perform the tasks, which leads to greater limitations in the execution of tasks by multiple models. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a data processing system, method and electronic device.

[0004] In a first aspect, an embodiment of the present application provides a data processing system for executing a target task;

[0005] The data processing system includes a plurality of applications, where different applications are used to perform different subtasks; each application has the ability to utilize a target model, where the number of trainable parameters of the target model is greater than 100 million;

[0006] Multiple applications determine the logic for executing the target task in a conversational manner.

[0007] In a possible implementation, any application executes a corresponding subtask, generates a session content that can represent a subresult, and outputs the session content to a target environment; the target environment is a running environment of multiple applications, and is used to establish a session connection between the multiple applications;

[0008] Other applications can obtain the sub-result from the target environment based on the session connection. If the sub-result meets the target condition, the other applications execute corresponding sub-tasks, generate session content that can represent the sub-result, and output the session content to the target environment;

[0009] If the sub-result does not meet the target condition, other application programs do not execute the corresponding sub-task; different application programs may correspond to the same or different target conditions.

[0010] In a possible implementation manner, the conversation content generated by any application is in natural language, and / or the data processing system includes a target interface, and the target interface is used to display the conversation content of the application.

[0011] In a possible implementation, the target model includes a plurality of sub-models;

[0012] Each application program can call a sub-model to execute the sub-task according to the sub-task it executes.

[0013] In a second aspect, the embodiment of the present application further provides a data processing method, including:

[0014] Obtaining a target instruction; the target instruction is used to enable the electronic device to perform a target task;

[0015] Determine a target task group; the target task group includes a plurality of applications, and different applications are used to perform different subtasks; each application has the ability to utilize a target model, and the number of trainable parameters of the target model is greater than 100 million;

[0016] A target environment is provided so that multiple application programs in the target task group can run in the target environment, and the logic for executing the target task is determined between the multiple application programs in a conversational manner.

[0017] In a third aspect, an embodiment of the present application further provides an electronic device, including:

[0018] The memory is used to store computer programs;

[0019] The plurality of processors are used to execute the computer program so that the electronic device can implement the following steps:

[0020] Obtaining a target instruction; the target instruction is used to enable the electronic device to perform a target task;

[0021] Determine a target task group; the target task group includes a plurality of applications, and different applications are used to perform different subtasks; each application has the ability to utilize a target model, and the number of trainable parameters of the target model is greater than 100 million;

[0022] A target environment is provided so that multiple application programs in the target task group can run in the target environment, and the logic for executing the target task is determined between the multiple application programs in a conversational manner.

[0023] In a possible implementation, the processor obtains a target task for diagnosing the electronic device, where the target task group includes at least a first application, a second application, a third application, and a fourth application;

[0024] The first application uses the first model to detect whether there is a problem with the electronic device, generates a first session content and sends it to the target environment, wherein the first session content includes the detection result;

[0025] The second application classifies the problems existing in the electronic device using the second model, generates second conversation content and sends it to the target environment, wherein the second conversation content includes the classification result;

[0026] A third application uses a third model to repair the problem in the electronic device, generates a third session content and sends it to the target environment, wherein the third session content includes a repair result;

[0027] A fourth application program evaluates the repaired problem data using a fourth model, generates fourth session content, and sends the fourth session content to the target environment, wherein the fourth session content includes an evaluation result;

[0028] The first application, the second application, the third application, and the fourth application can all obtain the first session content, the second session content, the third session content, and the fourth session content from the target environment.

[0029] In one possible implementation,

[0030] The second application classifies the problem existing in the electronic device using a second model when the first conversation content satisfies a first condition, wherein the first condition is that the detection result included in the first conversation content indicates that there is a problem in the electronic device;

[0031] The third application uses the third model to repair the problem existing in the electronic device when the second conversation content meets the second condition, and the second condition is that the classification result included in the second conversation content corresponds to the third model;

[0032] The fourth application uses a fourth model to evaluate the repaired problem data when the third conversation content meets a third condition, and the third condition is that the repair result included in the third conversation content indicates that the problem in the electronic device has been repaired.

[0033] In one possible implementation,

[0034] When the fourth conversation content satisfies a fourth condition, the third application uses the third model to re-repair the problem in the electronic device, or connects to other applications to repair the problem in the electronic device based on the other applications, and the fourth condition is that the evaluation result included in the fourth conversation content indicates that the problem has not been successfully repaired.

[0035] In a possible implementation manner, the electronic device further includes a target interface;

[0036] The target interface displays at least the first conversation content, the second conversation content, the third conversation content, and the fourth conversation content.

[0037] The data processing system of the embodiment of the present application includes multiple applications that determine the logic of executing the target task in a conversational manner, without the need to set a fixed process for them, thereby reducing the limitations of multiple applications in executing tasks. In addition, the system can be applied to different application scenarios, has high flexibility and wide applicability, and has certain scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 A flow chart of a data processing method provided by the present application is shown;

[0040] Figure 2 A schematic diagram of an electronic device provided by the present application is shown;

[0041] Figure 3 A schematic diagram of the structure of an electronic device provided by the present application for use in problem diagnosis and repair is shown. DETAILED DESCRIPTION

[0042] Various aspects and features of the present application are described herein with reference to the accompanying drawings.

[0043] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but only as an example of an embodiment. Other modifications within the scope and spirit of the present application will occur to those skilled in the art.

[0044] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0045] These and other characteristics of the present application will become apparent from the following description of a preferred form of embodiment given as a non-limiting example with reference to the accompanying drawings.

[0046] It should also be understood that, although the present application has been described with reference to some specific examples, those skilled in the art will be able to realize many other equivalent forms of the present application that have the features described in the claims and are therefore within the scope of protection defined thereby.

[0047] The above and other aspects, features and advantages of the present application will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.

[0048] Specific embodiments of the present application are described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments applied for are merely examples of the present application, which may be implemented in a variety of ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that obscure the present application. Therefore, the specific structural and functional details applied for herein are not intended to be limiting, but merely serve as a basis and representative basis for the claims to teach those skilled in the art to use the present application in a variety of ways with substantially any suitable detailed structure.

[0049] This specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," all of which may refer to one or more of the same or different embodiments according to the present application.

[0050] To facilitate understanding of the present application, a data processing system provided by the present application is first introduced in detail.

[0051] The data processing system provided in the embodiment of the present application is used to execute a target task, which may be a task in different scenarios or a task for different execution objects.

[0052] Furthermore, the data processing system includes multiple applications, and different applications are used to execute different subtasks, so that the purpose of executing the target task can be achieved by executing different subtasks by different applications. In addition, the logic of executing the target task is determined by the multiple applications in a conversational manner, that is, each application can complete the task independently, and its call flow is not fixed, that is, multiple applications are not limited to a fixed process, and the flexibility is high.

[0053] In actual applications, an application can execute multiple different subtasks or execute the same subtask multiple times. The embodiments of the present application do not specifically limit this. When executing the target task, you can judge and execute it according to actual needs. For example, the subtask executed by application A can be writing a paper or checking a paper, etc. For another example, the subtask executed by application B is to repair a fault. If the fault is not repaired successfully, application B can execute the subtask of repairing the fault again. This makes multiple applications have a certain degree of scalability.

[0054] Specifically, each application has the ability to use the target model. When executing a subtask, each application uses the corresponding target model to execute the subtask, thereby achieving the purpose of each application autonomously executing the subtask. Moreover, for different application scenarios, each application can use different target models to execute subtasks, thereby making the multiple applications applicable to different application scenarios, that is, the application is relatively wide. For example, each application is an agent application, and the agent application can be implemented based on the target model. Its behavior is determined by the model according to the current state of the agent and external input, and can be specifically used to solve a certain type of problem.

[0055] Among them, the number of trainable parameters of the target model is greater than 100 million. The target model is trained with a large number of parameters so that the target model has better performance, so as to achieve the purpose of the application autonomously executing subtasks and the purpose of the application and other applications autonomously deciding the logic of executing the target task. The target model can be a variety of generative models, which can be used to process natural language (NL) content and / or other inputs to generate outputs that reflect the generated content in response to the input. For example, large language models (LLM), small language models, etc. Large language models learn the statistical characteristics and laws of language by training a large amount of text data, so that they can understand and generate natural language text. Large language models usually have parameters ranging from billions to hundreds of billions, which can capture complex relationships and patterns in language text.

[0056] In actual applications, any application program that executes a corresponding subtask generates a session content that can represent a subresult, wherein the subresult is the result obtained by the application program executing the subtask. That is, for each application program, after the application program executes a corresponding subtask, the session content is generated based on the subresult obtained by executing the subtask. Here, the session content may be obtained by converting the subresult into a preset format, or the session content may be generated based on the subresult according to preset requirements, etc.

[0057] After generating the session content, each application outputs the session content to the target environment, wherein the target environment is the operating environment of multiple applications, and the target environment is used to establish a session connection between the multiple applications. That is, in the target environment, any two applications of the multiple applications can establish a session connection.

[0058] After an application generates session content and outputs the session content to the target environment, other applications can obtain sub-results from the target environment based on the session connection. Here, the application can output the session content to the target environment in the form of broadcasting, so that other applications in the target environment can obtain the session content, that is, the sub-result, and realize the session connection between multiple applications; the application can also store the session content in a specific location in the target environment, and the specific location is used to store the session content generated by each application, and the specific location can be accessed by each application, that is, each application can obtain the session content generated by other applications, that is, the sub-results obtained by other applications executing subtasks, from the specific location, thereby realizing the session connection between multiple applications.

[0059] For each application among other applications, after the application obtains the sub-result from the target environment, it is further determined whether the sub-result meets the target condition, wherein the target condition is related to the application and / or the sub-task corresponding to the application, such as whether the target condition is used to characterize whether the application is idle, whether the task to be performed to improve the sub-result is the sub-task corresponding to the application, or whether the sub-result meets the task requirements, etc. If different applications perform different sub-tasks, the target conditions corresponding to different applications are different accordingly.

[0060] Optionally, if the sub-result meets the target condition, at this time, the other application executes the corresponding sub-task, generates a session content that can represent the sub-result, and outputs the session content to the target environment. Here, when the sub-result meets the target condition, the other application that executes the corresponding sub-task can be one or more, and the other application can independently determine whether to execute the sub-task according to the sub-result.

[0061] Accordingly, if the sub-result does not satisfy the target condition, other applications do not execute the corresponding sub-task.

[0062] In the embodiment of the present application, multiple applications can obtain session content generated by other applications in a conversational manner, and dynamically determine the next application to execute a subtask based on the session content, which has high flexibility.

[0063] Optionally, the conversation content generated by any application is in natural language, so that multiple applications can work together as a team to perform target tasks, and multiple rounds of component dialogues can be conducted through the target environment, that is, each application can obtain the conversation content generated by other applications, so that multiple applications in the target environment can perform different target tasks, which has a wide range of applications.

[0064] As another example, the data processing system of the embodiment of the present application includes a target interface, which is used to display the conversation content of the application, that is, to visualize the conversation between multiple applications, so that the user can obtain the conversation process and conversation content between multiple applications in real time, and then can know the sub-results generated by each application executing the subtask in real time, which greatly improves the user experience. At the same time, if the final target task fails to execute, the conversation between the applications can also be traced back to find the cause of the failure. In addition, the conversation between the applications can also be monitored in real time. When a processing method that does not meet the user's expectations appears, the user intervenes in the execution logic of the application based on the interface control to guide the application to perform the task in accordance with the user's expectations.

[0065] The target model in the embodiment of the present application includes multiple sub-models, and different sub-models have different model parameters, and can perform different sub-tasks. In a specific implementation, each application can call a sub-model to perform the sub-task according to the sub-task it performs. For example, each sub-model is provided with identification information, and the identification information is used to characterize the field, input information, and output information to which the sub-model belongs. The application calls the sub-model corresponding to the sub-task to perform the sub-task according to the sub-task it performs and the mapping relationship between the sub-task and the sub-model.

[0066] The data processing system of the embodiment of the present application includes multiple applications that determine the logic of executing the target task in a conversational manner, without the need to set a fixed process for them, thereby reducing the limitations of multiple applications in executing tasks. In addition, the system can be applied to different application scenarios, has high flexibility and wide applicability, and has certain scalability.

[0067] Embodiments of the present application Figure 1 A flow chart of a data processing method provided in an embodiment of the present application is shown, wherein specific steps include S101-S103.

[0068] S101, obtaining a target instruction; the target instruction is used to enable the electronic device to perform a target task.

[0069] S102, determining a target task group; the target task group includes multiple applications, and different applications are used to perform different subtasks; each application has the ability to utilize a target model, and the number of trainable parameters of the target model is greater than 100 million.

[0070] S103, providing a target environment so that multiple application programs in the target task group can run in the target environment, and the multiple application programs determine the logic of executing the target task in a conversation manner.

[0071] In the embodiment of the present application, the data processing system of the electronic device is used as the execution subject. Specifically, it obtains the target instruction in real time, and the target instruction is used to make the electronic device perform the target task. Among them, the target instruction can be automatically triggered and generated. For example, in the scenario where the target task is to monitor the problem of the electronic device, the target instruction can be generated when the electronic device is started.

[0072] After obtaining the target instruction, a target task group is determined according to the target instruction, so as to use the target task group to execute the target task. The target task group includes a plurality of application programs, which may be obtained by screening from an application set according to the target task, and different application programs are used to execute different subtasks, and the number of application programs in the target task group is determined by the target task, and the number of application programs in the target task group corresponding to different target tasks is different.

[0073] Among them, each application has the ability to utilize the target model. Specifically, the target model includes multiple sub-models. Each application can call a sub-model to perform the sub-task according to the sub-task it performs, and the number of trainable parameters of the target model is greater than 100 million.

[0074] A target environment is provided while determining the target task group, so that multiple applications in the target task group can run in the target environment, and in the target environment, the logic of executing the target task is determined between the multiple applications in a conversational manner.

[0075] Specifically, after executing a subtask, any application generates conversation content that can represent the sub-result, and the conversation content is in natural language, so that multiple applications in the target task group can collaboratively execute the target task, and multiple applications can conduct multiple rounds of conversations within the target task group, so that each application can independently determine whether to execute the subtask without setting a fixed process for it, thereby reducing the limitations of multiple applications in executing tasks.

[0076] Embodiments of the present application Figure 2 A schematic diagram of an electronic device provided in an embodiment of the present application is shown, referring to Figure 2 It can be seen that the electronic device includes a memory, which is used to store computer programs and provide the stored computer programs to other devices such as a processor.

[0077] The electronic device also includes multiple processors, which are used to execute computer programs so that the electronic device can implement the following steps one to three.

[0078] Step 1: Obtain target instructions; target instructions are used to enable the electronic device to perform target tasks.

[0079] Step 2: Determine the target task group; the target task group includes multiple applications, and different applications are used to perform different subtasks; each application has the ability to utilize the target model, and the number of trainable parameters of the target model is greater than 100 million.

[0080] Step 3: Provide a target environment so that multiple applications in the target task group can run in the target environment, and multiple applications determine the logic of executing the target task in a conversational manner.

[0081] Taking the application scenario of electronic devices being used for problem diagnosis and repair as an example, the processor obtains the target task of diagnosing the electronic device, and the target task group includes at least the first application, the second application, the third application, and the fourth application. Figure 3 It is a schematic diagram of the structure of an electronic device used in application scenarios of problem diagnosis and repair.

[0082] In a specific implementation, when the electronic device is started, the target instruction is generated, and accordingly, the processor starts to execute the above-mentioned computer program, that is, obtains the target instruction.

[0083] After obtaining the target instruction, the target task group is determined according to the target task represented by the target instruction, that is, multiple applications are screened out according to the target task to form the target task group. Among them, the target task group in the scenario of the embodiment of the present application includes at least a first application, a second application, a third application and a fourth application. The first application, the second application, the third application and the fourth application all have the ability to utilize the target model, and can all run in the target environment, and then determine the logic of executing the target task in a conversational manner based on the target environment. Since the electronic device in this embodiment provides an operating environment for applications that have the ability to utilize the target model, that is, the above-mentioned application is running on the electronic device, the above-mentioned electronic device can be an AIPC. AIPC is a personal computer that integrates artificial intelligence technology. It integrates hardware such as NPU, CPU, GPU, etc., while achieving high energy and low consumption, fundamentally changing, reshaping and reconstructing the PC experience.

[0084] Specifically, for the application scenario of problem diagnosis and repair in AIPC, the target tasks include at least detecting problems in electronic devices, repairing problems, etc. Based on this, the first application uses the first model to detect whether there is a problem with the electronic device, generates a first session content and sends it to the target environment, and the first session content includes the detection result. Optionally, the first application can monitor and obtain the current information of the electronic device in real time, and the current information at least includes the operation log, hardware status, software status and user operation events of the electronic device. Among them, the monitoring frequency and monitoring indicators are pre-configured for the first application, and a reasonable monitoring frequency (such as once per second, once per minute, etc.) is set according to the needs; then the operation log is monitored according to the monitoring frequency. Here, key information can be extracted from the operation log to reduce the calculation amount of the first model. The extracted key information may include error messages, warnings, event records, etc.; at the same time, the user's operation information on the electronic device is monitored according to the monitoring frequency to extract user operation events and error reports, complaint feedback, etc.

[0085] A specific example of configuring the first application is as follows:

[0086] You are a problem monitoring application. Your task is to monitor system logs, hardware status, system performance indicators, and user operation events in real time. Parse the following system log files, identify abnormal patterns or error warnings, determine the abnormal type and cause, classify and archive, and generate a preliminary diagnostic report. The output format follows the example.

[0087] System log:

[0088] [2024-05-21 16:31:19]Error:Device driver failure detected.

[0089] [2024-05-21 16:32:00]Warning:High CPU usage detected.

[0090] [2024-05-21 16:33:45]Info:User opened application XYZ.

[0091] Please extract key information and generate a preliminary diagnostic report.

[0092] Output example:

[0093] Preliminary diagnostic report:

[0094] 1. Exception type: Device driver failure

[0095] Possible cause: The device driver is damaged or incompatible

[0096] Impact: May cause system to run slowly or equipment to not work properly

[0097] 2. Abnormal type: High CPU usage

[0098] Possible cause: An application is taking up too much CPU resources

[0099] Impact: May cause slow system response

[0100] 3. User action event: User opens application XYZ

[0101] The preliminary diagnosis report corresponds to the first session content generated by the first application.

[0102] Afterwards, the current information of the electronic device is input into the first model so that the first model processes the current information of the electronic device to obtain a detection result, which may include whether there is a problem with the electronic device, and if there is a problem with the electronic device, the detection result may also include the content of the problem, the possible cause of the problem, and the scope of influence of the problem.

[0103] After the first model obtains the detection result, the first application generates the first session content according to the detection result and sends it to the target environment. Optionally, the first model uses natural language processing (NLP) technology to parse the operation log to identify abnormal patterns or error warnings in the operation log. Of course, the embodiments of the present application are not limited to this.

[0104] After the first application sends the first conversation content to the target environment, the second application, the third application, and the fourth application can obtain the first conversation content and analyze the first conversation content to determine whether to execute their own subtasks. In the embodiment of the present application, the second application can trigger the execution of the subtask based on the first conversation content of the first application, and the detection result included in the first conversation content indicates that there is a problem with the electronic device.

[0105] In a specific implementation, the second application classifies the problems existing in the electronic device using the second model, generates and sends the second session content to the target environment, and the second session content includes the classification result. Optionally, when the second application determines that the first session content satisfies the first condition, it classifies the problems existing in the electronic device using the second model, wherein the first condition is that the detection result included in the first session content indicates that there is a problem in the electronic device.

[0106] Specifically, after the second application obtains the content of the first conversation and determines that the detection results included in the content of the first conversation indicate that there is a problem with the electronic device, the second application calls the trained second model to classify the problems in the electronic device. Among them, at least prompt engineering, RAG, and fine turning are used to train the second model, and in actual training, the comprehension of the second model is set through prompt rules, including necessary sample codes, and emphasizing that the large model returns a specified format; and the function and the passed parameters of the function are described in detail through prompt, so that the second model can accurately match the corresponding function; and for the unmatched function, prompt is used to encourage AI to autonomously generate executable function code to improve the intelligence of model training.

[0107] An example of configuring the second application is shown below:

[0108] You are a classification application. Your task is to classify the problems in the monitored electronic equipment and generate instructions in a unified format. According to the following preliminary diagnosis report, classify the problems and generate corresponding function call instructions. Note that the return function call must be output in the format.

[0109] Preliminary diagnostic report:

[0110] 1. Exception type: Device driver failure

[0111] Possible cause: The device driver is damaged or incompatible

[0112] Impact: May cause system to run slowly or equipment to not work properly

[0113] 2. Abnormal type: High CPU usage

[0114] Possible cause: An application is taking up too much CPU resources

[0115] Impact: May cause slow system response

[0116] Please generate the corresponding function call instruction.

[0117] Output example:

[0118] Function Call instruction:

[0119] 1.Function:diagnose_device_driver_issue

[0120] Parameters:{"error_type":"Device driver failure","possible_cause":"Driver corruption or incompatibility"}

[0121] 2.Function:diagnose_high_cpu_usage

[0122] Parameters:{"error_type":"High CPU usage","possible_cause":"Application XYZ consuming excessive CPU resources"}

[0123] After the second application is configured and the trained second model is obtained, the detection result included in the first session content is input into the second model so that the second model outputs the classification result. After that, the second application generates the second session content based on the classification result and sends it to the target environment. Of course, the second session content includes the classification result.

[0124] After the second application sends the second conversation content to the target environment, the first application, the third application, and the fourth application can obtain the second conversation content and analyze the second conversation content to determine whether to execute their own subtasks. In the embodiment of the present application, the third application can trigger the execution of the subtask based on the second conversation content of the second application.

[0125] In a specific implementation, the third application uses the third model to repair the problem in the electronic device, generates the third session content and sends it to the target environment, and the third session content includes the repair result. The third application uses the third model to repair the problem in the electronic device when the second session content meets the second condition, and the second condition is that the classification result included in the second session content corresponds to the third model, and the third model is a repair model, which includes multiple repair sub-models.

[0126] Specifically, different repair sub-models are set for different types of problems. After obtaining the second session content, the third application can call the repair sub-model corresponding to the classification result according to the classification result, and then use the repair sub-model to repair the problem in the electronic device. Among them, the third model has the function call code execution and code generation capabilities, among which a model specifically for code optimization such as code llama can be set to achieve code execution and code generation capabilities.

[0127] An example of configuration for a third application is shown below:

[0128] You are a repair application. Your task is to execute the function call instruction and repair the problem. According to the following function call instruction, perform the repair operation and feedback the execution result in real time. If the execution fails, you can try to search the Internet, find a solution, and generate executable code for repair.

[0129] Function Call instruction:

[0130] 1.Function:diagnose_device_driver_issue

[0131] Parameters:{"error_type":"Device driver failure","possible_cause":"Driver corruption or incompatibility"}

[0132] 2.Function:diagnose_high_cpu_usage

[0133] Parameters:{"error_type":"High CPU usage","possible_cause":"Application XYZ consuming excessive CPU resources"}

[0134] Please perform the repair operation and report back with the results.

[0135] Output example:

[0136] Repair operation results:

[0137] 1.Function:diagnose_device_driver_issue

[0138] Execution Result:Device driver reinstalled successfully.

[0139] 2.Function:diagnose_high_cpu_usage

[0140] Execution Result:Application XYZ terminated,CPU usage returned to normal

[0141] After the third application is configured and the third model is retrieved based on the classification result, the repair code is generated and executed using the third model to repair the problem. Then, the third application generates the third session content based on the repair result and sends it to the target environment.

[0142] After the third application sends the third conversation content to the target environment, the first application, the second application, and the fourth application can all obtain the third conversation content and analyze the third conversation content to determine whether to execute their own subtasks. In the embodiment of the present application, the fourth application can trigger the execution of a subtask based on the third conversation content of the third application, and the repair result included in the third conversation content is used to indicate whether the problem of the electronic device has been repaired.

[0143] In a specific implementation, the fourth application uses the fourth model to evaluate the repaired problem data, generates a fourth session content and sends it to the target environment, and the fourth session content includes the evaluation result. The fourth application requires that the third session content meets the third condition, and then uses the fourth model to evaluate the repaired problem data, and the third condition is that the repair result included in the third session content indicates that the problem in the electronic device has been repaired.

[0144] Specifically, after obtaining the content of the third session, the fourth application determines whether the problem in the electronic device has been repaired according to the repair result included in the third session content. If so, it triggers the execution of the subtask, that is, calling the fourth model to evaluate the repaired problem data. Specifically, the operation log, hardware status, software status and user operation events of the mobile phone after repair are rechecked to determine whether various indicators have returned to normal, such as whether the memory is no longer leaked, whether the network is unobstructed, and whether certain rule unit tests are successful, etc.; and the validity and integrity of the repair data are analyzed by combining the fourth model with context information; and user feedback is collected to evaluate the repair results.

[0145] An example of configuration for the fourth application is shown below:

[0146] You are an evaluation application. Your task is to evaluate and verify the problem data after repair. According to the following repair results, re-collect system logs, detect system status, and generate an acceptance report.

[0147] Repair operation results:

[0148] 1.Function:diagnose_device_driver_issue

[0149] Execution Result:Device driver reinstalled successfully.

[0150] 2.Function:diagnose_high_cpu_usage

[0151] Execution Result:Application XYZ terminated,CPU usage returned to normal.

[0152] Please generate an acceptance report.

[0153] Output example:

[0154] Acceptance Report:

[0155] 1. Repair result: The device driver was reinstalled successfully

[0156] System Status: The device is operating normally, no further errors reported

[0157] 2. Repair result: Application XYZ has been terminated and CPU usage has returned to normal

[0158] System status: System response speed is normal, no high CPU usage reported

[0159] Overall assessment: All issues have been successfully fixed and the system is operating normally

[0160] After completing the configuration of the fourth application, the fourth application collects the corresponding repaired problem data, evaluates the repaired problem data using the fourth model, and outputs the evaluation result, and the fourth application generates the fourth session content based on the evaluation result.

[0161] Of course, the fourth application also sends the fourth session content to the target environment, and the first application, the second application, and the third application can all obtain the fourth session content.

[0162] In another example, when the fourth session content meets the fourth condition, the third application uses the third model to re-repair the problem in the electronic device, or connects to other applications to repair the problem in the electronic device based on other applications, and the fourth condition is that the evaluation result included in the fourth session content indicates that the problem was not successfully repaired.

[0163] That is to say, there is a situation where the problem is not successfully repaired. Therefore, after the third application is set to obtain the fourth session content, the evaluation result included in the fourth session content is determined. If the evaluation result indicates that the problem is not successfully repaired, the fourth session content satisfies the fourth condition. At this time, the third application uses the third model to repair the problem in the electronic device again, or connects to other applications to repair the problem in the electronic device based on other applications. By connecting to other applications, the third application can have a higher repair capability, and can also continuously optimize its own performance without manual optimization, effectively reducing maintenance costs and maintenance time.

[0164] The electronic device of the embodiment of the present application includes a first application, a second application, a third application, and a fourth application that can conduct multiple rounds of dialogues, so that they can all independently determine whether to execute subtasks without setting a fixed process for them, thereby reducing the limitations of multiple applications in executing tasks.

[0165] It is worth noting that the first model, the second model, the third model and the fourth model are pre-trained, and the number of trainable parameters is greater than 100 million. Before using the training parameters to train each model, a large amount of text data is collected, such as system logs, error reports submitted by users, etc., and then data cleaning is performed on the large amount of text data to remove irrelevant or redundant information, thereby ensuring data quality and format consistency.

[0166] Since different text data correspond to different data sources, and thus different text data may have different formats, before using the training parameters to train each model, all text data need to be converted into a unified preset format to avoid interference with model calculations caused by different data formats.

[0167] Afterwards, the first model, the second model, the third model, and the fourth model are trained respectively using a large number of parameters in a unified format. Here, since model fine-tuning requires more computing power and time, it is not possible to quickly iterate and update in some scenarios. At this time, RAG is used to increase the knowledge base density of the model to cover the blind spots that still exist after model fine-tuning, and it can be quickly deployed and iterated in the form of documents or databases, effectively improving the training efficiency and accuracy of the model.

[0168] As another example, the electronic device of the embodiment of the present application also includes a target interface, which displays at least the first conversation content, the second conversation content, the third conversation content and the fourth conversation content, so that the user can clearly and conveniently understand the execution progress of the target task.

[0169] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present application with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. The elements in the claims will be interpreted broadly based on the language adopted in the claims, and are not limited to the examples described in this specification or during the implementation of the application, and the examples will be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered as examples only, and the true scope and spirit are indicated by the following claims and the full scope of their equivalents.

[0170] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. For example, those of ordinary skill in the art can use other embodiments when reading the above description. In addition, in the above-mentioned specific embodiments, various features can be grouped together to simplify the application. This should not be interpreted as an intention that a disclosed feature that does not require protection is necessary for any claim. On the contrary, the subject matter of the present application may be less than all the features of a specific disclosed embodiment. Thus, the following claims are incorporated into the specific embodiments as examples or embodiments, wherein each claim is independently used as a separate embodiment, and it is considered that these embodiments can be combined with each other in various combinations or arrangements. The scope of the present application should be determined with reference to the attached claims and the full scope of equivalent forms granted by these claims.

[0171] Multiple embodiments of the present application are described in detail above, but the present application is not limited to these specific embodiments. Those skilled in the art can make various variations and modifications to the embodiments based on the concept of the present application, and these variations and modifications should all fall within the scope of protection required by the present application.

Claims

1. A data processing system for performing a target task; The data processing system includes a plurality of applications, where different applications are used to perform different subtasks; each application has the ability to utilize a target model, where the number of trainable parameters of the target model is greater than 100 million; Multiple applications determine the logic for executing the target task in a conversational manner.

2. The data processing system according to claim 1, Any application executes the corresponding subtask, generates session content that can represent the subresult, and outputs the session content to the target environment; The target environment is a running environment of multiple applications, and is used to establish a session connection between the multiple applications; Other applications can obtain the sub-result from the target environment based on the session connection. If the sub-result meets the target condition, the other applications execute corresponding sub-tasks, generate session content that can represent the sub-result, and output the session content to the target environment; If the sub-result does not meet the target condition, other application programs do not execute the corresponding sub-task; different application programs may correspond to the same or different target conditions.

3. According to the data processing system of claim 2, the conversation content generated by any application is in natural language, and / or the data processing system includes a target interface, and the target interface is used to display the conversation content of the application.

4. The data processing system according to claim 1, wherein the target model comprises a plurality of sub-models; Each application program can call a sub-model to execute the sub-task according to the sub-task it executes.

5. A data processing method, comprising: Get target instructions; The target instruction is used to enable the electronic device to perform a target task; Identify target task groups; The target task group includes a plurality of applications, and different applications are used to perform different subtasks; each application has the ability to utilize a target model, and the number of trainable parameters of the target model is greater than 100 million; A target environment is provided so that multiple application programs in the target task group can run in the target environment, and the logic for executing the target task is determined between the multiple application programs in a conversational manner.

6. An electronic device comprising: The memory is used to store computer programs; The plurality of processors are used to execute the computer program so that the electronic device can implement the following steps: Obtaining a target instruction; the target instruction is used to enable the electronic device to perform a target task; Determine a target task group; the target task group includes a plurality of applications, and different applications are used to perform different subtasks; each application has the ability to utilize a target model, and the number of trainable parameters of the target model is greater than 100 million; A target environment is provided so that multiple application programs in the target task group can run in the target environment, and the logic for executing the target task is determined between the multiple application programs in a conversational manner.

7. The electronic device according to claim 6, wherein the processor obtains a target task for diagnosing the electronic device, the target task group comprising at least a first application, a second application, a third application, and a fourth application; The first application uses the first model to detect whether there is a problem with the electronic device, generates a first session content and sends it to the target environment, wherein the first session content includes the detection result; The second application classifies the problems existing in the electronic device using the second model, generates second conversation content and sends it to the target environment, wherein the second conversation content includes the classification result; A third application uses a third model to repair the problem in the electronic device, generates a third session content and sends it to the target environment, wherein the third session content includes a repair result; A fourth application program evaluates the repaired problem data using a fourth model, generates fourth session content, and sends the fourth session content to the target environment, wherein the fourth session content includes an evaluation result; The first application, the second application, the third application, and the fourth application can all obtain the first session content, the second session content, the third session content, and the fourth session content from the target environment.

8. The electronic device according to claim 7, The second application classifies the problem existing in the electronic device using a second model when the first conversation content satisfies a first condition, wherein the first condition is that the detection result included in the first conversation content indicates that there is a problem in the electronic device; The third application uses the third model to repair the problem existing in the electronic device when the second conversation content meets the second condition, and the second condition is that the classification result included in the second conversation content corresponds to the third model; The fourth application uses a fourth model to evaluate the repaired problem data when the third conversation content meets a third condition, and the third condition is that the repair result included in the third conversation content indicates that the problem in the electronic device has been repaired.

9. The electronic device according to claim 7, When the fourth conversation content satisfies a fourth condition, the third application uses the third model to re-repair the problem in the electronic device, or connects to other applications to repair the problem in the electronic device based on the other applications, and the fourth condition is that the evaluation result included in the fourth conversation content indicates that the problem has not been successfully repaired.

10. The electronic device according to claim 7, further comprising a target interface; The target interface displays at least the first conversation content, the second conversation content, the third conversation content, and the fourth conversation content.

Images