Task processing method and device based on large language model, equipment and medium

Abstract

The invention discloses a task processing method and device based on a large language model, equipment and a medium, and the method comprises the steps: carrying out the disassembly of a target task through the large language model, so as to obtain a plurality of subtasks and subtask information corresponding to the subtasks; based on the task description, the task category and the task difficulty corresponding to the sub-tasks, determining execution models corresponding to the sub-tasks; executing the plurality of sub-tasks by using an execution model to obtain a plurality of sub-execution results; and integrating the plurality of sub-execution results to obtain an execution result of the target task. The category and difficulty of an input task are identified by using a large model, and a complex task is disassembled into a plurality of sub-tasks. According to the difficulty level of each subtask, determining an execution model which is most matched with the current task in difficulty and scene in a model pool; large models of different parameter volumes are dynamically called, and resource waste is reduced to the maximum extent while task quality is guaranteed.

Description

Technical Field

[0001] The present application relates to a field, and specifically to a task processing method, device, equipment and medium based on a large language model. Background Art

[0002] In recent years, with the rapid development of artificial intelligence technology, large-scale language models (LLMs) have shown great potential and application value in natural language processing, machine translation, intelligent question answering and other fields. With their powerful language understanding and generation capabilities, these models are changing the way people interact with computers and are widely used in all walks of life. However, with the continuous expansion of model scale and the increasing complexity of application scenarios, the operation of large models faces many challenges, which are mainly manifested in the following aspects:

[0003] Resource consumption problem: Large-scale language models usually consume a lot of computing resources and energy. For example, models such as GPT-3 contain hundreds of billions of parameters, which require a lot of GPU resources and electricity support during training and inference. This not only brings high operating costs, but also creates a significant environmental burden. At present, many studies are devoted to how to reduce resource consumption while ensuring model performance, but this is still a difficult problem that needs to be solved urgently.

[0004] Inefficient reasoning: Large models often require a long reasoning time when processing complex tasks, which may lead to a decline in user experience in application scenarios with high real-time requirements. Although some studies have proposed methods such as model compression and knowledge distillation to improve reasoning speed, these methods usually sacrifice model performance to a certain extent, making it difficult to achieve a perfect balance between efficiency and performance.

[0005] Insufficient task adaptability: Although large models perform well on a wide range of tasks, they may have insufficient adaptability for specific fields or complex tasks. This may require additional fine-tuning or specialized model versions when dealing with certain highly specialized or complex tasks, increasing the complexity of model maintenance and deployment.

[0006] Unreasonable resource allocation: In practical applications, different tasks have very different requirements for computing resources. However, the existing large-scale model operation methods often adopt a unified resource allocation strategy, which cannot dynamically adjust resource allocation according to the specific needs of the task, resulting in inefficient resource utilization. Sometimes simple tasks occupy too many resources, while complex tasks have insufficient resources.

[0007] Insufficient parallel processing capabilities: Complex tasks can usually be broken down into multiple subtasks, which can theoretically be processed in parallel to improve efficiency. However, existing large model operation methods are still relatively limited in their ability to decompose tasks and process them in parallel, making it difficult to fully utilize the parallel computing capabilities of hardware.

[0008] To address these challenges, researchers have proposed a variety of solutions. For example, model quantization techniques are used to reduce the storage and computing requirements of the model; model pruning methods reduce the size of the model by removing unimportant neurons; distributed training and inference techniques are widely used to improve the processing efficiency of large models. In addition, some studies have attempted to improve the task adaptability of the system through multi-model fusion or dynamic model selection.

[0009] However, these methods can usually only solve part of the problem and lack a unified framework to comprehensively improve the operational efficiency of large models. In particular, existing technologies still have obvious shortcomings in dynamic task processing, intelligent resource allocation, and parallel execution of tasks. Summary of the invention

[0010] In order to solve the above problems, the present application proposes a method, a device and a medium, wherein the method includes:

[0011] Decompose the target task through a large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; determine the step numbers, task descriptions, pre-step numbers, task categories and task difficulties corresponding to the multiple subtasks according to the subtask information; determine the execution models corresponding to the multiple subtasks based on the task descriptions, task categories and task difficulties corresponding to the multiple subtasks; use the execution model to execute the multiple subtasks to obtain sub-execution results of the multiple subtasks; integrate the sub-execution results of the multiple subtasks to obtain the execution result of the target task.

[0012] In one example, based on the task descriptions, task categories and task difficulties respectively corresponding to the multiple subtasks, the execution models respectively corresponding to the multiple subtasks are determined, specifically including: selecting any subtask from the multiple subtasks as a target subtask; determining an execution tool corresponding to the target subtask based on the task description and task category of the target subtask; determining a model size of the execution model corresponding to the target subtask based on the task difficulty corresponding to the target subtask; selecting a target execution model from a preset model pool based on the execution tool and the model size; and traversing the multiple subtasks until the corresponding execution models are determined for the multiple subtasks.

[0013] In one example, the preset model pool maintains a plurality of large language models with different specialization directions and different parameter sizes.

[0014] In one example, the use of the execution model to execute the multiple subtasks specifically includes: using the execution model in the preset model pool with a model size corresponding to the target subtask as the ideal model of the target subtask; if the ideal model is in a called state when determining the execution model of the target subtask, then based on the task difficulty and model size of the target subtask, selecting a preset model with a similar model size from the preset model pool as the model to be executed for the target subtask; when executing the target subtask, determining whether the ideal model is in a called state; if the ideal model is in a called state, executing the target subtask through the model to be executed; if the ideal model is in an idle state, executing the target subtask through the ideal model.

[0015] In one example, before executing the multiple subtasks, using the execution model specifically includes: determining multiple single-line execution orders corresponding to the multiple subtasks based on the step numbers and the preceding step numbers respectively corresponding to the multiple subtasks; determining a serial-parallel execution sequence of the multiple subtasks based on the multiple single-line execution sequences corresponding to the multiple subtasks and the number of idle threads; and executing the multiple subtasks based on the serial-parallel execution sequence and the execution model.

[0016] In one example, the sub-execution results of the multiple sub-tasks are integrated to obtain the execution result of the target task, specifically including: determining an intermediate execution result from the sub-execution results of the multiple sub-tasks according to the serial-parallel execution sequence; inputting the intermediate execution result into the execution model of the next sub-task according to the serial-parallel execution sequence until sequence results corresponding to the multiple execution sequences are obtained; and integrating the sequence results corresponding to the multiple execution sequences to obtain the execution result of the target task.

[0017] In one example, the task categories include TOOL type and Chat type, and the TOOL type tasks are tasks that need to be completed by calling various execution tools.

[0018] The present application also provides a task processing method device based on a large language model, including: a disassembly module, which disassembles the target task through a large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; a subtask information module, which determines the step numbers, task descriptions, pre-step numbers, task categories and task difficulties corresponding to the multiple subtasks according to the subtask information; an execution model module, which determines the execution models corresponding to the multiple subtasks based on the task descriptions, task categories and task difficulties corresponding to the multiple subtasks; an execution module, which uses the execution model to execute the multiple subtasks to obtain sub-execution results of the multiple subtasks; and an integration module, which integrates the sub-execution results of the multiple subtasks to obtain the execution result of the target task.

[0019] The present application also provides a task decomposition processing device based on a large language model, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute: decomposing a target task through a large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks; determining, according to the subtask information, the step numbers, task descriptions, preceding step numbers, task categories and task difficulties corresponding to the plurality of subtasks; determining, based on the task descriptions, task categories and task difficulties corresponding to the plurality of subtasks, the execution models corresponding to the plurality of subtasks; executing the plurality of subtasks using the execution models to obtain sub-execution results of the plurality of subtasks; and integrating the sub-execution results of the plurality of subtasks to obtain the execution result of the target task.

[0020] The present application also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to: decompose a target task through a large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks; determine the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the plurality of subtasks according to the subtask information; determine the execution models corresponding to the plurality of subtasks based on the task descriptions, task categories, and task difficulties corresponding to the plurality of subtasks; execute the plurality of subtasks using the execution model to obtain sub-execution results of the plurality of subtasks; and integrate the sub-execution results of the plurality of subtasks to obtain the execution result of the target task.

[0021] The method proposed in this application can bring the following beneficial effects: using a large model to automatically identify the category and difficulty of the input task, and breaking down complex tasks into multiple subtasks through the large model. On the basis of task decomposition, the large model is used to determine the execution model in the model pool that best matches the current task in terms of difficulty and scenario according to the difficulty level of each subtask and the actual execution process of each module; according to the specific difficulty of the task, it is possible to dynamically call large models with different parameter volumes, and assign appropriate subtasks to models of corresponding capabilities for execution, thereby minimizing resource waste while ensuring task quality. This dynamic call and task allocation mechanism not only significantly reduces the reasoning time of the large model, but also realizes the rational allocation of computing resources in the intelligent agent. Through this innovative method, the present invention effectively improves the overall operating efficiency of the intelligent agent and provides new ideas and solutions for the optimization of large-scale artificial intelligence systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 A flowchart of a task decomposition processing method based on a large language model in an embodiment of the present application;

[0024] Figure 2 This is a schematic diagram of a task disassembly prompt word template in an embodiment of the present application;

[0025] Figure 3 This is a structural diagram of a task decomposition processing device based on a large language model in an embodiment of the present application;

[0026] Figure 4 It is a structural diagram of a task decomposition processing device based on a large language model in an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0028] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0029] Figure 1A flowchart of a task decomposition processing method based on a large language model is provided for one or more embodiments of this specification. The method can be applied to different business fields, such as the Internet financial business field, the e-commerce business field, the instant messaging business field, the game business field, the official business field, etc. The process can be executed by a computing device in the corresponding field (for example, a risk control server or a smart mobile terminal corresponding to the payment business, etc.), and some input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.

[0030] The analysis method involved in the embodiments of the present application may be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For the convenience of understanding and description, the following embodiments are described in detail by taking a terminal device as an example.

[0031] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific limitations on this.

[0032] like Figure 1 As shown, the embodiment of the present application provides a task decomposition processing method based on a large language model, including:

[0033] S101: Decomposing a target task by using a large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks.

[0034] First, the user specifies the target task that needs to be performed by the large language model, and then the server uses the large language model to deeply analyze the input task and automatically identify its category and difficulty level. Then the target task is decomposed into multiple subtasks and the subtask information corresponding to the multiple subtasks is obtained.

[0035] Among them, the task decomposition prompt word template is as follows Figure 2 As shown in the figure. With this comprehensive task understanding, the system can make more intelligent decisions and lay the foundation for subsequent task processing. When encountering complex tasks, it can perform detailed task decomposition and decompose them into multiple independent but interrelated subtasks. This decomposition not only takes into account the logical structure of the task, but also fully considers the characteristics and resource requirements of each subtask, providing precise guidance for subsequent dynamic calls and resource allocation.

[0036] S102: Determine, based on the subtask information, the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the plurality of subtasks.

[0037] The subtask information includes the step numbers, task descriptions, prerequisite step numbers, task categories, and task difficulties corresponding to multiple subtasks. Task categories include TOOL type and Chat type. TOOL type tasks are tasks that need to be completed by calling various execution tools, and sometimes may require a combination of multiple skills to complete.

[0038] S103: Determine the execution models corresponding to the plurality of subtasks respectively based on the task descriptions, task categories and task difficulties corresponding to the plurality of subtasks respectively.

[0039] After determining the subtask information corresponding to each subtask, corresponding execution models may be determined for the multiple subtasks according to the subtask information. Specifically, when determining the execution model, the task description, task category, and task difficulty corresponding to the subtask need to be considered.

[0040] S104: Using the execution model, executing the multiple subtasks to obtain sub-execution results of the multiple subtasks.

[0041] The most appropriate model is selected for each subtask. The selection process takes into account the characteristics and complexity of the subtask, as well as the currently available model resources. The selector weighs the model's capabilities and resource consumption, striving to find the best balance between performance and efficiency. Then, the execution model is used to execute the corresponding subtask to obtain the sub-execution results of multiple subtasks.

[0042] S105: Integrate the sub-execution results of the multiple sub-tasks to obtain the execution result of the target task.

[0043] When all subtasks are completed, the result integrator collects the outputs of each subtask. It analyzes the relationship between these outputs and integrates them intelligently. This process involves optimizing the results, removing redundant information, and reconstructing logical relationships to ensure the quality and coherence of the final output.

[0044] Finally, the system returns the integrated results to the user. At the same time, the system will analyze the entire process of this task execution, including resource utilization efficiency, task completion quality, etc., and use this information for continuous optimization of the system.

[0045] In one embodiment, when determining the execution model of a subtask, it is necessary to select any subtask from multiple subtasks as the target subtask; based on the task description and task category of the target subtask, determine the execution tool corresponding to the target subtask; based on the task difficulty corresponding to the target subtask, determine the model size of the execution model corresponding to the target subtask; based on the execution tool and the model size, select the target execution model from the preset model pool; traverse multiple subtasks until the corresponding execution models of multiple subtasks are determined.

[0046] In one embodiment, the preset model pool includes skill model pools with different parameter scales and specialization directions, such as online search engines, retrieval models for searching databases in professional fields such as law and medicine, weather query models, and large language models of different sizes that have been fine-tuned in various professional fields. Each model in the model pool has its own specific advantages and applicable scenarios, and is used to complete reasoning generation tasks in different scenarios.

[0047] In one embodiment, when using an execution model to execute a subtask, there may be a situation where an execution model that is adapted to the size of the target subtask is in a called state. If you insist on using this execution model, you need to wait until the current task of the execution model is completed before processing the target subtask, which will cause thread blockage and delay the overall processing flow of the target task. To avoid this situation, if an execution model that is adapted to the size is called, a model with a similar size can be selected as the model to be executed, that is, the backup model. Specifically, when executing the target subtask, determine whether the ideal model is in a called state; if the ideal model is in a called state, execute the target subtask through the model to be executed; if the ideal model is in an idle state, execute the target subtask through the ideal model.

[0048] In one embodiment, using an execution model, before executing the multiple subtasks, it is also necessary to determine the execution order of the subtasks. Specifically, the multiple single-line execution orders corresponding to the multiple subtasks can be determined based on the step numbers and the preceding step numbers corresponding to the multiple subtasks. The single-line execution order here refers to the execution order in which multiple subtasks can be connected in series in order to achieve a certain function. For example, in order to achieve a certain early warning function, it is necessary to execute subtask B, and the result or intermediate parameters of subtask A are required to execute subtask B. At this time, the single-line execution order corresponding to the early warning function is AB. Then, based on the multiple single-line execution orders corresponding to the multiple subtasks and the number of idle threads, the serial-parallel execution sequence of the multiple subtasks can be determined. The serial-parallel execution sequence here refers to multiple single-line execution sequences that can be executed simultaneously according to the number of idle threads. Then, according to the serial-parallel execution sequence and the execution model, multiple subtasks can be executed.

[0049] Furthermore, when integrating the sub-execution results of multiple sub-tasks, the execution order of the sub-tasks also needs to be considered. The intermediate execution results can be determined from the sub-execution results of the multiple sub-tasks according to the serial-parallel execution sequence. The intermediate execution result here refers to the execution result used as the input of the next execution model. The intermediate execution result is input into the execution model of the next sub-task according to the serial-parallel execution sequence until the sequence results corresponding to the multiple execution sequences are obtained; the sequence results corresponding to the multiple execution sequences are integrated to obtain the execution result of the target task.

[0050] like Figure 3 As shown, the embodiment of the present application also provides a task processing method device based on a large language model, including:

[0051] The decomposition module 301 decomposes the target task through a large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks.

[0052] The subtask information module 302 determines the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the plurality of subtasks, respectively, based on the subtask information.

[0053] The execution model module 303 determines the execution models corresponding to the plurality of subtasks respectively based on the task descriptions, task categories and task difficulties respectively corresponding to the plurality of subtasks.

[0054] The execution module 304 executes the multiple subtasks using the execution model to obtain sub-execution results of the multiple subtasks.

[0055] The integration module 305 integrates the sub-execution results of the multiple sub-tasks to obtain the execution result of the target task.

[0056] like Figure 4 As shown, the embodiment of the present application also provides a task decomposition processing device based on a large language model, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:

[0057] Decompose the target task through a large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; determine the step numbers, task descriptions, pre-step numbers, task categories and task difficulties corresponding to the multiple subtasks according to the subtask information; determine the execution models corresponding to the multiple subtasks based on the task descriptions, task categories and task difficulties corresponding to the multiple subtasks; use the execution model to execute the multiple subtasks to obtain sub-execution results of the multiple subtasks; integrate the sub-execution results of the multiple subtasks to obtain the execution result of the target task.

[0058] The embodiment of the present application further provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured as follows:

[0059] Decompose the target task through a large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; determine the step numbers, task descriptions, pre-step numbers, task categories and task difficulties corresponding to the multiple subtasks according to the subtask information; determine the execution models corresponding to the multiple subtasks based on the task descriptions, task categories and task difficulties corresponding to the multiple subtasks; use the execution model to execute the multiple subtasks to obtain sub-execution results of the multiple subtasks; integrate the sub-execution results of the multiple subtasks to obtain the execution result of the target task.

[0060] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0061] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0062] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0063] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0064] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0066] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0067] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0068] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include transitory media such as modulated data signals and carrier waves.

[0069] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0070] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A task decomposition processing method based on a large language model, characterized in that: include: Decomposing the target task through the large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; Determine, based on the subtask information, the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the multiple subtasks respectively; Determining execution models corresponding to the plurality of subtasks respectively based on task descriptions, task categories, and task difficulties respectively corresponding to the plurality of subtasks; Using the execution model, executing the multiple subtasks to obtain sub-execution results of the multiple subtasks; The sub-execution results of the multiple sub-tasks are integrated to obtain the execution result of the target task.

2. The method according to claim 1, characterized in that Determining execution models corresponding to the plurality of subtasks based on the task descriptions, task categories, and task difficulties corresponding to the plurality of subtasks respectively, specifically includes: Select any subtask from the multiple subtasks as a target subtask; Determine an execution tool corresponding to executing the target subtask based on the task description and task category of the target subtask; Determining a model size of an execution model corresponding to the target subtask based on a task difficulty corresponding to the target subtask; Based on the execution tool and the model size, selecting a target execution model from a preset model pool; The multiple subtasks are traversed until the corresponding execution models are determined for the multiple subtasks.

3. The method according to claim 2, characterized in that The preset model pool maintains a plurality of large language models with different specialization directions and different parameter scales.

4. The method according to claim 2, characterized in that: The using the execution model to execute the multiple subtasks specifically includes: Taking the execution model of the preset model pool with the model size corresponding to the target subtask as the ideal model of the target subtask; If the ideal model is in a called state when determining the execution model of the target subtask, based on the task difficulty and model size of the target subtask, a preset model with a similar model size is selected from the preset model pool as the model to be executed for the target subtask; When executing the target subtask, determining whether the ideal model is in a called state; If the ideal model is in a called state, the target subtask is executed through the model to be executed; If the ideal model is in an idle state, the target subtask is executed by the ideal model.

5. The method according to claim 1, characterized in that The using the execution model, before executing the plurality of subtasks, specifically includes: Determine the execution order of multiple single lines corresponding to the multiple subtasks based on the step numbers and the preceding step numbers respectively corresponding to the multiple subtasks; Determining a serial-parallel execution sequence of the multiple subtasks based on the multiple single-line execution orders corresponding to the multiple subtasks and the number of idle threads; The plurality of subtasks are executed based on the serial-parallel execution sequence and the execution model.

6. The method according to claim 5, characterized in that The integrating the sub-execution results of the plurality of sub-tasks to obtain the execution result of the target task specifically includes: Determining an intermediate execution result from the sub-execution results of the plurality of sub-tasks according to the serial parallel execution sequence; Inputting the intermediate execution result into the execution model of the next subtask according to the serial-parallel execution sequence until sequence results corresponding to the plurality of execution sequences are obtained; The sequence results corresponding to the multiple execution sequences are integrated to obtain the execution result of the target task.

7. The method according to claim 1, characterized in that The task categories include TOOL type and Chat type. The TOOL type tasks are tasks that need to be completed by calling various execution tools.

8. A task processing method and device based on a large language model, characterized in that: include: A disassembly module, which disassembles the target task through a large language model to obtain multiple subtasks and subtask information corresponding to the multiple subtasks; A subtask information module, which determines the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the plurality of subtasks, respectively, based on the subtask information; An execution model module, which determines the execution models corresponding to the plurality of subtasks respectively based on the task descriptions, task categories and task difficulties respectively corresponding to the plurality of subtasks; An execution module, using the execution model, executes the plurality of subtasks to obtain sub-execution results of the plurality of subtasks; The integration module integrates the sub-execution results of the multiple sub-tasks to obtain the execution result of the target task.

9. A task decomposition processing device based on a large language model, characterized in that: include: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute: Decomposing the target task through the large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks; Determine, based on the subtask information, the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the multiple subtasks respectively; Determining execution models corresponding to the plurality of subtasks respectively based on task descriptions, task categories, and task difficulties respectively corresponding to the plurality of subtasks; Using the execution model, executing the multiple subtasks to obtain sub-execution results of the multiple subtasks; The sub-execution results of the multiple sub-tasks are integrated to obtain the execution result of the target task.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Decomposing the target task through the large language model to obtain a plurality of subtasks and subtask information corresponding to the plurality of subtasks; Determine, based on the subtask information, the step numbers, task descriptions, pre-step numbers, task categories, and task difficulties corresponding to the multiple subtasks respectively; Determining execution models corresponding to the plurality of subtasks respectively based on task descriptions, task categories, and task difficulties respectively corresponding to the plurality of subtasks; Using the execution model, executing the multiple subtasks to obtain sub-execution results of the multiple subtasks; The sub-execution results of the multiple sub-tasks are integrated to obtain the execution result of the target task.

Images