User document generation method and device, medium and equipment

Abstract

The invention provides a user document generation method and device, a medium and equipment, and relates to the technical field of artificial intelligence, in particular to the technical field of deep learning, natural language processing and POC. According to the implementation scheme, in response to a received first demand for a target model, first information of the target model is extracted from the first demand; based on the first information, obtaining a resource link used for deploying deployment resources of the target model; and processing the first information, the resource link and the first prompt text based on the first language model to enable the first language model to generate a target user document based on the guidance of the first prompt text, including: recalling at least one first document fragment corresponding to the at least one content demand in a first database based on the first information; and generating a target user document based on the structure demand, the at least one first document fragment, the first information and the resource link.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the field of deep learning, natural language processing, and proof of concept (POC) technology, and specifically to a user document generation method, device, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] Artificial intelligence is a discipline that studies how to use computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.). It includes both hardware-level and software-level technologies. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, and big data processing; artificial intelligence software technologies mainly include computer vision technology, speech recognition technology, natural language processing technology, as well as machine learning / deep learning, big data processing technology, knowledge graph technology, and other major directions.

[0003] User documentation is used to provide users with detailed guidance on how to use, configure, and deploy models. By providing sufficient documentation support, it can reduce the learning cost of users when using the model and improve the ease of use and acceptability of the model. The production of user documentation is an important part of the POC work of deep learning models.

[0004] The methods described in this section are not necessarily methods that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any method described in this section is considered to be prior art simply because it is included in this section. Similarly, unless otherwise indicated, the issues mentioned in this section should not be considered to have been recognized in any prior art. Summary of the invention

[0005] The present disclosure provides a user document generation method, device, electronic device, computer-readable storage medium, and computer program product.

[0006] According to one aspect of the present disclosure, a user document generation method is provided, comprising: in response to receiving a first requirement for a target model, extracting first information of the target model from the first requirement, wherein the first requirement includes a production requirement for a user document of the target model, the user document is used to guide the deployment of the target model, the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; based on the first information, obtaining a resource link of a deployment resource for deploying the target model; and processing the first information, the resource link and a first prompt text based on a first language model, so that the first language model generates a target user document based on the guidance of the first prompt text, the first prompt text includes a structural requirement and at least one content requirement for the user document, and the operations performed by the first language model include: based on the first information, recalling at least one first document fragment corresponding to at least one content requirement in a first database, the first database includes at least one user document fragment containing corresponding second information, the second information includes model information and deployment environment information, and the semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and generating a target user document based on the structural requirement, the at least one first document fragment, the first information and the resource link.

[0007] According to another aspect of the present disclosure, a user document generation device is provided, comprising: a first extraction unit, configured to extract first information of the target model from the first requirement in response to receiving a first requirement for the target model, wherein the first requirement includes a production requirement of a user document of the target model, the user document is used to guide the deployment of the target model, the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; a first acquisition unit, configured to acquire a resource link of a deployment resource for deploying the target model based on the first information; and a first generation unit, configured to generate the first information, the resource link and the deployment environment information based on a first language model. The method further comprises the steps of: processing a first prompt text and a first hint text so that a first language model generates a target user document based on the guidance of the first prompt text, wherein the first prompt text includes a structural requirement and at least one content requirement for the user document, and the operations performed by the first language model include: based on the first information, recalling at least one first document fragment corresponding to the at least one content requirement in a first database, wherein the first database includes at least one user document fragment containing corresponding second information, the second information includes model information and deployment environment information, and the semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and generating a target user document based on the structural requirement, the at least one first document fragment, the first information, and the resource link.

[0008] According to another aspect of the present disclosure, an electronic device is provided, 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 the above-mentioned user document generation method.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the above-mentioned user document generating method.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements the above-mentioned user document generating method when executed by a processor.

[0011] According to one or more embodiments of the present disclosure, it is possible to realize automatic generation of target user documents for deploying a target model, thereby improving the efficiency of producing user documents.

[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 accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0014] Figure 1 A schematic diagram showing an exemplary system in which the various methods described herein may be implemented according to an embodiment of the present disclosure;

[0015] Figure 2 A flowchart of a method for generating a user document according to an embodiment of the present disclosure is shown;

[0016] Figure 3 A flowchart of a method for generating a user document according to an embodiment of the present disclosure is shown;

[0017] Figure 4 A flowchart of a method for generating a user document according to an exemplary embodiment of the present disclosure is shown;

[0018] Figure 5 A structural block diagram of a user document generating device according to an embodiment of the present disclosure is shown;

[0019] Figure 6A structural block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0020] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0021] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, temporal relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same instance of the element, and in some cases, based on the description of the context, they may also refer to different instances.

[0022] The terms used in the description of various examples in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. In addition, the term "and / or" used in this disclosure covers any one of the listed items and all possible combinations.

[0023] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0024] Figure 1 FIG. 1 is a schematic diagram of an exemplary system 100 in which various methods and apparatuses described herein may be implemented according to an embodiment of the present disclosure. Figure 1 , the system 100 includes one or more client devices 101, 102, 103, 104, 105, and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105, and 106 may be configured to execute one or more applications.

[0025] In an embodiment of the present disclosure, the server 120 may run one or more services or software applications that enable the user document generation method of the present disclosure to be executed.

[0026] In some embodiments, server 120 may also provide other services or software applications, which may include non-virtualized environments and virtualized environments. In some embodiments, these services may be provided as web-based services or cloud services, such as provided to users of client devices 101, 102, 103, 104, 105, and / or 106 under a software as a service (SaaS) model.

[0027] exist Figure 1 In the configuration shown, the server 120 may include one or more components that implement the functions performed by the server 120. These components may include software components, hardware components, or a combination thereof that can be executed by one or more processors. Users operating client devices 101, 102, 103, 104, 105, and / or 106 may in turn utilize one or more client applications to interact with the server 120 to utilize the services provided by these components. It should be understood that a variety of different system configurations are possible, which may differ from the system 100. Therefore, Figure 1 is one example of a system for implementing the various methods described herein and is not intended to be limiting.

[0028] The user may use client devices 101, 102, 103, 104, 105 and / or 106 to send a user document generation request. The client device may provide an interface that enables the user of the client device to interact with the client device. The client device may also output information to the user via the interface. Figure 1 Only six client devices are depicted, but one skilled in the art will appreciate that the present disclosure may support any number of client devices.

[0029] Client devices 101, 102, 103, 104, 105 and / or 106 may include various types of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptop computers), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, game systems, thin clients, various messaging devices, sensors or other sensing devices, etc. These computer devices may run various types and versions of software applications and operating systems, such as MICROSOFT Windows, APPLE iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as GOOGLE Chrome OS); or include various mobile operating systems, such as MICROSOFT Windows Mobile OS, iOS, Windows Phone, Android. Portable handheld devices may include cellular phones, smart phones, tablet computers, personal digital assistants (PDAs), etc. Wearable devices may include head-mounted displays (such as smart glasses) and other devices. Game systems may include various handheld game devices, Internet-enabled game devices, etc. Client devices are capable of executing various different applications, such as various Internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and may use various communication protocols.

[0030] The network 110 may be any type of network known to those skilled in the art that may support data communications using any of a variety of available protocols, including but not limited to TCP / IP, SNA, IPX, etc. By way of example only, the one or more networks 110 may be a local area network (LAN), an Ethernet-based network, a token ring, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0031] Server 120 may include one or more general purpose computers, dedicated server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running virtual operating systems, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that may be virtualized to maintain a server's virtual storage device). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0032] The computing units in the server 120 may run one or more operating systems including any of the above operating systems and any commercially available server operating systems. The server 120 may also run any of a variety of additional server applications and / or middle-tier applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0033] In some implementations, server 120 may include one or more applications to analyze and consolidate data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and / or 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and / or 106.

[0034] In some embodiments, the server 120 may be a server of a distributed system, or a server combined with a blockchain. The server 120 may also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in a cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services.

[0035] The system 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. The databases 130 may reside in various locations. For example, the database used by the server 120 may be local to the server 120, or may be remote from the server 120 and may communicate with the server 120 via a network-based or dedicated connection. The databases 130 may be of different types. In some embodiments, the databases used by the server 120 may be, for example, relational databases. One or more of these databases may store, update, and retrieve data to and from the databases in response to commands.

[0036] In some embodiments, one or more of the databases 130 may also be used by applications to store application data. The databases used by the applications may be different types of databases, such as a key-value store, an object store, or a conventional store backed by a file system.

[0037] Figure 1The system 100 may be configured and operated in various ways to enable the application of various methods and apparatuses described in the present disclosure.

[0038] According to the embodiments of the present disclosure, Figure 2 As shown, a user document generation method is provided, comprising: step S201, in response to receiving a first requirement for a target model, extracting first information of the target model from the first requirement, wherein the first requirement includes a production requirement of a user document of the target model, the user document is used to guide the deployment of the target model, the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; step S202, based on the first information, obtaining a resource link of a deployment resource for deploying the target model; and step S203, processing the first information, the resource link and the first prompt text based on a first language model so that the first The language model generates a target user document based on the guidance of a first prompt text, wherein the first prompt text includes structural requirements and at least one content requirement for the user document, and the operations performed by the first language model include: step S2031, based on the first information, recalling at least one first document fragment corresponding to at least one content requirement in the first database, the first database including at least one user document fragment containing corresponding second information, the second information including model information and deployment environment information, and the semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and step S2032, generating a target user document based on the structural requirements, at least one first document fragment, the first information and the resource link.

[0039] Therefore, after receiving the first requirement for user document production sent by the user, the model information and deployment environment information of the target model in the first requirement are first extracted, and the resource link of the deployment resource is obtained; then, the first language model is applied, and based on the first information and resource link and the guidance for content requirements in the first prompt text, at least one user document fragment is recalled in the first database, and then based on the reference to the user document fragment and the requirements for user document structure in the first prompt text, the target user document is generated, thereby realizing the automatic generation of the target user document for deploying the target model and improving the efficiency of user document production.

[0040] User documentation can be used to guide users in deploying the target model, including detailed guidance on how to use, configure, and deploy the model. The target model can be any deep learning model, without limitation.

[0041] In some embodiments, the first requirement may be a requirement for making a user document of the target model. The requirement may be issued by the user in the form of a requirement document or a requirement text, which may include relevant information of the target model. In some embodiments, the first requirement may also include requirement information for the user document, such as content requirements and structure requirements.

[0042] In some embodiments, in response to receiving a first requirement for a target model, extracting the first information of the target model from the first requirement may be processing the first requirement through a pre-trained keyword extraction model to extract keywords (i.e., the first information) therein. The keyword extraction model may be obtained by training based on sample requirement texts with keywords annotated.

[0043] In some embodiments, in response to receiving a first requirement for a target model, extracting the first information of the target model from the first requirement may also be inputting the first requirement together with a corresponding prompt text into a large language model, and the large language model, under the guidance of the prompt text, performs semantic analysis on the first requirement and extracts the first information. The prompt text may be, for example, a prompt text for guiding the large language model to extract information related to the target model from the first requirement.

[0044] In some embodiments, the first information may include model information and deployment environment information of the target model.

[0045] Among them, the model information may include information such as model type and parameter quantity; the deployment environment information may include hardware environment information and software environment information, wherein the hardware environment information may include, for example, the model and performance parameters of the central processing unit (CPU), the model and performance parameters of the graphics processing unit (GPU) and parameters of other types of hardware, and the software environment information may include, for example, the type and version of the operating system, the type and version of the software on which the model runs, and other information. It is understandable that the specific information contained in the above-mentioned first information can be determined based on actual needs and is not limited here.

[0046] After the first information is extracted, a resource link of a deployment resource for deploying the target model may be acquired based on the first information.

[0047] The deployment resources may include the target model and other software, toolkits, data, etc. that the model depends on for running. Each deployment resource may be open source, and the resource link may be a download link for obtaining each deployment resource. It is understandable that the above deployment resources may be determined based on actual needs and are not limited here.

[0048] After obtaining the resource link, the first information, the resource link and the first prompt text may be further processed based on the first language model, so that the first language model generates a target user document based on the guidance of the first prompt text.

[0049] In some embodiments, the first language model may be a pre-trained generative model for generating user documents.

[0050] In some embodiments, the first language model can also be a generative large language model, which is obtained through training with corpus data of a preset scale. It has deep semantic understanding and generation capabilities across modalities and languages, and has broader application scenarios in many fields such as search question and answer, content creation and generation, and smart office.

[0051] In some embodiments, the first prompt text may be pre-set, which may include structural requirements and at least one content requirement for the user document. For example, the first prompt text may include text for guiding the first language model to generate the user document and text for specifying what content needs to be included in the user document and the order and hierarchical relationship between each content.

[0052] In some embodiments, a plurality of first prompt texts may be preset, each prompt text may correspond to a different type of model, and each prompt text may contain different content requirements and structure requirements for different models.

[0053] In some embodiments, after the first information is obtained, a first prompt text that is most suitable for the current target model can be obtained from multiple first prompt texts through text matching or semantic similarity matching.

[0054] In some embodiments, after obtaining the first information, the first information can also be input into a large language model so that the large language model performs semantic analysis on the first information and each first prompt text to select a first prompt text that is most suitable for the current target model.

[0055] In some embodiments, the first database may include at least one user document fragment containing corresponding second information, wherein the second information is used to identify each user document fragment and includes model information and deployment environment information. The above-mentioned first database may be constructed in the following manner: firstly, a plurality of completed user documents are obtained, each of which corresponds to different model information and deployment environment information (i.e., second information); each user document is divided into a plurality of user document fragments according to the content, and each user document fragment is stored together with its corresponding second information to obtain the first database.

[0056] In some embodiments, the first database may also store semantic vectors of each user document fragment and the second information. When recalling at least one first document fragment corresponding to at least one content requirement from the first database based on the first information, one or more first document fragments whose semantic similarity between the second information and the first information meets the first preset condition may be recalled for each content requirement.

[0057] The first preset condition may be that the semantic similarity between the first information and the second information is greater than a preset threshold, or that the semantic similarity between the first information and the second information is the highest, which is not limited here.

[0058] In some embodiments, the first prompt text may further include a sentence for guiding the first language model to screen document fragments. When multiple document fragments are recalled for each content requirement, the first language model may select a document fragment that best meets the requirement for each content requirement under the guidance of the first prompt text, and generate a target user document based on the selected document fragment.

[0059] In some embodiments, generating a target user document based on structural requirements, at least one first document fragment, first information and resource links can be inputting the above-mentioned structural requirements, at least one first document fragment, first information and resource links into a first language model, so as to refer to the first document fragment through the first language model, perform semantic analysis on the first information and resource links, and organize the above-mentioned information into one document according to the structural requirements to generate a target user file.

[0060] In some embodiments, the first requirement also includes a test requirement for the target model.

[0061] In some embodiments, Figure 3 As shown, the above-mentioned user document generation method may also include: step S301, before obtaining the resource link, based on the first information, selecting a target server corresponding to the deployment environment information from multiple local servers, so as to deploy the target model on the target server according to the model information, wherein each server in the multiple servers corresponds to different deployment environment information; step S302, extracting test requirement information for the target model from the first requirement, wherein the test requirement information includes indicators to be tested and a data set for testing the target model; step S303, based on the test requirement information, obtaining a data set; step S304, based on the data set, testing the indicators to be tested of the target model to obtain a test result for the target model; and step S305, detecting whether there is an abnormality in the test result.

[0062] And wherein, based on the first information, obtaining a resource link of deployment resources for deploying the target model may be: in response to determining that there is no abnormality in the test result, packaging the deployment resources of the target model on the target server and uploading them to the cloud to obtain a resource link.

[0063] Therefore, by recalling at least one relatively similar first test report from the second database and inputting it into the second language model, the second language model refers to the information therein to detect the current test result, thereby being able to efficiently and accurately complete the abnormal detection of the test result.

[0064] In some embodiments, multiple local servers may be communicatively connected to a host for executing the user document generation method of the present disclosure, and the multiple servers may cover various hardware environment configurations and software environment configurations.

[0065] In some embodiments, before obtaining the resource link, the target server for deploying the target model can be selected from multiple local servers directly based on the first information, and based on the model information, relevant installation packages, dependency packages, data, etc. can be downloaded to complete the deployment of the target model on the target server.

[0066] In some embodiments, the above operations can be completed manually. In some embodiments, the above operations can also be performed by the host generating a configuration plan based on the model information, and sending the configuration plan to the target server, which automatically completes the deployment of the target model.

[0067] In some embodiments, the above-mentioned user document generation method may also include: before deploying the target model, matching in the user document database based on the first information to determine whether the user document database contains a target pre-stored user document whose pre-stored information matches the first information, wherein the user document database includes at least one pre-stored user document, and the pre-stored information includes model information and deployment environment information.

[0068] And wherein, based on the first information, selecting a target server corresponding to the deployment environment information from multiple local servers to deploy the target model on the target server according to the model information includes: in response to the user document database not containing the target pre-stored user document, based on the first information, selecting a target server corresponding to the deployment environment information from multiple servers to deploy the target model on the target server according to the model information.

[0069] Therefore, after receiving the first demand sent by the user and extracting the first information therein, a match is first performed in the user document database, and only when no reusable user document is matched, the target model is deployed locally and subsequent work is performed, thereby enabling the reuse of the generated user documents and further improving the efficiency of user document production.

[0070] In some embodiments, after obtaining the above-mentioned first information, a match may be first performed in the user document database based on the first information to determine whether a user document for the same target model (i.e., the model information and deployment environment information are the same) already exists in the user document database. The user document database stores one or more user documents that have been prepared, and each user document is annotated with its corresponding model information and deployment environment information (i.e., pre-stored information).

[0071] If a user document can be matched in the user document database, there is no need to perform subsequent user document generation work, and the matched user document can be directly fed back to the user, thereby achieving the reuse of existing user documents.

[0072] If the user document cannot be matched in the user document database, then continue to execute the deployment operation of the above target model.

[0073] In some embodiments, while the target model is being deployed, test requirement information for the target model may be extracted from the first requirement, wherein the test requirement information may include indicators to be tested and a data set for testing the target model.

[0074] In some embodiments, the test for the target model may include at least one of a performance test and an accuracy verification test. The above-mentioned indicators to be tested may include various model performance indicators and / or model accuracy indicators, which are not limited here.

[0075] In some embodiments, the above-mentioned test requirement information may further include parameters in the model testing process, including but not limited to input length, output length, batch size, number of iterations, etc., which are not limited here.

[0076] In some embodiments, extracting the test requirement information for the target model from the first requirement may be performed by processing the first requirement through a pre-trained keyword extraction model to extract the keywords (i.e., the test requirement information) therein. The keyword extraction model may be obtained by training based on sample requirement texts with the keywords annotated.

[0077] In some embodiments, extracting the test requirement information for the target model from the first requirement may also be inputting the first requirement together with a corresponding prompt text into the large language model, and the large language model, under the guidance of the prompt text, performs semantic analysis on the first requirement and extracts the above test requirement information. The above prompt text may be, for example, a prompt text for guiding the large language model to extract information related to the model test from the first requirement.

[0078] After obtaining the test requirement information, a data set for model testing can be obtained based on the test requirement information; then, after the above preparations are completed, the corresponding test tools (such as at least one of the performance testing tools and the accuracy verification tool) can be called to perform corresponding tests on the target model to obtain the test results of the target model.

[0079] The test results may then be examined to determine whether there are any abnormalities in the test results.

[0080] In some embodiments, the above detection can be performed manually.

[0081] In some embodiments, detecting whether a test result is abnormal may include: based on the first information, recalling at least one first test report in a second database, wherein the second database includes at least one historical test report and each historical test report corresponds to its own third information, the third information includes model information and deployment environment information, and the semantic similarity between the third information of each first test report in the at least one first test report and the first information meets a second preset condition; and inputting at least the test result, at least one first test report, and a second prompt text into a second language model, so that the second language model, under the guidance of the second prompt text, refers to the information in the at least one first test report, analyzes the test result, and obtains a first analysis result output by the second language model, wherein the first analysis result indicates whether the test result is abnormal.

[0082] Therefore, by recalling at least one relatively similar first test report from the second database and inputting it into the second language model, the second language model refers to the information therein to detect the current test result, thereby being able to efficiently and accurately complete the abnormal detection of the test result.

[0083] In some embodiments, the second database may include at least one historical test report, wherein each historical test report is a test report of a corresponding model that has completed testing and detected no abnormalities, and is annotated with model information and deployment environment information (ie, the third information).

[0084] In some embodiments, the second database may also include one or more test reports of completed tests and including annotations of abnormal results.

[0085] In some embodiments, the second database may also store each test report together with a semantic vector of the corresponding third information.

[0086] In some embodiments, at least one first test report can be obtained in the second database based on the first information. The second preset condition can be that the semantic similarity between the third information and the first information is greater than a preset threshold, or one or more test reports with the highest semantic similarity between the third information and the first information are used as at least one first test report.

[0087] In some embodiments, the second language model may be a pre-trained language model for detecting abnormal test results, which may be trained based on labeled sample test results, at least one sample test report, and a second prompt text.

[0088] In some embodiments, the second language model may also be a generative large language model that is the same as the first language model described above.

[0089] In some embodiments, the second prompt text may be an instruction text for guiding the second language model to refer to the information in the first test report to analyze and detect anomalies on the current test result.

[0090] In some embodiments, the test results, at least one first test report and a second prompt text may be input into a second language model, so that the second language model analyzes the test results under the guidance of the second prompt text and with reference to information in the at least one first test report.

[0091] In some embodiments, the first information, test requirement information and the other information mentioned above can also be input into the second language model, so that the second language model, under the guidance of the second prompt text, refers to the information in at least one first test report, conducts a comprehensive analysis of various types of information to output a first analysis result of the test result.

[0092] When the first analysis result indicates that there is no abnormality in the current test result, the deployment resources of the target model on the target server can be further packaged and uploaded to the cloud to obtain a resource link, thereby further improving the reliability of the deployment resources.

[0093] In some embodiments, the user document generation method may further include: in response to determining that there is no abnormality in the test result, generating a test report of the target model through a fourth language model based on the test result.

[0094] Therefore, by applying the fourth language model, a test report is generated based on the test results, thereby improving the readability of the test results and enhancing the user experience.

[0095] In some embodiments, the fourth language model may be a pre-trained generative language model, which may be trained based on sample test results and corresponding sample test reports.

[0096] In some embodiments, the fourth language model may also be a generative large language model that is the same as the first language model described above.

[0097] In some embodiments, the test report that passes the detection may be input into the fourth language model to obtain the test report outputted by the fourth language model.

[0098] In some embodiments, the test result, at least one first test report and a prompt text may be input into the fourth language model together, so that the fourth language model, under the guidance of the prompt text, refers to the content and structure in the first test report, semantically understands and organizes the test result, and thus generates a corresponding test report.

[0099] In some embodiments, after obtaining the test report, it can be stored together with the corresponding first information in the second database to update the second database, thereby providing more reference data for subsequent abnormal test result detection.

[0100] In some embodiments, the above-mentioned user document generation method may also include: in response to determining that there is an abnormality in the test result, based on the first information, recalling at least one first abnormal test result in the third database, wherein the third database includes at least one historical abnormal test result, and each historical abnormal test result includes corresponding fourth information, abnormal cause and abnormal handling solution, the fourth information includes model information and deployment environment information, and the semantic similarity between the fourth information of each of the at least one first abnormal test result and the first information meets the third preset condition; and inputting the test result, at least one first abnormal test result and the third prompt text into the third language model, so that the third language model, under the guidance of the third prompt text, refers to the abnormal cause and abnormal handling solution of each of the at least one first abnormal test result, to generate an abnormal cause analysis result and an abnormal handling solution for the test result.

[0101] Therefore, by recalling at least one relatively similar first abnormal test result from the third database and inputting it into the third language model, the third language model refers to its abnormal cause and abnormal handling solution, analyzes the abnormal cause and generates a handling solution for the current abnormal test result, thereby being able to efficiently and accurately handle the abnormal test result.

[0102] In some embodiments, at least one historical abnormal test result is stored in the third database, and each historical abnormal test result is annotated with corresponding model information and deployment environment information (ie, fourth information), abnormal cause and abnormal handling solution.

[0103] In some embodiments, when the first analysis result indicates that there is an abnormality in the test result, at least one first abnormal test result can be obtained in the third database based on the first information. The third preset condition can be that the semantic similarity between the fourth information and the first information is greater than a preset threshold, or one or more abnormal test results with the highest semantic similarity between the fourth information and the first information are used as at least one first abnormal test result.

[0104] In some embodiments, the third language model may be a pre-trained generative language model for performing exception cause analysis and exception handling solution generation, which may be trained based on sample test results annotated with exception causes and exception handling solutions and a third prompt text.

[0105] In some embodiments, the third language model may also be a generative large language model that is the same as the first language model.

[0106] In some embodiments, the third prompt text may be an instruction text for guiding the third language model to refer to the abnormal causes and abnormal handling solutions of each first abnormal test result, and to analyze the abnormal causes and generate a handling solution for the current test result.

[0107] In some embodiments, the test results, at least one first abnormal test result, and a third prompt text can be input into a third language model so that the third language model, under the guidance of the third prompt text, refers to the abnormal cause and abnormal handling solution of each of the at least one first abnormal test result to generate an abnormal cause analysis result and an abnormal handling solution for the test result.

[0108] In some embodiments, the first information, test requirements and the other information mentioned above can also be input into a third language model, so that the third language model, under the guidance of the third prompt text, refers to each abnormal cause and abnormal handling solution in at least one first abnormal test result, conducts a comprehensive analysis of various types of information, and generates abnormal cause analysis results and abnormal handling solutions for the test results.

[0109] In some embodiments, after obtaining the exception handling solution, corresponding processing can be performed accordingly, and the above-mentioned test operation and exception detection operation can be performed again until it is determined that there is no exception in the test result, and then subsequent work can be continued.

[0110] In some embodiments, processing the first information, resource links, and first prompt text based on the first language model so that the first language model generates a target user document based on the guidance of the first prompt text may include: processing the first information, resource links, test reports, and first prompt text based on the first language model so that the first language model generates a target user document based on the guidance of the first prompt text.

[0111] Therefore, by inputting the test report together with the above-mentioned other information into the first language model, the target user document can further include model test result information, thereby improving the comprehensiveness of the content in the user document; in addition, by first organizing the test results into a test report and then inputting the test report into the first language model, the processing difficulty of the language model can be further reduced, thereby improving the efficiency and accuracy of user document production.

[0112] In some embodiments, generating a target user document based on structural requirements, at least one first document fragment, first information, and resource links may include: based on structural requirements, at least one first document fragment, and fourth prompt text, allowing the first language model to generate a user document template based on structural requirements and at least one first document fragment under the guidance of the fourth prompt text; and generating a target user document based on at least the first user document, the first information, the resource link, the test report, and the user document template.

[0113] Therefore, by first generating a user document template based on at least one first document fragment and structural requirements, and then generating a target user document based on the user document template through the first language model, the accuracy of the user document can be further improved.

[0114] In some embodiments, the content requirements in the first prompt text may include requirements for content related to model configuration and deployment and requirements for content related to model testing. When multiple first prompt texts are preset, the first prompt text that is most suitable for the current first requirement can be recalled based on the first information and the test report by matching semantic similarity or selecting through a large language model.

[0115] In some embodiments, the fourth prompt text may be an independent prompt text or may be included in the first prompt text, and is used to guide the first language model to organize the at least one first document fragment into a user document template according to structural requirements.

[0116] In some embodiments, after the above-mentioned first information, resource link, test report and first prompt text are input into the first language model, the first language model can first recall one or more first document fragments for each content requirement in the first database based on the first information, resource link and test report under the guidance of the first prompt text.

[0117] Subsequently, the first language model may further, under the guidance of the fourth prompt text in the first prompt text and according to the structural requirements, perform semantic understanding on the at least one first document fragment and integrate it into a user document template.

[0118] On this basis, under the guidance of the first prompt text, the first language model can further fill the first information, resource link, and corresponding content in the test report into the corresponding position of the user document template based on the user document template to generate a target user document.

[0119] In some embodiments, the target user document may include multiple steps for deploying and testing the target model and the target execution results of each step. The above user document generation method may also include: for each of the multiple steps, performing the following operations: executing the step to obtain the execution result of the step; and comparing the execution result of the step with the target execution result corresponding to the step to verify whether there is an abnormality in the execution result of the step; and in response to judging that there is no abnormality in the execution result of each of the multiple steps, saving the target user document corresponding to the first information in the user document database.

[0120] In some embodiments, the target user document may include multiple steps for deploying and testing the target model and target execution results of each step. When executing each step, each step may be written as an operation instruction for computer execution based on the large language model, and then each operation instruction is executed to obtain the corresponding operation result.

[0121] In some embodiments, the target user document may also include operation instructions corresponding to the above steps. When executing each step, the operation instructions corresponding to each step may be directly executed and the corresponding operation results may be obtained.

[0122] Therefore, after the target user document is generated, the execution results of each step are verified, so that the validity of the target user document can be further guaranteed; and the reuse of the user document can be achieved by saving the verified target user document in the user document database.

[0123] In some embodiments, the user document generation method may further include: in response to determining that there is an abnormal execution result in the first step of the multiple steps, based on the first information and the abnormal information of the first step, recalling at least one first abnormal information in a fourth database, wherein the fourth database includes at least one historical abnormal information, and each historical abnormal information includes corresponding fifth information, abnormal cause and abnormal handling solution, the fifth information includes model information and deployment environment information, and the semantic similarity between the fifth information of each of the at least one first abnormal information and the first information meets the fourth preset condition; and

[0124] The exception information of the first step, at least one first exception information and the fifth prompt text are input into the fourth language model, so that the fourth language model, under the guidance of the fifth prompt text, refers to the exception cause and exception handling solution of each of the at least one first exception information, and generates an exception cause analysis result and an exception handling solution for the exception information of the first step.

[0125] Therefore, by recalling the more similar first exception information from the fourth database and inputting it into the fourth language model, the fourth language model refers to its exception cause and exception handling solution, analyzes the exception cause and generates a handling solution for the current exception execution result, thereby being able to efficiently and accurately handle the exception execution result.

[0126] In some embodiments, at least one historical exception information is stored in the fourth database, and each historical exception information is annotated with corresponding model information and deployment environment information (ie, fifth information), exception cause and exception handling solution.

[0127] In some embodiments, when an execution result of a certain step is abnormal, at least one first abnormal information can be obtained in the fourth database based on the above-mentioned first information. The fourth preset condition can be that the semantic similarity between the fifth information and the first information is greater than a preset threshold, or one or more historical abnormal information with the highest semantic similarity between the fifth information and the first information is used as the at least one first abnormal information.

[0128] In some embodiments, the fourth language model may be a pre-trained generative language model for performing exception cause analysis and exception handling solution generation, which may be trained based on sample exception information annotated with exception causes and exception handling solutions and the fifth prompt text.

[0129] In some embodiments, the fourth language model may also be a generative large language model that is the same as the first language model described above.

[0130] In some embodiments, the fifth prompt text may be an instruction text for guiding the fourth language model to refer to the exception causes and exception handling solutions of each first exception information, and to perform exception cause analysis and handling solution generation on the current exception execution result.

[0131] In some embodiments, the exception information of the first step, at least one first exception information and the fifth prompt text can be input into the fourth language model, so that the fourth language model, under the guidance of the fifth prompt text, refers to the exception cause and exception handling solution of each of the at least one first exception information, and generates an exception cause analysis result and an exception handling solution for the abnormal execution result of the first step.

[0132] In some embodiments, after obtaining the exception handling solution, corresponding processing can be performed accordingly, and the above steps can be executed again, and it can be verified whether the execution result is correct, until it is determined that there is no exception in the execution result, and then the subsequent work can be continued.

[0133] In some exemplary embodiments, the first requirement may be a proof of concept (POC) requirement for a target model.

[0134] The POC work of the target model is a comprehensive and complex task, which mainly includes the following tasks:

[0135] 1) Accuracy verification: When doing POC work, you first need to verify the accuracy of the model. This usually involves using a test data set to evaluate the predictive ability of the model. Accuracy verification may include cross-validation, holdout validation, or other validation techniques to ensure the generalization ability of the model on different data sets. By comparing the performance of different models or algorithms, the best model configuration can be selected.

[0136] 2) Performance testing: In addition to accuracy, the performance of the model is also an aspect that needs to be focused on in POC. Performance testing aims to evaluate the efficiency, response time, and resource consumption of the model when processing real data. Performance testing can include testing the model's inference speed, memory usage, and performance under different hardware configurations. Performance testing helps identify potential bottlenecks and optimization points to ensure that the model can run efficiently in a production environment.

[0137] 3) Deployment package preparation: Once the accuracy and performance of the model are verified, the next step is to prepare the deployment package. This usually involves packaging the model and its dependencies into an executable image or program. The preparation of the deployment package needs to take into account the characteristics and requirements of the target environment to ensure that the model can run smoothly on different platforms. In addition, security and maintainability need to be considered, such as protecting model files through encryption and signing, and providing necessary logging and monitoring functions.

[0138] 4) User documentation: User documentation is an integral part of the POC process. It provides users with detailed guidance on how to use, configure, and deploy the model. The user documentation should clearly explain the model's input and output formats, parameter settings, expected performance, and possible problems and solutions. By providing adequate documentation support, the user's learning cost when using the model can be reduced, and the model's ease of use and acceptability can be improved.

[0139] 5) Integration testing: During the POC process, you also need to consider the integration of the model with other systems. This may include docking with databases, API interfaces, or front-end applications. Integration testing aims to ensure that the model can work seamlessly with other components to achieve the expected functions and performance. Through automated testing tools and methods, potential problems and errors can be efficiently detected and fixed.

[0140] Through accuracy verification, performance testing, deployment package preparation, user documentation writing, and integration testing, the reliability, efficiency, and ease of use of the model in practical applications can be ensured.

[0141] Figure 4 A flowchart of a method for generating a user document according to an exemplary embodiment of the present disclosure is shown.

[0142] In some exemplary embodiments, Figure 4As shown, the user document generation method may include: step S401, in response to receiving a POC requirement, searching in the user document database whether there is a reusable user document based on the information in the POC requirement; step S402, in response to not finding a reusable user document in the user document database, deploying the target model on the local target server based on the model deployment module, and providing a calling interface to the agent (Agent); step S403, the agent performs performance testing and accuracy verification on the target model based on the POC requirement by calling the target model and the accuracy verification tool and the performance testing tool to obtain a performance test report and an accuracy verification report; step S404, detecting whether there is an abnormality in the results of the performance test report and the accuracy verification report; step S405, if there is an abnormality, obtaining an abnormality handler Step S406: If there is no exception, the deployment resources (image, data, program code, model, etc.) are packaged through the packaging module; Step S407: Upload the packaged resources to the cloud storage; Step S408: Based on the document generation module, call the large language model and generate user documents through the method described above; Step S409: Send the user document to the agent to verify the execution results of each step; Step S410: Manually verify the execution results of each step; Step S411: If there is no exception in the verification, save the user document to the user document database; Step S412: If there is an exception in the verification, obtain the exception handling plan to handle the exception, repackage the resources and generate the user document.

[0143] In some embodiments, Figure 5As shown, a user document generation device 500 is provided, comprising: a first extraction unit 510, configured to extract first information of the target model from the first requirement in response to receiving a first requirement for the target model, wherein the first requirement includes a production requirement of a user document of the target model, the user document is used to guide the deployment of the target model, the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; a first acquisition unit 520, configured to acquire a resource link of a deployment resource for deploying the target model based on the first information; and a first generation unit 530, configured to generate a resource link of a deployment resource for deploying the target model based on a first language model. The source link and the first prompt text are processed so that the first language model generates a target user document based on the guidance of the first prompt text, the first prompt text includes a structural requirement and at least one content requirement for the user document, and the operations performed by the first language model include: based on the first information, recalling at least one first document fragment corresponding to the at least one content requirement in the first database, the first database includes at least one user document fragment containing corresponding second information, the second information includes model information and deployment environment information, and the semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and generating the target user document based on the structural requirement, the at least one first document fragment, the first information and the resource link.

[0144] The operations performed by the units 510 to 530 in the user document generating device 500 are similar to the operations of steps S201 to S203 in the user document generating method, and are not described in detail herein.

[0145] In some embodiments, the first requirement may also include a test requirement for the target model, and the above-mentioned user document generation device may also include: a deployment unit, configured to select a target server corresponding to the deployment environment information from multiple local servers based on the first information before obtaining the resource link, so as to deploy the target model on the target server according to the model information, wherein each of the multiple servers corresponds to different deployment environment information; a second extraction unit, configured to extract test requirement information for the target model from the first requirement, wherein the test requirement information includes indicators to be tested and a data set for testing the target model; a second acquisition unit, configured to acquire a data set based on the test requirement information; a third acquisition unit, configured to test the indicators to be tested of the target model based on the data set to obtain a test result for the target model; and a first detection unit, configured to detect whether there is an abnormality in the test result; and wherein the first acquisition unit can be further configured to: in response to judging that there is no abnormality in the test result, package the deployment resources of the target model on the target server and upload them to the cloud to obtain a resource link.

[0146] In some embodiments, the first detection unit may include: a first recall subunit, configured to recall at least one first test report in a second database based on the first information, wherein the second database includes at least one historical test report and each historical test report corresponds to its own third information, the third information includes model information and deployment environment information, and the semantic similarity between the third information of each first test report in at least one first test report and the first information meets a second preset condition; and a first acquisition subunit, configured to input at least the test result, at least one first test report, and a second prompt text into a second language model, so that the second language model, under the guidance of the second prompt text, refers to the information in the at least one first test report, analyzes the test result, and obtains a first analysis result output by the second language model, wherein the first analysis result indicates whether there is an abnormality in the test result.

[0147] In some embodiments, the user document generation device may further include: a first recall unit, configured to recall at least one first abnormal test result in a third database based on the first information in response to determining that the test result is abnormal, wherein the third database includes at least one historical abnormal test result, and each historical abnormal test result includes corresponding fourth information, abnormal cause and abnormal handling solution, the fourth information includes model information and deployment environment information, and the semantic similarity between the fourth information of each of the at least one first abnormal test result and the first information meets a third preset condition; and a second generation unit, configured to input the test result, at least one abnormal test result and a third prompt text into a third language model, so that the third language model, under the guidance of the third prompt text, refers to the abnormal cause and abnormal handling solution of each of the at least one abnormal test result, and generates an abnormal cause analysis result and an abnormal handling solution for the test result.

[0148] In some embodiments, the user document generating apparatus may further include: a third generating unit configured to generate a test report of the target model through a fourth language model based on the test result in response to determining that the test result does not have an abnormality.

[0149] In some embodiments, the first generating unit may be further configured to: process the first information, the resource link, the test report and the first prompt text based on the first language model, so that the first language model generates a target user document based on the guidance of the first prompt text.

[0150] In some embodiments, generating a target user document based on structural requirements, at least one first document fragment, first information, and resource links may include: based on structural requirements, at least one first document fragment, and fourth prompt text, allowing the first language model to generate a user document template based on structural requirements and at least one first document fragment under the guidance of the fourth prompt text; and generating a target user document based on at least the first user document, the first information, the resource link, the test report, and the user document template.

[0151] According to an embodiment of the present disclosure, an electronic device, a readable storage medium and a computer program product are also provided.

[0152] refer to Figure 6 , a block diagram of an electronic device 600 that can be used as a server or client of the present disclosure will now be described, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, 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.

[0153] like Figure 6 As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0154] Multiple components in the electronic device 600 are connected to the I / O interface 605, including: an input unit 606, an output unit 607, a storage unit 608, and a communication unit 609. The input unit 606 can be any type of device that can input information to the electronic device 600. The input unit 606 can receive input digital or character information and generate key signal input related to user settings and / or function control of the electronic device, and can include but is not limited to a mouse, a keyboard, a touch screen, a track pad, a track ball, a joystick, a microphone, and / or a remote controller. The output unit 607 can be any type of device that can present information, and can include but is not limited to a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 608 can include but is not limited to a disk, an optical disk. The communication unit 609 allows the electronic device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include but is not limited to a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, an 802.11 device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0155] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 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 601 performs the various methods and processes described above, such as the above-mentioned user document generation method. For example, in some embodiments, the above-mentioned user document generation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the above-mentioned user document generation method may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the above-mentioned user document generation method by any other appropriate means (e.g., by means of firmware).

[0156] 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.

[0157] 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.

[0158] 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 equipment, 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.

[0159] 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).

[0160] 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.

[0161] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0162] 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 described in this disclosure can be performed 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.

[0163] Although the embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it should be understood that the above-mentioned methods, systems and devices are merely exemplary embodiments or examples, and the scope of the present invention is not limited by these embodiments or examples, but only by the claims after authorization and their equivalent scope. Various elements in the embodiments or examples can be omitted or replaced by their equivalent elements. In addition, each step can be performed in an order different from that described in the present disclosure. Further, the various elements in the embodiments or examples can be combined in various ways. It is important that with the evolution of technology, many elements described herein can be replaced by equivalent elements that appear after the present disclosure.

Claims

1. A method for generating a user document, include: In response to receiving a first requirement for a target model, extracting first information of the target model from the first requirement, wherein the first requirement includes a requirement for making a user document of the target model, the user document is used to guide the deployment of the target model, and the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; Based on the first information, obtaining a resource link of a deployment resource for deploying the target model; and The first information, the resource link, and the first prompt text are processed based on a first language model, so that the first language model generates a target user document based on the guidance of the first prompt text, wherein the first prompt text includes a structural requirement and at least one content requirement for the user document, and the operations performed by the first language model include: Based on the first information, recalling at least one first document fragment corresponding to the at least one content requirement in a first database, the first database comprising at least one user document fragment containing corresponding second information, the second information comprising model information and deployment environment information, and a semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and The target user document is generated based on the structural requirement, the at least one first document fragment, the first information, and the resource link.

2. The method according to claim 1, in, The first requirement also includes a test requirement for the target model, and the method further includes: Before acquiring the resource link, based on the first information, selecting a target server corresponding to the deployment environment information from a plurality of local servers, so as to deploy the target model on the target server according to the model information, wherein each server in the plurality of servers corresponds to different deployment environment information; Extracting test requirement information for the target model from the first requirement, wherein the test requirement information includes an indicator to be tested and a data set for testing the target model; Based on the test requirement information, acquiring the data set; Based on the data set, testing the to-be-tested indicator of the target model to obtain a test result for the target model; and Detecting whether the test result is abnormal; and wherein, The acquiring, based on the first information, a resource link of a deployment resource for deploying the target model comprises: In response to determining that there is no abnormality in the test result, the deployment resources of the target model on the target server are packaged and uploaded to the cloud to obtain the resource link.

3. The method according to claim 2, in, The detecting whether the test result is abnormal comprises: Based on the first information, recall at least one first test report in a second database, wherein the second database includes at least one historical test report and each historical test report corresponds to respective third information, the third information includes model information and deployment environment information, and the semantic similarity between the third information of each first test report in the at least one first test report and the first information meets a second preset condition; and At least the test result, the at least one first test report and the second prompt text are input into a second language model, so that the second language model, under the guidance of the second prompt text, refers to the information in the at least one first test report, analyzes the test result, and obtains a first analysis result output by the second language model, wherein the first analysis result indicates whether there is an abnormality in the test result.

4. The method according to claim 2 or 3, further comprising: include: In response to determining that the test result is abnormal, based on the first information, recalling at least one first abnormal test result in a third database, wherein the third database includes at least one historical abnormal test result, and each historical abnormal test result includes corresponding fourth information, an abnormal cause, and an abnormal handling solution, the fourth information includes model information and deployment environment information, and the semantic similarity between the fourth information of each of the at least one first abnormal test result and the first information meets a third preset condition; and The test result, the at least one first abnormal test result and the third prompt text are input into a third language model, so that the third language model, under the guidance of the third prompt text, refers to the abnormal cause and abnormal handling solution of each of the at least one first abnormal test result, and generates an abnormal cause analysis result and an abnormal handling solution for the test result.

5. The method according to any one of claims 2 to 4, further comprising: include: In response to determining that there is no abnormality in the test result, a test report of the target model is generated by a fourth language model based on the test result.

6. The method according to claim 5, in, The processing of the first information, the resource link, and the first prompt text based on the first language model so that the first language model generates a target user document based on the guidance of the first prompt text includes: The first information, the resource link, the test report and the first prompt text are processed based on the first language model, so that the first language model generates the target user document based on the guidance of the first prompt text.

7. The method according to claim 6, in, The generating the target user document based on the structural requirement, the at least one first document fragment, the first information and the resource link comprises: Based on the structural requirement, the at least one first document fragment and the fourth prompt text, enabling the first language model to generate a user document template based on the structural requirement and the at least one first document fragment under the guidance of the fourth prompt text; and The target user document is generated based at least on the first user document, the first information, the resource link, the test report, and the user document template.

8. The method according to claim 6 or 7, in, The target user document includes multiple steps for deploying and testing the target model and target execution results of each step. The method further includes: For each of the multiple steps, perform the following operations: Execute the step to obtain the execution result of the step; and Compare the execution result of the step with the target execution result corresponding to the step to verify whether there is any abnormality in the execution result of the step; and In response to determining that there is no abnormality in the execution result of each of the multiple steps, the target user document and the first information are stored in a user document database correspondingly.

9. The method according to claim 8, further comprising: include: In response to determining that there is a first step in the multiple steps with an abnormal execution result, at least one first exception information is recalled from a fourth database based on the first information and the exception information of the first step, wherein the fourth database includes at least one historical exception information, and each historical exception information includes corresponding fifth information, an exception cause and an exception handling solution, the fifth information includes model information and deployment environment information, and the semantic similarity between the fifth information of each of the at least one first exception information and the first information meets a fourth preset condition; and The exception information of the first step, the at least one first exception information and the fifth prompt text are input into a fourth language model, so that the fourth language model, under the guidance of the fifth prompt text, refers to the exception cause and exception handling solution of each of the at least one first exception information, and generates an exception cause analysis result and an exception handling solution for the exception information of the first step.

10. The method according to any one of claims 2 to 9, further comprising: include: Before deploying the target model, matching is performed in a user document database based on the first information to determine whether the user document database contains a target pre-stored user document whose pre-stored information matches the first information, wherein the user document database includes at least one pre-stored user document, and the pre-stored information includes model information and deployment environment information; and wherein, The step of selecting a target server corresponding to the deployment environment information from a plurality of local servers based on the first information, so as to deploy the target model on the target server according to the model information includes: In response to the user document database not containing the target pre-stored user document, based on the first information, selecting the target server corresponding to the deployment environment information from the multiple servers to deploy the target model on the target server according to the model information.

11. A user document generating device, include: A first extraction unit is configured to, in response to receiving a first requirement for a target model, extract first information of the target model from the first requirement, wherein the first requirement includes a requirement for making a user document of the target model, the user document is used to guide the deployment of the target model, and the first information includes model information and deployment environment information of the target model, the model information includes a model type and a parameter amount, and the deployment environment information includes hardware environment information and software environment information; A first acquisition unit is configured to acquire a resource link of a deployment resource for deploying the target model based on the first information; and A first generating unit is configured to process the first information, the resource link, and a first prompt text based on a first language model, so that the first language model generates a target user document based on the guidance of the first prompt text, wherein the first prompt text includes a structural requirement and at least one content requirement for the user document, and the operations performed by the first language model include: Based on the first information, recalling at least one first document fragment corresponding to the at least one content requirement in a first database, the first database comprising at least one user document fragment containing corresponding second information, the second information comprising model information and deployment environment information, and a semantic similarity between the second information of each first document fragment and the first information meets a first preset condition; and The target user document is generated based on the structural requirement, the at least one first document fragment, the first information, and the resource link.

12. The device according to claim 11, in, The first requirement also includes a test requirement for the target model, and the device further includes: a deployment unit configured to select, before acquiring the resource link, a target server corresponding to the deployment environment information from a plurality of local servers based on the first information, so as to deploy the target model on the target server according to the model information, wherein each server in the plurality of servers corresponds to different deployment environment information; A second extraction unit is configured to extract test requirement information for the target model from the first requirement, wherein the test requirement information includes an indicator to be tested and a data set for testing the target model; A second acquisition unit is configured to acquire the data set based on the test requirement information; a third acquisition unit, configured to test the to-be-tested indicator of the target model based on the data set to obtain a test result for the target model; and The first detection unit is configured to detect whether the test result is abnormal; and wherein, The first acquisition unit is further configured to: in response to determining that there is no abnormality in the test result, package the deployment resources of the target model on the target server and upload them to the cloud to obtain the resource link.

13. The device according to claim 12, in, The first detection unit comprises: A first recall subunit is configured to recall at least one first test report in a second database based on the first information, wherein the second database includes at least one historical test report and each historical test report corresponds to respective third information, the third information includes model information and deployment environment information, and a semantic similarity between the third information of each first test report in the at least one first test report and the first information meets a second preset condition; and The first acquisition subunit is configured to input at least the test result, the at least one first test report and the second prompt text into a second language model, so that the second language model, under the guidance of the second prompt text, refers to the information in the at least one first test report to analyze the test result, and obtain a first analysis result output by the second language model, wherein the first analysis result indicates whether the test result is abnormal.

14. The device according to claim 12 or 13, further comprising: include: a first recall unit configured to, in response to determining that the test result is abnormal, recall at least one first abnormal test result in a third database based on the first information, wherein the third database includes at least one historical abnormal test result, and each historical abnormal test result includes corresponding fourth information, an abnormal cause, and an abnormal handling solution, the fourth information includes model information and deployment environment information, and a semantic similarity between the fourth information of each of the at least one first abnormal test result and the first information meets a third preset condition; and The second generating unit is configured to input the test result, the at least one first abnormal test result and the third prompt text into a third language model, so that the third language model, under the guidance of the third prompt text, refers to the abnormal cause and abnormal handling solution of each of the at least one first abnormal test result to generate an abnormal cause analysis result and an abnormal handling solution for the test result.

15. The device according to any one of claims 12 to 14, further comprising: include: The third generating unit is configured to generate a test report of the target model by using a fourth language model based on the test result in response to judging that the test result does not have an abnormality.

16. The device according to claim 15, in, The first generating unit is further configured to: The first information, the resource link, the test report and the first prompt text are processed based on the first language model, so that the first language model generates the target user document based on the guidance of the first prompt text.

17. The device according to claim 16, in, The generating the target user document based on the structural requirement, the at least one first document fragment, the first information and the resource link comprises: Based on the structural requirement, the at least one first document fragment and the fourth prompt text, enabling the first language model to generate a user document template based on the structural requirement and the at least one first document fragment under the guidance of the fourth prompt text; and The target user document is generated based at least on the first user document, the first information, the resource link, the test report, and the user document template.

18. An electronic device, include: at least one processor; as well as a memory communicatively coupled to the at least one processor; in The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

19. A non-transitory computer-readable storage medium storing computer instructions, in, The computer instructions are used to make a computer execute the method according to any one of claims 1-10.

20. A computer program product comprising a computer program, in, When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images