Dynamic parameter setting method and system for large language model

Abstract

The invention discloses a parameter dynamic setting method and system for a large language model, and the method comprises the steps: dividing a business demand into a plurality of specific business types, distributing a corresponding business plug-in for each type of business based on the business types, and distinguishing the business plug-ins through a URL or a parameter; setting a model parameter and a prompt instruction template corresponding to each type of business plug-in, and determining a parameter range of each model parameter; obtaining a service request submitted by a user, determining the category of the service request, and configuring model parameters and a prompt instruction based on the category of the service request; and forming a model request parameter based on the configured model parameter and the prompt instruction, and calling the large language model based on the model request parameter. According to the method, the flexibility and response speed of service processing are greatly improved, more importantly, parameter optimization can be completed on the premise that existing service calling is not interrupted, and smooth transition of service continuity and user experience is ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to a method and system for dynamically setting parameters of a large language model. Background Art

[0002] In recent years, with the rapid development of artificial intelligence technology, large language models (LLM) have shown great potential and broad application prospects in the field of natural language processing. These models are trained with massive text data to generate high-quality text content and complete a variety of complex tasks such as text generation, summary extraction, and sentiment analysis. However, in actual applications, different business scenarios need to complete the writing of parameters and prompt instructions according to the business when calling large language models. Complex business scenarios and parameter configurations pose many challenges to model calling.

[0003] A common practice is to hardcode specific model call parameters and prompt templates in the application to adapt to different functional scenarios. For example, in some text generation systems, developers will set the model's sampling temperature, prompts and other parameters directly in the code according to business needs to guide the model to generate specific types of text. However, this approach has obvious limitations: when business needs change, the application needs to be manually modified and redeployed, resulting in high development and maintenance costs; at the same time, the hard-coded method lacks flexibility and is difficult to meet the needs of diverse and rapidly changing scenarios; in addition, due to the fixed nature of parameters and prompts, it is difficult for the system to dynamically adjust parameters based on user feedback or business changes, thereby limiting the further improvement of model performance.

[0004] Therefore, a method for dynamically setting parameters of a large language model is needed. Summary of the invention

[0005] The present invention provides a method and system for dynamically setting parameters of a large language model to solve the problem of how to dynamically set parameters of a large language model.

[0006] In order to solve the above problem, according to one aspect of the present invention, a method for dynamically setting parameters of a large language model is provided, the method comprising:

[0007] Divide business requirements into multiple specific business categories, assign corresponding business plug-ins to each business category based on the business category, and distinguish business plug-ins through URLs or parameters;

[0008] Set the model parameters and prompt instruction template corresponding to each type of business plug-in, and determine the parameter range of each model parameter;

[0009] Obtaining a service request submitted by a user, determining the category of the service request, and configuring model parameters and prompt instructions based on the category of the service request;

[0010] Model request parameters are formed based on the configured model parameters and prompt instructions, so as to call the large language model based on the model request parameters.

[0011] Preferably, the method performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a public parameter layer and a plug-in parameter layer.

[0012] Preferably, the common parameter layer is provided with basic model parameters, including: input and output length limits and learning rates, which are used as common configurations for all service plug-ins;

[0013] The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

[0014] Preferably, the method further comprises:

[0015] Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction;

[0016] When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

[0017] Preferably, the key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

[0018] According to another aspect of the present invention, a system for dynamically setting parameters of a large language model is provided, the system comprising:

[0019] A service plug-in allocation unit is used to divide service requirements into multiple specific service categories, allocate corresponding service plug-ins to each service category based on the service category, and distinguish service plug-ins by URL or parameters;

[0020] Model parameter and setting unit, used to set the model parameters and prompt instruction template corresponding to each type of business plug-in, and determine the parameter range of each model parameter;

[0021] A configuration unit, used to obtain a service request submitted by a user, determine the category of the service request, and configure model parameters and prompt instructions based on the category of the service request;

[0022] The model request parameter determination unit is used to form the model request parameters based on the configured model parameters and the prompt instruction, so as to call the large language model based on the model request parameters.

[0023] Preferably, the model parameter and setting unit performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a public parameter layer and a plug-in parameter layer.

[0024] Preferably, the common parameter layer is provided with basic model parameters, including: input and output length limits and learning rates, which are used as common configurations for all service plug-ins;

[0025] The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

[0026] Preferably, the system further comprises: an updating unit, configured to:

[0027] Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction;

[0028] When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

[0029] Preferably, the key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

[0030] The present invention provides a method and system for dynamically setting parameters of a large language model, including: dividing business requirements into multiple specific business categories, assigning corresponding business plug-ins to each type of business based on the business category, and distinguishing the business plug-ins through URLs or parameters; setting model parameters and prompt instruction templates corresponding to each type of business plug-in, and determining the parameter range of each model parameter; obtaining a business request submitted by a user, determining the category of the business request, and configuring the model parameters and prompt instructions based on the category of the business request; forming model request parameters based on the configured model parameters and prompt instructions, so as to call a large language model based on the model request parameters. The parameter dynamic generation method of the present invention supports the real-time dynamic adjustment of configuration parameters by the server, and can automatically or manually trigger a parameter adjustment mechanism based on factors such as user feedback evaluation, model performance monitoring and algorithm upgrades during the continuous operation of business services, so as to ensure that the parameter settings are always kept in the optimal state. This greatly improves the flexibility and response speed of business processing, and more importantly, it can complete parameter optimization without interrupting existing business calls, ensuring business continuity and smooth transition of user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0032] Figure 1 is a flowchart of a method 100 for dynamically setting parameters of a large language model according to an embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a dynamic parameter update process according to an embodiment of the present invention;

[0034] Figure 3 It is a structural schematic diagram of a system 300 for dynamically setting parameters of a large language model according to an embodiment of the present invention;

[0035] Figure 4 FIG. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] Now, exemplary embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to disclose the present invention in detail and completely and to fully convey the scope of the present invention to those skilled in the art. The terms used in the exemplary embodiments shown in the accompanying drawings are not intended to limit the present invention. In the accompanying drawings, the same units / elements are marked with the same reference numerals.

[0037] Unless otherwise specified, the terms (including technical terms) used herein have the commonly understood meanings to those skilled in the art. In addition, it is understood that the terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0038] The present invention provides a method for dynamically setting parameters of a large language model, which performs parameter setting and instruction assembly on large model requests at the intermediate business layer, and performs request forwarding processing. Specifically, the intermediate business layer can classify requests according to business rules, form different business plug-in service interfaces, and define unique model parameters and prompt instructions for different business plug-in categories. With the continuous improvement and optimization of the business, the device can also dynamically update the model parameters according to the feedback content, and all of this can be dynamically configured on the server side without affecting the application docking.

[0039] Figure 1 FIG. 1 is a flow chart of a method 100 for dynamically setting parameters of a large language model according to an embodiment of the present invention. Figure 1 As shown, the method for dynamically setting parameters of a large language model provided by an embodiment of the present invention supports real-time dynamic adjustment of configuration parameters by the server, and can automatically or manually trigger the parameter adjustment mechanism based on factors such as user feedback evaluation, model performance monitoring, and algorithm upgrades during the continuous operation of business services, so as to ensure that the parameter settings are always kept in the optimal state. This greatly improves the flexibility and response speed of business processing. More importantly, it can complete parameter optimization without interrupting existing business calls, ensuring business continuity and a smooth transition of user experience. The method 100 for dynamically setting parameters of a large language model provided by an embodiment of the present invention starts from step 101. In step 101, business requirements are divided into multiple specific business categories, corresponding business plug-ins are assigned to each type of business based on the business category, and business plug-ins are distinguished by URL or parameters.

[0040] In step 102, model parameters and prompt instruction templates corresponding to each type of service plug-in are set, and a parameter range of each model parameter is determined.

[0041] Preferably, the method performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a public parameter layer and a plug-in parameter layer.

[0042] Preferably, the common parameter layer is provided with basic model parameters, including: input and output length limits and learning rates, which are used as common configurations for all service plug-ins;

[0043] The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

[0044] In step 103, a service request submitted by a user is obtained, the category of the service request is determined, and model parameters and prompt instructions are configured based on the category of the service request.

[0045] In step 104, model request parameters are formed based on the configured model parameters and the prompt instruction, so as to call the large language model based on the model request parameters.

[0046] In the present invention, it is first necessary to determine the request classification according to the business category; and set the model parameters of different business categories.

[0047] In the present invention, in order to improve the output effect of the model, it is first divided according to different business categories. The extensive and complex business needs are refined into multiple specific business categories, such as text-based question-answering systems, efficient text summary generation, complex text data conversion into intuitive chart presentations, and automatic information filling based on preset templates. This classification method not only reflects an in-depth understanding of different business scenarios, but also lays a solid foundation for the customized configuration of subsequent model parameters. After determining the business classification, we assign a unique plug-in to each business and distinguish them through different URLs or parameters. This design realizes isolation and expansion between businesses. At the same time, plug-ins can be quickly updated or replaced according to changes in business needs, without the need for large-scale reconstruction or upgrade of the entire system, making the access of new businesses or the adjustment of old businesses simple and fast.

[0048] After determining the business plug-ins, we tailored the corresponding model parameters and prompt instructions for each type of business plug-in to ensure that the model can perform at its best when processing specific business requests. When setting parameters, in order to improve efficiency and scalability, we conducted a two-layer parameter design: the public parameter layer and the plug-in parameter layer.

[0049] Among them, in the public parameter layer, a set of basic model parameters, such as input and output length limits, learning rate, etc., are set as common configurations for all business plug-ins to improve the uniformity and efficiency of parameter management.

[0050] In the plug-in parameter layer, for each business plug-in, a parameter set specific to the business is designed, allowing plug-in parameters to override common parameters to achieve a more refined control strategy. For example, in the information filling plug-in, the input and output length limits can be extended to 30,000 to meet the data processing requirements unique to the plug-in.

[0051] At the same time, after determining the business scope of the request, we design prompt instruction templates that are closely related to the specific functions of the business plug-in, such as icon generation and information filling. These templates not only improve the pertinence of the instructions, but also reduce interference between cross-business plug-ins. Within the same business plug-in, a parameter passing mechanism is designed to allow prompt instructions to be dynamically adjusted and expanded according to specific business requests, achieving more accurate business matching and instruction execution. For example, in the information filling plug-in, the background parameters and templates of the information filling are passed as parameters to facilitate information filling in different business scenarios.

[0052] For example, in the design of the information filling plug-in, in order to allow the model to extract content more quickly and accurately, while taking into account dynamic settings, background information and filling templates are designed as plug-in parameters. The reference instructions are set as follows:

[0053]

[0054] The application side can pass background information and fill in templates to the plug-in based on its own business scenarios. The plug-in will splice parameters and finally form the prompt command as follows:

[0055]

[0056]

[0057] Preferably, the method further comprises:

[0058] Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction;

[0059] When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

[0060] Preferably, the key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

[0061] The present invention can also dynamically update service parameters based on model operation status and service feedback. Figure 2As shown, in order to maintain the continuous optimization of model performance and the continuous improvement of adaptability, the present invention innovatively introduces a dynamic update mechanism based on model operation and business feedback. This mechanism monitors key indicators (such as response time, accuracy, user satisfaction, etc.) during the operation of the model and conducts a comprehensive analysis in combination with user feedback (such as evaluation, suggestions, etc.). When the satisfaction of the answer is low, the model can be trained and optimized in combination with the reasons, and the plug-in parameters or prompt instruction template can be adjusted at the same time to achieve the purpose of improving the answer effect. Based on the device of the present invention, only one business plug-in can be optimized without affecting the operation of other business plug-ins. After the test optimization is completed, the optimized parameters or prompt instruction template are updated through the parameter update device. Without modifying any code on the application side, the model input optimization can be completed, providing users with more efficient, accurate and personalized services.

[0062] The key points of the present invention are:

[0063] 1) By classifying requests according to business, user intent can be identified more accurately. The model parameters and prompt instructions configured based on business rules form the final model request parameters to ensure that parameter splicing meets output expectations.

[0064] 2) Dynamic parameter adjustment device: During the model calling process, the present invention will collect user feedback and performance indicator data in real time, such as user satisfaction, output effect, etc. By analyzing the evaluation data and optimizing the latest parameters, the configuration can be directly implemented on the application side without affecting the docking application call.

[0065] The present invention determines the request category through precise business division, achieving a deep understanding and precise identification of user intentions. Compared with traditional processing schemes, this step not only simplifies the complexity of request processing, reduces the main intent identification steps, and accurately distinguishes the nature of the business, which can better understand and answer the business. At the same time, it also ensures that the configuration of model parameters can closely fit the actual business needs. The model parameters and prompt instructions carefully configured based on business rules form efficient and expected model request parameters, thereby greatly improving the accuracy and pertinence of the model output.

[0066] In addition, the present invention also realizes the optimization and dynamic adjustment of model parameters during the calling process. The device can be based on user feedback and performance indicator data, such as user satisfaction, output effect and other key information. Model optimization is performed through business analysis, and model parameter adjustment or prompt instruction template adjustment can be performed simultaneously. Compared with traditional processing methods, parameters can be dynamically set and processed on the server side without any modification or adjustment to the existing docking application, ensuring the stable operation and seamless upgrade of the system.

[0067] Figure 3 FIG. 3 is a schematic diagram of a system 300 for dynamically setting parameters of a large language model according to an embodiment of the present invention. Figure 3 As shown, a large language model parameter dynamic setting system 300 provided by an embodiment of the present invention includes: a service plug-in allocation unit 301, a model parameter and setting unit 302, a configuration unit 303 and a model request parameter determination unit 304.

[0068] Preferably, the service plug-in allocation unit 301 is used to divide the service requirements into a plurality of specific service categories, allocate a corresponding service plug-in to each service category based on the service category, and distinguish the service plug-ins through URLs or parameters.

[0069] Preferably, the model parameter setting unit 302 is used to set the model parameters and prompt instruction template corresponding to each type of business plug-in, and determine the parameter range of each model parameter.

[0070] Preferably, the model parameter setting unit 302 performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a common parameter layer and a plug-in parameter layer.

[0071] Preferably, the common parameter layer is provided with basic model parameters, including: input and output length limits and learning rates, which are used as common configurations for all service plug-ins;

[0072] The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

[0073] Preferably, the configuration unit 303 is used to obtain a service request submitted by a user, determine the category of the service request, and configure model parameters and prompt instructions based on the category of the service request.

[0074] Preferably, the model request parameter determining unit 304 is used to form model request parameters based on the configured model parameters and prompt instructions, so as to call the large language model based on the model request parameters.

[0075] Preferably, the system further comprises: an updating unit, configured to:

[0076] Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction;

[0077] When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

[0078] Preferably, the key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

[0079] The system 300 for dynamically setting parameters of a large language model according to an embodiment of the present invention corresponds to the method 100 for dynamically setting parameters of a large language model according to another embodiment of the present invention, and will not be described in detail herein.

[0080] Figure 4 The electronic device provided by an exemplary embodiment of the present invention may be any one or both of the first device and the second device, or a stand-alone device independent of them, and the stand-alone device may communicate with the first device and the second device to receive the collected input signals from them. Figure 4 FIG. 1 is a block diagram of an electronic device according to an embodiment of the present disclosure. Figure 4 As shown, the electronic device 400 includes one or more processors 401 and a memory 402 .

[0081] The processor 401 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0082] The memory 402 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 401 may run the program instructions to implement the method of the method for dynamically setting parameters of the large language model of the software program of each embodiment of the present disclosure described above and / or other desired functions. In one example, the electronic device may also include: an input device 403 and an output device 404, which are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0083] In addition, the input device 403 may also include, for example, a keyboard, a mouse, and the like.

[0084] The output device 404 can output various information to the outside. The output device 604 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto.

[0085] Of course, to simplify, Figure 4 Only some of the components related to the present disclosure in the electronic device are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device may further include any other appropriate components.

[0086] Exemplary computer program products and computer-readable storage media

[0087] In addition to the above-mentioned methods and devices, an embodiment of the present disclosure may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the method for dynamically setting parameters of a large language model according to various embodiments of the present disclosure described in the above-mentioned "Exemplary Method" section of this specification.

[0088] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the disclosed embodiments, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0089] In addition, an embodiment of the present disclosure may also be a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, causes the processor to execute the steps of the method for dynamically setting parameters of a large language model according to various embodiments of the present disclosure described in the above “Exemplary Method” section of this specification.

[0090] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable 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 above.

[0091] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0092] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system embodiment, since it basically corresponds to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0093] The block diagrams of the devices, apparatuses, equipment, and systems involved in this disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including," "comprising," "having," and the like are open words, referring to "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or," and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0094] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0095] It should also be noted that in the apparatus, equipment and method of the present disclosure, each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present disclosure. The above description of the disclosed aspects is provided to enable any technician in the field to make or use the present disclosure. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown here, but to the widest scope consistent with the principles and novel features disclosed herein.

[0096] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A method for dynamically setting parameters of a large language model, characterized in that: The method comprises: Divide business requirements into multiple specific business categories, assign corresponding business plug-ins to each business category based on the business category, and distinguish business plug-ins through URLs or parameters; Set the model parameters and prompt instruction template corresponding to each type of business plug-in, and determine the parameter range of each model parameter; Obtaining a service request submitted by a user, determining the category of the service request, and configuring model parameters and prompt instructions based on the category of the service request; Model request parameters are formed based on the configured model parameters and prompt instructions, so as to call the large language model based on the model request parameters.

2. The method according to claim 1, characterized in that The method performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a public parameter layer and a plug-in parameter layer.

3. The method according to claim 2, characterized in that The common parameter layer is provided with basic model parameters, including: input and output length limits and learning rate, which are used as common configurations for all business plug-ins; The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

4. The method according to claim 1, characterized in that: The method further comprises: Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction; When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

5. The method according to claim 4, characterized in that The key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

6. A system for dynamically setting parameters of a large language model, characterized in that: The system comprises: A service plug-in allocation unit is used to divide service requirements into multiple specific service categories, allocate corresponding service plug-ins to each service category based on the service category, and distinguish service plug-ins by URL or parameters; Model parameter and setting unit, used to set the model parameters and prompt instruction template corresponding to each type of business plug-in, and determine the parameter range of each model parameter; A configuration unit, used to obtain a service request submitted by a user, determine the category of the service request, and configure model parameters and prompt instructions based on the category of the service request; The model request parameter determination unit is used to form the model request parameters based on the configured model parameters and the prompt instruction, so as to call the large language model based on the model request parameters.

7. The system according to claim 6, characterized in that The model parameter and setting unit performs double-layer parameter design when setting plug-in parameters; wherein the double-layer parameters include: a public parameter layer and a plug-in parameter layer.

8. The system according to claim 7, characterized in that The common parameter layer is provided with basic model parameters, including: input and output length limits and learning rate, which are used as common configurations for all business plug-ins; The plug-in parameter layer is provided with dedicated service parameters of the corresponding service plug-in, and allows the dedicated service parameters to overwrite the public parameters.

9. The system according to claim 6, characterized in that The system further comprises an updating unit, configured to: Monitor key indicators during the operation of large language models, and conduct comprehensive analysis based on user feedback to determine analysis satisfaction; When the analysis satisfaction is less than or equal to a preset satisfaction threshold, the large language model is trained and optimized, and the plug-in parameters or prompt instruction template are adjusted to improve the answer effect of the large language model. When the test optimization is completed, the parameters are updated based on the optimized parameters or prompt instruction template.

10. The system according to claim 9, characterized in that The key indicators include: response time, accuracy and user satisfaction; the user feedback information includes: evaluation and suggestions.

Images