Task execution method based on large model

Abstract

The invention provides a task execution method based on a large model, and relates to the technical field of artificial intelligence, in particular to the technical fields of large models, deep learning, software development and the like. The task execution method comprises the steps that in response to a timed task instruction, a first processing unit carries out fine adjustment on a first large model by calling a plurality of task tools used for achieving a plurality of functions to obtain a second large model, and the timed task instruction is used for triggering fine adjustment operation aiming at the first large model at a preset moment; and the first processing unit sends model information of the second large model to the second processing unit in response to the fact that the performance evaluation result of the second large model meets the preset performance condition, so that the second processing unit replaces the first large model with the second large model based on the model information.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to technical fields such as large models, deep learning, and software development, and more specifically, to a task execution method based on large models. Background Art

[0002] With the development of artificial intelligence technology, large models have emerged. Large models refer to "large parameter" models trained using large-scale data and powerful computing power. These models have high versatility and generalization capabilities and can be applied to natural language processing, image recognition, speech recognition and other fields.

[0003] In the process of applying the large model to the above fields, the large model can be fine-tuned. Large model fine-tuning refers to further training the pre-trained large model to adapt to the needs of specific tasks or fields. Summary of the invention

[0004] The present disclosure provides a task execution method based on a large model, a task processing method and apparatus based on a large model, an intelligent agent, equipment, medium, and product.

[0005] According to one aspect of the present disclosure, a large-model-based task execution method is provided, comprising: in response to a scheduled task instruction, a first processing unit fine-tunes a first large model by calling a plurality of task tools for realizing a plurality of functions to obtain a second large model, wherein the scheduled task instruction is used to trigger a fine-tuning operation on the first large model at a predetermined time; and, in response to a performance evaluation result on the second large model meeting a predetermined performance condition, the first processing unit sends model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information.

[0006] According to another aspect of the present disclosure, a large model-based task processing method is provided, including: in response to receiving model information, a second processing unit deploys a second large model according to the above model information to replace the first large model with the above second large model, and the above model information is obtained using the task execution method; wherein the above first large model is used to process the business data of the target business before a predetermined time, and the above second large model is used to process the business data of the above target business after the above predetermined time.

[0007] According to another aspect of the present disclosure, a task execution device based on a large model is provided, including: a fine-tuning module, for responding to a scheduled task instruction, a first processing unit fine-tuning the first large model by calling multiple task tools for realizing multiple functions to obtain a second large model, wherein the above-mentioned scheduled task instruction is used to trigger the fine-tuning operation for the above-mentioned first large model at a predetermined time; and a sending module, for responding to the performance evaluation result for the above-mentioned second large model meeting the predetermined performance condition, the above-mentioned first processing unit sending the model information of the above-mentioned second large model to the second processing unit, so that the above-mentioned second processing unit replaces the above-mentioned first large model with the above-mentioned second large model based on the above-mentioned model information.

[0008] According to another aspect of the present disclosure, a task processing device based on a large model is provided, including: a deployment module, for responding to receiving model information, a second processing unit deploying a second large model according to the above model information to replace the first large model with the above second large model, and the above model information is obtained using a task execution device; wherein the above first large model is used to process the business data of the target business before a predetermined time, and the above second large model is used to process the business data of the above target business after the above predetermined time.

[0009] According to another aspect of the present disclosure, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0010] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the steps of the above method are implemented.

[0011] According to another aspect of the present disclosure, a computer program product is provided, including a computer program or instructions, and the steps of the above method are implemented when the computer program or instructions are executed by a processor.

[0012] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0014] Figure 1The system architecture to which the large model-based task execution method and the large model-based task processing method according to the embodiments of the present disclosure can be applied is schematically shown;

[0015] Figure 2 A flowchart of a task execution method based on a large model according to an embodiment of the present disclosure is schematically shown;

[0016] Figure 3 Schematically shows an example schematic diagram of a process in which a first processing unit according to an embodiment of the present disclosure fine-tunes a first large model by calling multiple task tools for realizing multiple functions to obtain a second large model;

[0017] Figure 4A The schematic diagram schematically shows an example process in which the first processing unit according to an embodiment of the present disclosure calls multiple task tools in the execution order and sequentially executes multiple subtasks to fine-tune the first large model to obtain the second large model;

[0018] Figure 4B An example schematic diagram schematically shows a process of acquiring a training data set for a target business according to an embodiment of the present disclosure;

[0019] Figure 4C An example schematic diagram schematically shows a process of fine-tuning a first large model using a training data set to obtain a second large model according to an embodiment of the present disclosure;

[0020] Figure 4D An example schematic diagram schematically shows a process of obtaining a performance evaluation result for the second largest model according to an embodiment of the present disclosure;

[0021] Figure 5 An example schematic diagram of a task execution process based on a large model according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 6 A flowchart of a task processing method based on a large model according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 7 A block diagram of a task execution device based on a large model according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 8 A block diagram of a task processing device based on a large model according to an embodiment of the present disclosure is schematically shown;

[0025] Fig. 9 A block diagram schematically shows a structure of an intelligent agent of a large model according to an embodiment of the present disclosure; and

[0026] Fig.10A block diagram of an electronic device suitable for implementing a large model-based task execution method and a large model-based task processing method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0027] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0028] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0029] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0030] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0031] Since the model fine-tuning function of the platform cannot achieve automated scheduled execution, and manually organizing related application programming interface (API) calls to complete the entire process from data processing, file uploading to training, evaluation, and deployment is too complicated, it is not conducive to users achieving automated and rapid iteration of model effects on a daily basis.

[0032] In one example, the relevant technology can only fine-tune small models in the traditional sense, and there is no fine-tuning scenario for large models; and the relevant technology usually only integrates the fine-tuning function, lacking data processing, model evaluation and other links. Based on this, for platforms that do not expose the fine-tuning interface, it is impossible to build a code-based automated pipeline; for platforms that expose the fine-tuning interface, since the process is usually left to the user to build by themselves, the cost of getting started is high.

[0033] To this end, the embodiment of the present disclosure proposes a task execution scheme based on a large model. For example, in response to a scheduled task instruction, the first processing unit fine-tunes the first large model by calling multiple task tools for implementing multiple functions to obtain a second large model, wherein the scheduled task instruction is used to trigger the fine-tuning operation for the first large model at a predetermined time; and in response to the performance evaluation result for the second large model meeting the predetermined performance condition, the first processing unit sends the model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information.

[0034] According to an embodiment of the present disclosure, the first processing unit automatically starts the fine-tuning pipeline for the first large model at a predetermined time by calling multiple task tools for realizing multiple functions according to a preset timed task instruction, so as to obtain a second large model with better performance. On this basis, when the performance evaluation result of the second large model meets the predetermined performance conditions, the first processing unit sends the model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information, thereby realizing seamless update of the large model, which not only ensures that the large model can continuously optimize and adapt to environmental changes, but also avoids manual intervention, reduces maintenance costs, and improves the automation and stability of task execution, so as to realize automatic continuous fine-tuning iteration of the large model, which is conducive to improving the performance of the large model in user scenarios.

[0035] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0036] In the technical solution of the present invention, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0037] Figure 1 The system architecture to which the task execution method based on a large model and the task processing method based on a large model can be applied according to the embodiment of the present disclosure is schematically shown. It should be noted that: Figure 1What is shown is merely an example of a system architecture to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but it does not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios.

[0038] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first processing unit 110 , a server 120 , a server 130 , a server 140 , and a network 150 .

[0039] The first processing unit 110 can be a one-stop large model development platform for developers. The large model development platform designs a set of SDK (Software Development Kit) around the large model tool chain. The SDK performs complete logic encapsulation based on each business module, such as fast authentication integration, local data set operation, and training task pipeline, so as to facilitate users to be simpler and more focused on each process in the large model tool chain.

[0040] Server 120, server 130 and server 140 can be set in an electronic device, for example, and the electronic device can be a computer. Server 120, server 130 and server 140 can be various types of servers that provide various services. For example, server 120, server 130 and server 140 can be cloud servers, also known as cloud computing servers or cloud hosts, which are host products in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services (Virtual Private Server). Server 120, server 130 and server 140 can also be servers of a distributed system, or servers combined with blockchain.

[0041] The server 120 may include a second processing unit 121, the server 130 may include a second processing unit 131, and the server 140 may include a second processing unit 141. The second processing unit 121, the second processing unit 131, and the second processing unit 141 may be processing units with computing power and storage power, such as a central processing unit (CPU) in a server. The network 150 is used as a medium for providing a communication link between the second processing unit 121, the second processing unit 131, the second processing unit 141, and the first processing unit 110. The network 150 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0042] It should be noted that the task execution method provided in the embodiment of the present disclosure may be executed by the first processing unit 110. Accordingly, the task execution device provided in the embodiment of the present disclosure may generally be disposed in the first processing unit 110.

[0043] The task processing method provided in the embodiment of the present disclosure may be executed by the server 120 , the server 130 , and the server 140 . Accordingly, the task processing device provided in the embodiment of the present disclosure may be disposed in the server 120 , the server 130 , and the server 140 .

[0044] It should be understood that Figure 1 The number of the first processing unit, the server, and the second processing unit in the embodiment is only illustrative. According to the implementation requirements, there may be any number of the first processing unit, the server, and the second processing unit.

[0045] It should be noted that the sequence numbers of the operations in the following method are only used as representations of the operations for the purpose of description, and should not be regarded as representing the execution order of the operations. Unless explicitly stated, the method does not need to be executed completely in the order shown.

[0046] The above describes the system architecture that can be applied to the task execution method based on a large model and the task processing method based on a large model provided by the present disclosure. Figure 2 For example, the task execution process based on the large model of the present invention is further explained.

[0047] Figure 2 The flowchart of the large model-based task execution method according to an embodiment of the present disclosure is schematically shown.

[0048] like Figure 2 As shown, the large model-based task execution method 200 includes operations S210 to S220.

[0049] In operation S210, in response to a scheduled task instruction, the first processing unit fine-tunes the first large model by calling multiple task tools for implementing multiple functions to obtain a second large model, wherein the scheduled task instruction is used to trigger the fine-tuning operation on the first large model at a predetermined time.

[0050] In operation S220, in response to the performance evaluation result for the second large model meeting the predetermined performance condition, the first processing unit sends the model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information.

[0051] The scheduled task instruction refers to a preset instruction for triggering the model fine-tuning operation. The scheduled task instruction can be automatically started at a predetermined time to ensure that the first large model can automatically trigger the fine-tuning operation for the first large model. The scheduled task instruction can be set by the system administrator or dynamically adjusted according to actual needs.

[0052] The fine-tuning operation may include multiple subtasks, and a subtask may refer to a smaller, specific task unit decomposed from the fine-tuning operation, and these subtasks have a specific execution order. The execution order refers to the need to execute these subtasks step by step in a certain order when completing the entire fine-tuning operation. For example, the fine-tuning operation includes subtask 1->subtask 2->subtask 3->subtask 4. In the process of executing the fine-tuning operation for the first large model, it is necessary to execute subtask 1 first, and after subtask 1 is executed, execute subtask 2, and after subtask 2 is executed, execute subtask 3, so as to complete the execution of the fine-tuning operation.

[0053] Multiple subtasks can be used to implement multiple functions. Multiple functions can be configured according to actual business needs and are not limited here. Multiple functions may include at least one of the following: data processing function, model training function, evaluation function, and deployment function, etc. For example, each subtask can be used to implement different functions, that is, subtask 1 is used to implement the data processing function, subtask 2 is used to implement the model training function, subtask 3 is used to implement the evaluation function, and subtask 4 is used to implement the deployment function. Alternatively, at least two of the multiple subtasks can also be used to implement the same function, that is, subtask 1 and subtask 2 are both used to implement the data processing function, subtask 3 is used to implement the model training function, and subtask 4 is used to implement the bottle function, etc.

[0054] In response to the scheduled task instruction, the first processing unit can fine-tune the first large model by calling multiple task tools for realizing multiple functions to obtain the second large model. The first processing unit refers to a one-stop large model development platform for developers. The first processing unit designs a set of SDKs around the large model tool chain to realize the automatic calling of multiple task tools and the automatic execution of multiple subtasks. The second large model refers to an improved model obtained by fine-tuning the first large model, and the performance of the second large model is improved compared with the first large model.

[0055] After obtaining the second largest model, the performance of the second largest model can be evaluated to obtain a performance evaluation result. Performance evaluation refers to the process of quantitatively analyzing the performance of the second largest model on a specific task or data set. Performance evaluation can use a variety of indicators, such as accuracy, recall, F1 value, mean square error, etc., so that the performance evaluation result can comprehensively measure the performance of the second largest model.

[0056] After obtaining the performance evaluation results, it can be determined whether the performance evaluation results meet the predetermined performance conditions. The predetermined performance conditions can be determined based on the target business applicable to the first large model. If the performance evaluation results do not meet the predetermined performance conditions, the online process of the second large model can be terminated. If the performance evaluation results meet the predetermined performance conditions, the model information of the second large model can be sent to the second processing unit. Model information refers to data information that describes the characteristics of the second large model, such as structure, parameters and performance. The second processing unit refers to the subject responsible for receiving model information and performing replacement operations on the first large model. For example, the second processing unit can be a cloud or local server. After receiving the model information, the second processing unit can remove the original first large model from the local area and load the second large model to the local area according to the received model information.

[0057] According to an embodiment of the present disclosure, the first processing unit automatically starts the fine-tuning pipeline for the first large model at a predetermined time by calling multiple task tools for realizing multiple functions according to the preset timed task instructions, so as to obtain the second large model with better performance. On this basis, when the performance evaluation result of the second large model meets the predetermined performance conditions, the first processing unit sends the model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information, thereby realizing seamless update of the large model, which not only ensures that the large model can continuously optimize and adapt to environmental changes, but also avoids manual intervention, reduces maintenance costs, and improves the automation and stability of task execution, so as to realize the automatic continuous fine-tuning iteration of the large model, which is conducive to improving the performance of the large model in user scenarios.

[0058] Figure 3 The diagram schematically shows an example process in which the first processing unit according to an embodiment of the present disclosure fine-tunes the first large model by calling multiple task tools for implementing multiple functions to obtain the second large model.

[0059] like Figure 3 As shown, in 300, the timed task instruction 310 is used to trigger a fine-tuning operation for the first large model 350 at a predetermined time, and the fine-tuning operation may include multiple subtasks 340 with an execution order. The execution order means that the multiple subtasks 340 need to be executed in a certain order to ensure the correctness and effectiveness of the fine-tuning process.

[0060] In one example, in response to the timed task instruction 310, the first processing unit 320 can execute multiple subtasks 340 in sequence by calling multiple task tools in the execution order. Specifically, for each subtask 340, the first large model 350 can be fine-tuned by calling the task tool used to execute the subtask 340 to obtain the second large model 360.

[0061] For example, the fine-tuning operation includes N subtasks 340. Task tool 330_1 can be called first to execute the first subtask 340; then task tool 330_2 can be called to execute the second subtask 340; then task tool 330_3 can be called to execute the third subtask 340, and so on. Finally, task tool 330_N can be called to execute the Nth subtask 340, where N is a positive integer greater than 1.

[0062] According to the embodiments of the present disclosure, the fine-tuning operation is subdivided into multiple subtasks with a clear execution order, and the first processing unit calls the corresponding task tools in this order to execute them in sequence. This step-by-step and orderly execution of fine-tuning allows each subtask to optimize and adjust the model in a targeted manner based on the previous task, thereby achieving efficient and accurate fine-tuning of the first large model, which not only improves the efficiency of fine-tuning, but also ensures the reliability and effectiveness of the fine-tuning results.

[0063] Figure 4A The diagram schematically shows an example process in which the first processing unit according to an embodiment of the present disclosure calls multiple task tools in an execution order and executes multiple subtasks in sequence to fine-tune the first large model to obtain the second large model.

[0064] like Figure 4A As shown, in 400A, the first large model 406 can be used to process data related to the target business. The task tool can include at least one of the following: a data processing tool 403 and a model training tool 405 .

[0065] Data processing tool 403 refers to software or tools used to clean, preprocess, and annotate raw data. Model training tool 405 refers to software or framework used to train and optimize models. The specific implementation form of data processing tool 403 and model training tool 405 can be configured according to actual business needs and is not limited here. For example, in order to adapt the model fine-tuning service to the hardware performance of the first processing unit 402, the tool of the first processing unit 402 adapted according to the model fine-tuning service can be used. Data processing tool 403 and model training tool 405 may include at least one of the following tools: an instruction set, an instruction library, and a software development kit, etc.

[0066] In one example, in response to the scheduled task instruction 401, the first processing unit 402 can call the data processing tool 403 to obtain a training data set 404 for the target business, which refers to a data set that meets the model training requirements after the original data is processed by the data processing tool 403, and the data set contains the key features and label information of the target business. On this basis, the first processing unit 402 can call the model training tool 405 to use the training data set 404 to fine-tune the first large model 406 to obtain the second large model 407.

[0067] According to the embodiments of the present disclosure, the first processing unit can effectively improve the quality and availability of training data by calling data processing tools to clean, preprocess and annotate the original data, ensure that the training data set accurately reflects the characteristics and laws of the target business, and provide high-quality input for fine-tuning the model. On this basis, the first processing unit can call the model training tool and use the processed training data set to fine-tune the first large model, so that the model can better adapt to the specific needs and scenarios of the target business, and improve the performance of the second large model on the target business.

[0068] Figure 4B An example schematic diagram of a process of acquiring a training data set for a target business according to an embodiment of the present disclosure is schematically shown.

[0069] like Figure 4B As shown, in 400B, in response to the timed task instruction 408, a fine-tuning operation for the first large model can be triggered, and the first large model is used to process data related to the target business 409. The target business can be configured with at least one data source for providing historical business data. In the data processing stage, the user can load the initial data set 411 from different data sources according to the business requirements 412 of the target business 409, and process the initial data set 411 according to the business requirements 412 to obtain a training data set 414 that meets the format document requirements of the first processing unit.

[0070] In one example, the initial data set 411 can be loaded from at least one data source configured according to the target business 409. The specific method of loading the initial data set 411 can be configured according to actual business requirements and is not limited here. For example, the Dataset.load() function can be used to load and construct the initial data set 411 from the data source supported by the first processing unit.

[0071] The first processing unit is associated with a candidate data source set 410, which may include a candidate data source 410_1, a candidate data source 410_2, ..., a candidate data source 410_P, where P is a positive integer. The candidate data source may include a file system, an object storage, or a Python object, etc. When the data sources configured by the target business 409 are the candidate data source 410_1 and the candidate data source 410_2, the initial data set 411 may be loaded from the candidate data source 410_1 and the candidate data source 410_2.

[0072] In an embodiment of the present disclosure, the user's consent or authorization may be obtained before obtaining the initial data set 411. For example, before obtaining the initial data set 411, a request to obtain the initial data set 411 may be issued to the user. When the user agrees or authorizes that the initial data set 411 may be obtained, an operation of obtaining the initial data set 411 from at least one data source is performed.

[0073] In the embodiments of the present disclosure, a corresponding operation entry may be provided for the user to choose to agree or reject the automated decision result. That is, before the initial data set 411 is processed, the user may obtain the instruction of agreeing or rejecting the processing input through the corresponding operation entry. If the user agrees to process, the initial data set 411 is processed, that is, the format adjustment operation is performed. If the user rejects the processing, the expert decision process is entered.

[0074] In one example, the format of the initial data set 411 can be adjusted according to the business requirements 412 of the target business 409 to obtain the intermediate data set 413. The format adjustment may include regularizing the fields in the initial data set 411 and formatting the content. For example, the initial data set 411 can be cleaned and the fields can be regularized by calling methods such as .map() and .filter() to obtain the intermediate data set 413 that meets the format requirements of the first processing unit.

[0075] After obtaining the intermediate data set 413, the intermediate data set 413 can be verified. The specific content of the verification can be configured according to actual business needs and is not limited here. For example, the verification may include at least one of the following: integrity verification, accuracy verification, and consistency verification. Integrity verification refers to checking whether the intermediate data set 413 is missing key fields. Accuracy verification refers to verifying whether the intermediate data set 413 is accurate. Consistency verification refers to verifying whether the data provided by different data sources in the intermediate data set 413 are consistent in terms of format, unit, naming rules, etc.

[0076] If the intermediate data set 413 fails to pass the verification, the above data processing stage can be re-executed to obtain a new intermediate data set 413. If the intermediate data set 413 passes the verification, the intermediate data set 413 can be determined as the training data set 414. After obtaining the training data set 414, the .save() method of the Dataset object can be used to upload the training data set 414 to the first processing unit for subsequent training.

[0077] According to the embodiments of the present disclosure, by configuring the data source of the target business and obtaining the initial data set from it, the historical business data related to the target business can be effectively integrated and utilized to provide a rich and reliable data foundation for subsequent data processing and model training. By adjusting the format of the initial data set to meet the business needs of the target business and the format requirements of model training, the availability and consistency of the data can be improved. On this basis, by performing data verification on the intermediate data set obtained after format adjustment, the integrity, accuracy and consistency of the data are ensured, thereby ensuring the quality of the training data set, which helps to improve the training effect and performance of large models.

[0078] Figure 4C An example schematic diagram of the process of fine-tuning the first large model using a training data set to obtain a second large model according to an embodiment of the present disclosure is schematically shown.

[0079] like Figure 4C As shown, in 400C, in the model fine-tuning stage, the user can define and initiate a fine-tuning task according to the actual needs of the target business 415. For example, the target business 415 can be configured with a fine-tuning template 416 and fine-tuning parameters for the fine-tuning operation.

[0080] The fine-tuning template 416 refers to a template or framework designed in advance to guide the fine-tuning operation according to the characteristics and requirements of the target business 415. For example, the fine-tuning template 416 may specify key aspects and steps that need to be paid attention to during the fine-tuning process, such as the level of fine-tuning, parameter adjustment strategy, etc. The fine-tuning template 416 may be selected by the user from the trainer package provided by the first processing unit. For example, the fine-tuning template 416 may include Finetune, PostPreTrain, etc.

[0081] Fine-tuning parameters refer to model parameters that need to be specifically adjusted and optimized during the fine-tuning process, such as learning rate, number of training rounds, and batch size. The settings of these parameters directly affect the effect of fine-tuning and the performance of the model. For example, users can set fine-tuning parameters according to their actual expectations.

[0082] During the process of the first processing unit fine-tuning the first large model 418, the first large model 418 can be fine-tuned based on the fine-tuning template 416 and the training data set 417, and during the fine-tuning process, the operation S410 is continuously performed for the intermediate model parameters obtained by the fine-tuning. In operation S410, it can be determined whether the intermediate model parameters satisfy the fine-tuning parameters?

[0083] If not, the fine-tuning step for the first large model 418 can be continued. If yes, the large model corresponding to the intermediate model parameters that meet the fine-tuning parameters can be determined as the second large model 419. In the above process, the first processing unit can track the status of the fine-tuning task through the local event loop, and use the user-predefined event processing function to process the phased events to ensure the smooth completion of the fine-tuning task.

[0084] According to the embodiments of the present disclosure, when the first large model is fine-tuned based on the fine-tuning template and using the training data set, the model can fully learn and adapt to the data characteristics and rules of the target business, further optimize the parameters and structure of the large model, and improve the large model's understanding and processing capabilities of the target business, thereby improving the performance of the second large model on the target business, which helps to improve the processing efficiency and user experience of the target business.

[0085] Figure 4D An example schematic diagram schematically illustrates a process of obtaining a performance evaluation result for a second largest model according to an embodiment of the present disclosure.

[0086] like Figure 4D As shown, in 400D, in the model effect evaluation stage, the user can use a customized verification machine to evaluate the fine-tuned second largest model through the evaluation function of the first processing unit, and determine whether the fine-tuned effect can meet the online standard based on business needs and evaluation performance.

[0087] In one example, the task tool may also include an evaluation tool 421. An evaluation tool refers to a software tool or framework for evaluating and analyzing model performance, which can quantitatively and qualitatively analyze the performance of the model on a specific data set according to preset evaluation indicators, thereby providing an objective evaluation of the model performance. The specific implementation form of the evaluation tool 421 can be configured according to actual business needs and is not limited here. For example, in order to adapt the model fine-tuning service to the hardware performance of the first processing unit 402, a tool of the first processing unit 402 adapted according to the model fine-tuning service can be used. The evaluation tool 421 may include at least one of the following tools: an instruction set, an instruction library, and a software development kit, etc.

[0088] During the evaluation process, the first processing unit 420 can call the evaluation tool 421 to perform a performance evaluation on the second largest model 425 based on the evaluation strategy 424 configured for the target business 423, using the verification data set 422, to obtain a performance evaluation result 426. The evaluation strategy 424 indicates the specific method of evaluation. For example, the evaluation strategy 424 may include automatic referee evaluation, statistical indicator evaluation, and manual evaluation, etc. The performance evaluation result 426 may include indicator data reflecting the performance of the model. The form of the performance evaluation result 426 can be configured according to actual business needs and is not limited here. For example, the form of the performance evaluation result 426 may include at least one of the following: report, text, picture, etc.

[0089] In a specific example, the evaluation tool 421 may include a first sub-tool for platform evaluation and a second sub-tool for local evaluation. For example, for platform evaluation, it provides an out-of-the-box evaluation method, and the first sub-tool can be used to place the second largest model after fine-tuning in the platform environment. In this case, the verification data set 422 can be a data set for the target business 423.

[0090] Alternatively, for local evaluation, which provides a higher degree of customization, the second sub-tool can be used to evaluate the second large model using the evaluation strategy 424 in the local environment. In this case, the verification dataset 422 can be a dataset generated via the second large model, which can be understood as evaluation answer data.

[0091] According to the embodiments of the present disclosure, by integrating the evaluation tool into the task tool and configuring the evaluation strategy for the target business, a comprehensive and accurate performance evaluation of the second largest model is achieved, which provides a reference for the further optimization and application of the large model, thereby ensuring the reliability and effectiveness of the second largest model in the target business.

[0092] Figure 5 An example schematic diagram of a task execution process based on a large model according to an embodiment of the present disclosure is schematically shown.

[0093] like Figure 5 As shown, in 500, in response to the scheduled task instruction 501, the first processing unit can obtain a training data set 502 for the target business by calling a data processing tool. On this basis, the first processing unit can fine-tune the first large model 503 by calling the model training tool and using the training data set 502 to obtain a second large model 504.

[0094] After obtaining the second largest model 504, the first processing unit may also perform performance evaluation on the second largest model 504 by calling an evaluation tool based on an evaluation strategy and using a verification data set 505 to obtain a performance evaluation result 506. After obtaining the performance evaluation result 506, operation S510 may be performed.

[0095] In operation S510, is it determined whether the performance evaluation result 506 meets the predetermined performance condition? If not, the first processing unit may terminate the iteration and continue to wait for a response to the next timed task instruction 501. If so, the first processing unit may send the model information 507 of the second large model 504 to the second processing unit, so that the second processing unit replaces the first large model 503 with the second large model 504 based on the model information 507.

[0096] The above are only exemplary embodiments, but are not limited thereto and may also include other large model-based task execution methods known in the art, as long as they can achieve automated continuous fine-tuning and iteration of the large model.

[0097] The above describes the task execution process based on the large model provided by the present disclosure. Figure 6 For example, the task processing process based on the large model of the present disclosure is further explained.

[0098] Figure 6 The flowchart of the task processing method based on the large model according to the embodiment of the present disclosure is schematically shown.

[0099] like Figure 6 As shown, the large model-based task processing method 600 includes operation S610.

[0100] In operation S610, in response to receiving the model information, the second processing unit deploys the second large model according to the model information to replace the first large model with the second large model, and the model information is obtained using the task execution method.

[0101] In the model deployment phase, the virtual routing of the first processing unit can be used to divert the dynamic small traffic of online requests to the fine-tuned second largest model to verify the effect in the actual production environment. On this basis, the flow switching strategy can be used to implement the grayscale upgrade of the first largest model to the second largest model, thereby realizing the dynamic update of the model effect.

[0102] The deployment method of the second largest model can be configured according to actual business needs and is not limited here. For example, the model information of the second largest model can be uploaded to the designated directory of the second processing unit, the input and output interfaces of the second largest model can be set, and the dependent libraries and environment variables required for the operation of the second largest model can be configured to ensure that the second largest model can be loaded and run normally on the second processing unit.

[0103] In one example, the dynamic update frequency of the model effect can be set according to the business needs of the target business. For example, 00:00 of each day can be set as the scheduled time, so that the scheduled task execution can be used to fine-tune the first large model at 00:00 of each day to achieve dynamic daily updates of the model effect. The first large model can be used to process the business data of the target business before the scheduled time, and the second large model can be used to process the business data of the target business after the scheduled time.

[0104] According to an embodiment of the present disclosure, the second processing unit responds to the received model information and deploys a second large model to replace the first large model according to the model information, thereby achieving automatic smooth transition and performance improvement of the large model, thereby improving the processing efficiency and accuracy of the target business.

[0105] The above are only exemplary embodiments, but are not limited thereto, and may also include other large model-based task processing methods known in the art, as long as they can achieve an automated and smooth transition of the large model.

[0106] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned, which should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present invention, but it does not mean that the present invention has or must use this solution.

[0107] Based on the above-mentioned task execution method based on a large model, the present invention also provides a task execution device based on a large model. Figure 7 The device is described in detail.

[0108] Figure 7 A block diagram of a task execution device based on a large model according to an embodiment of the present disclosure is schematically shown.

[0109] like Figure 7 As shown, the large model-based task execution device 700 may include a fine-tuning module 710 and a sending module 720 .

[0110] The fine-tuning module 710 is used to respond to the scheduled task instruction. The first processing unit fine-tunes the first large model by calling multiple task tools for implementing multiple functions to obtain the second large model, wherein the scheduled task instruction is used to trigger the fine-tuning operation on the first large model at a predetermined time.

[0111] The sending module 720 is used for sending the model information of the second large model to the second processing unit in response to the performance evaluation result of the second large model meeting the predetermined performance condition, so that the second processing unit replaces the first large model with the second large model based on the model information.

[0112] According to an embodiment of the present disclosure, the fine-tuning operation includes a plurality of subtasks having an execution order.

[0113] According to an embodiment of the present disclosure, the fine-tuning module 710 may include an execution submodule.

[0114] The execution submodule is used for the first processing unit to call multiple task tools according to the execution order and execute multiple subtasks in sequence to fine-tune the first large model to obtain the second large model.

[0115] According to an embodiment of the present disclosure, the first large model is used to process data related to the target business, and the task tool includes at least one of the following: a data processing tool and a model training tool.

[0116] According to an embodiment of the present disclosure, the execution submodule may include an acquisition unit and a fine-tuning unit.

[0117] The acquisition unit is used for the first processing unit to call the data processing tool to obtain a training data set for the target business.

[0118] The fine-tuning unit is used for the first processing unit to call the model training tool to use the training data set to fine-tune the first large model to obtain the second large model.

[0119] According to an embodiment of the present disclosure, the target service is configured with at least one data source for providing historical service data.

[0120] According to an embodiment of the present disclosure, the acquisition unit may include an acquisition subunit, an adjustment subunit and a verification subunit.

[0121] The acquisition subunit is used to acquire an initial data set from at least one data source.

[0122] The adjustment subunit is used to adjust the format of the initial data set according to the business requirements of the target business to obtain an intermediate data set.

[0123] The verification subunit is configured to determine the intermediate data set as a training data set in response to the intermediate data set passing the verification.

[0124] According to an embodiment of the present disclosure, the target service is configured with a fine-tuning template and fine-tuning parameters for fine-tuning operations.

[0125] According to an embodiment of the present disclosure, the fine-tuning unit may include a fine-tuning sub-unit.

[0126] The fine-tuning subunit is used to fine-tune the first large model based on the fine-tuning template and the training data set until the fine-tuning parameters are met to obtain the second large model.

[0127] According to an embodiment of the present disclosure, the task tool further includes an evaluation tool, and the target business is configured with an evaluation strategy.

[0128] The task execution device 700 may further include an evaluation module.

[0129] An evaluation module is used for the first processing unit to call an evaluation tool to perform performance evaluation on the second largest model based on an evaluation strategy and using a verification data set to obtain a performance evaluation result, wherein the verification data set includes at least one of the following: a data set for a target business and a data set generated by the second largest model.

[0130] Based on the above task processing method, the present invention also provides a task processing device. Figure 8 The device is described in detail.

[0131] Figure 8 A block diagram of a task processing device based on a large model according to an embodiment of the present disclosure is schematically shown.

[0132] like Figure 8 As shown, the large model-based task processing apparatus 800 may include a deployment module 810 .

[0133] The deployment module 810 is used for, in response to receiving the model information, the second processing unit deploys the second large model according to the model information to replace the first large model with the second large model, and the model information is obtained using the task execution device.

[0134] The first large model is used to process the business data of the target business before the predetermined time, and the second large model is used to process the business data of the target business after the predetermined time.

[0135] Fig. 9 The structural block diagram of the intelligent agent of the large model according to the embodiment of the present disclosure is schematically shown.

[0136] In the embodiments of the present disclosure, inspired by the von Neumann structure in modern computer theory, such as Fig. 9 As shown, the AI ​​agent 900 may include five core modules: an input module 910 , a control module 920 , a storage module 930 , a calculation module 940 and an output module 950 .

[0137] The input module 910 is responsible for receiving or sensing information such as queries, requests, instructions, signals or data from the outside world (such as users or external environment), and converting them into a format that can be understood and processed by the AI ​​agent 900. The input module 910 is the primary link for the AI ​​agent 900 to interact with the outside world. It enables the AI ​​agent 900 to efficiently and accurately obtain the necessary "sensory" information from the outside world and respond to this information.

[0138] In an example, the input module 910 may input the initial data set described above.

[0139] The control module 920 is the core support for the AI ​​agent 900 to handle complex tasks. The control module 920 can execute the task execution method and task processing method described above.

[0140] In the example, the control module 920 will continuously interact with the storage module 930, the computing module 940 and / or the output module 950 during operation. However, it should be noted that in the embodiment of the present disclosure, the control module 920 acts as a single initiator to initiate communication with the storage module 930, the computing module 940 and / or the output module 950, and there is no communication coupling between the storage module 930, the computing module 940 and the output module 950.

[0141] In the example, the performance of the control module 920 may be closely related to the large model on which the AI ​​agent 900 is based. In order to give full play to the capabilities of the large language model, the internal structure of the control module 920 may be designed to be highly configurable and scalable to cope with various types of tasks and requirements in real scenarios.

[0142] The storage module 930 may be responsible for memorizing candidate category mappings for different application scenarios and the multimodal large model applicable to the application scenario. The multimodal large model as described above may be included in the storage module 930 .

[0143] In the example, after receiving the scheduled task instruction, the AI ​​agent 900 can trigger the task execution process, execute the task execution method through the first processing unit, and obtain the second largest model. The model information of the second largest model can be stored in the storage module 930. The AI ​​agent 900 can obtain the model information of the second largest model from the storage module 930 and feed it back to the control module 920. Then, the control module 920 can pass the model information of the second largest model fed back to the output module 950.

[0144] The operation module 940 can be viewed as a predefined tool library. Tools for performing a plurality of tasks as described above can be included in the operation module 940 .

[0145] In the example, when the AI ​​agent 900 needs to process data, it can call the relevant data processing tool from the operation module 940 and feed it back to the control module 920. Then, the control module 920 can use the fed-back data processing tool to obtain a training data set for the target business and pass the training data set to the output module 950. It can be understood that although the large language model has excellent language understanding and generation capabilities, it is the same as a human being. Without the help of any tools, the tasks that can be solved are very limited. When the AI ​​agent 900 is given the ability to call tools, it can achieve tasks such as determining the frequency of category descriptions with the help of tools for determining the frequency of category descriptions.

[0146] The output module 950 can output the model information of the second largest model described above.

[0147] The AI ​​agent 900 according to the embodiment of the present disclosure can simply and effectively improve the level of intelligence, and improve flexibility and versatility.

[0148] Fig.10 A block diagram of an electronic device suitable for implementing a large model-based task execution method and a large model-based task processing method according to an embodiment of the present disclosure is schematically shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0149] like Fig.10 As shown, the device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1002 or a computer program loaded from a storage unit 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the device 1000 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0150] A number of components in the device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0151] The computing unit 1001 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1001 performs the various methods and processes described above, such as task execution methods and task processing methods. For example, in some embodiments, the task execution method and the task processing method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into the RAM 1003 and executed by the computing unit 1001, one or more steps of the task execution method and the task processing method described above may be executed. Alternatively, in other embodiments, the computing unit 1001 may be configured to execute a large model-based task execution method or a large model-based task processing method in any other appropriate manner (for example, by means of firmware).

[0152] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0153] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0154] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0155] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0156] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0157] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0158] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0159] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A task execution method based on a large model, comprising: In response to a scheduled task instruction, the first processing unit fine-tunes the first large model by calling a plurality of task tools for implementing a plurality of functions to obtain a second large model, wherein the scheduled task instruction is used to trigger a fine-tuning operation on the first large model at a predetermined time; and In response to the performance evaluation result of the second large model meeting the predetermined performance condition, the first processing unit sends the model information of the second large model to the second processing unit, so that the second processing unit replaces the first large model with the second large model based on the model information.

2. The method according to claim 1, wherein: The fine-tuning operation includes a plurality of subtasks having an execution order; The first processing unit fine-tunes the first large model by calling multiple task tools for realizing multiple functions, and obtains the second large model including: The first processing unit calls the multiple task tools according to the execution order, and executes the multiple subtasks in sequence to fine-tune the first large model to obtain the second large model.

3. The method according to claim 2, wherein: The first large model is used to process data related to the target business, and the task tool includes at least one of the following: a data processing tool and a model training tool; The first processing unit calls the multiple task tools according to the execution order and executes the multiple subtasks in sequence to fine-tune the first large model to obtain the second large model including: The first processing unit calls the data processing tool to obtain a training data set for the target business; and The first processing unit calls the model training tool to use the training data set to fine-tune the first large model to obtain the second large model.

4. The method according to claim 3, wherein: The target business is configured with at least one data source for providing historical business data; The acquiring of a training data set for the target service includes: Acquire an initial data set from the at least one data source; According to the business requirements of the target business, the format of the initial data set is adjusted to obtain an intermediate data set; and In response to the intermediate data set passing the verification, the intermediate data set is determined as the training data set.

5. The method according to claim 3, wherein: The target service is configured with a fine-tuning template and fine-tuning parameters for the fine-tuning operation; The step of fine-tuning the first large model by using the training data set to obtain the second large model comprises: Based on the fine-tuning template, the first large model is fine-tuned using the training data set until the fine-tuning parameters are met, thereby obtaining the second large model.

6. The method according to claim 3, wherein: The task tool also includes an evaluation tool, and the target business is configured with an evaluation strategy; The method further comprises: The first processing unit calls the evaluation tool to perform performance evaluation on the second large model based on the evaluation strategy and using a verification data set to obtain the performance evaluation result, wherein the verification data set includes at least one of the following: a data set for the target business and a data set generated by the second large model.

7. A task processing method based on a large model, comprising: In response to receiving the model information, the second processing unit deploys the second large model according to the model information to replace the first large model with the second large model, wherein the model information is obtained using the method of claims 1 to 6; The first large model is used to process the business data of the target business before a predetermined time, and the second large model is used to process the business data of the target business after the predetermined time.

8. A task execution device based on a large model, comprising: a fine-tuning module, configured to, in response to a scheduled task instruction, cause the first processing unit to fine-tune the first large model by calling a plurality of task tools for realizing a plurality of functions, so as to obtain a second large model, wherein the scheduled task instruction is configured to trigger a fine-tuning operation on the first large model at a predetermined time; and A sending module is used for sending the model information of the second large model to the second processing unit by the first processing unit in response to the performance evaluation result of the second large model meeting the predetermined performance condition, so that the second processing unit replaces the first large model with the second large model based on the model information.

9. A task processing device based on a large model, comprising: a deployment module, configured to, in response to receiving the model information, deploy the second large model according to the model information by the second processing unit to replace the first large model with the second large model, wherein the model information is obtained by using the apparatus according to claim 8; The first large model is used to process the business data of the target business before a predetermined time, and the second large model is used to process the business data of the target business after the predetermined time.

10. An artificial intelligence agent, configured to execute the method according to any one of claims 1 to 7.

11. An electronic device, comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

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

13. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images