Equipment firmware upgrading risk identification method

Abstract

The invention discloses an equipment firmware upgrading risk identification method. The method comprises the following steps: acquiring upgrading information of target equipment; the upgrading information comprises equipment information of the target equipment, a firmware type of to-be-upgraded target firmware in the target equipment, current version information and upgrading target version information; generating a description text of the upgrading information according to the input template, and taking the description text as a cue word; the input template is consistent with the template of the input example; the input example is used for constructing a paradigm sample, the paradigm sample is used for constructing a training set, and the training set is used for pre-training the large model; inputting the cue word into the large model to obtain at least one model output of the upgrading information; the model outputs upgrade risk information indicating a single category of the target firmware. The risk identification process does not need manual participation, so that the automation and intelligence level of equipment firmware upgrading risk identification of the system can be improved.

Description

Technical Field

[0001] This application relates to the technical field of server management, and in particular to a method for identifying risks in device firmware upgrades. Background Art

[0002] With the wide application of cloud computing and big data technologies, the data center industry is developing rapidly and its scale is constantly expanding. For example, ultra-large data centers with tens of thousands to hundreds of thousands of servers are not uncommon. Facing such a large scale of equipment, complex system architectures, and strict security requirements, more and more experts believe that the future operation of data centers must rely on new technologies and new solutions to improve operation and maintenance efficiency and ensure high efficiency and security of operations through automation and intelligence.

[0003] Taking server firmware upgrade as an example, currently, it mainly relies on manual risk investigation before upgrade according to the operation and maintenance documents or version upgrade guides provided by manufacturers. This method has disadvantages such as high omission rate of investigation, incomplete coverage, and low efficiency.

[0004] Therefore, how to achieve automation and intelligence in risk identification for devices in the firmware upgrade application scenario in data centers has become an urgent challenge to be solved. Summary of the Invention

[0005] Embodiments of this application provide a method for identifying risks in device firmware upgrades and a computing device, which can achieve automatic risk identification for devices in the firmware upgrade application scenario.

[0006] In a first aspect, embodiments of this application provide a method for identifying risks in device firmware upgrades. The method includes: obtaining upgrade information of a target device; the upgrade information includes device information of the target device, firmware type, current version information, and upgrade target version information of the target firmware to be upgraded in the target device; generating a description text of the upgrade information according to an input template and determining a prompt word; the input template is the same as the template of the input example; the input example is used to construct an example sample, the example sample is used to construct a training set, and the training set is used to pre-train a large model; inputting the prompt word into the large model to obtain at least one model output of the upgrade information; the model output indicates upgrade risk information of a single category of the target firmware; each of the at least one model output indicates a different category.

[0007] Since this risk identification process does not require manual participation, it can improve the automation and intelligence levels of the system for identifying risks in device firmware upgrades.

[0008] In a possible implementation, the method further includes: obtaining the current device alarm information of the target device; inputting the prompt into the large model to obtain at least one model output of the upgrade information, including: inputting the device alarm information and the prompt into the large model to obtain at least one model output of the upgrade information.

[0009] Thus, the efficiency and accuracy of the system for identifying the risks of device firmware upgrade can be further improved.

[0010] In a possible implementation, the example sample further includes an output example, and the output example includes multiple fields and their field values. The multiple fields include a risk category field and a risk description field; the risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information.

[0011] Thus, the upgrade risk information of a single category of the target firmware can be obtained according to the output of the large model.

[0012] In a possible implementation, the prompt further includes at least one example sample.

[0013] In a possible implementation, the prompt further includes an output format description. The output format description includes multiple fields related to the upgrade information and their field descriptions. The multiple fields include a risk category field and a risk description field; the risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information.

[0014] Thus, by providing an output format description or an example sample including input examples and output examples in the prompt, it helps to provide a clear and consistent information output representation for the large model in data processing tasks.

[0015] In a possible implementation, obtaining the upgrade information of the target device includes: obtaining multiple upgrade information of the target device; the method further includes: classifying the model output corresponding to each upgrade information according to the firmware type of the target firmware to obtain a list of upgrade risk information of the target device.

[0016] Thus, users can more intuitively obtain the upgrade risk information of each target firmware in the target device to efficiently conduct risk investigation and handling of the firmware upgrade of the target device.

[0017] Second aspect, an embodiment of the present application provides a large model training method, including: obtaining a plurality of example samples; the example samples include input examples and output examples, the input examples are description texts of the upgrade guidance information of the first device, and the upgrade guidance information includes the device information of the first device, the firmware type, the current version information, the upgrade target version information, and the upgrade risk information of a single category of the first firmware to be upgraded in the first device; the output examples include multiple fields and their field values, the multiple fields include a risk category field and a risk description field, the risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information; the classification method of the category is predefined; training the large model with the training set constructed by the multiple example samples.

[0018] Thus, the large model can better adapt to the device firmware upgrade risk identification task.

[0019] In a possible implementation, the multiple fields further include a number of entity category fields; the number of entity category fields is used to indicate the entity information in the input example; the number of entity category fields includes the device manufacturer, the device type, the firmware type, the current firmware version, and the firmware upgrade target version.

[0020] Thus, the readability of the large model output can be improved.

[0021] In a possible implementation, the upgrade guidance information of the first device is determined by the firmware upgrade guide of the first device, and the firmware upgrade guide is released by the device manufacturer of the first device.

[0022] Thus, multiple device firmware upgrade guidance information can be obtained to construct a training set.

[0023] Third aspect, an embodiment of the present application provides a computing device, including:

[0024] At least one memory for storing programs;

[0025] At least one processor for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used to execute the methods described in the first aspect or any possible implementation of the first aspect, and the second aspect or any possible implementation of the second aspect.

[0026] Fourth aspect, an embodiment of the present application provides a computer storage medium, in which instructions are stored. When the instructions run on a computer, the computer is caused to execute the methods described in the first aspect or any possible implementation of the first aspect, and the second aspect or any possible implementation of the second aspect.

[0027] Fifth aspect, an embodiment of the present application provides a computer program product including instructions. When the instructions run on a computer, the computer is caused to execute the methods described in the first aspect or any possible implementation manner of the first aspect, and the second aspect or any possible implementation manner of the second aspect.

[0028] It can be understood that for the beneficial effects of the above third aspect to fifth aspect, reference can be made to the relevant descriptions in the above first aspect and second aspect, and details are not elaborated herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0030] Figure 1 It is an architecture diagram of a device firmware upgrade risk identification system provided by an embodiment of the present application;

[0031] Figure 2 It is a flowchart of a method for identifying risks in device firmware upgrade provided by an embodiment of the present application;

[0032] Figure 3 It is a flowchart of a large model training method provided by an embodiment of the present application;

[0033] Figure 4 It is a schematic diagram of the process of identifying risks in server firmware upgrade provided by an embodiment of the present application;

[0034] Figure 5 It is a schematic diagram of the structure of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will describe the technical solutions in the embodiments of the present application with reference to the drawings.

[0036] In the description of the embodiments of the present application, any embodiment or design solution with "exemplary", "for example", or "for instance" should not be understood as being more preferred or having more advantages than other embodiments or design solutions. Exactly, the words such as "exemplary", "for example", or "for instance" are used to present relevant concepts in a specific manner.

[0037] In addition, the terms "first" and "second" are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the technical features indicated. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "comprise", "include", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways. "At least one" can be one or more, where multiple means two or more.

[0038] Taking server firmware upgrade as an example, the server firmware upgrade in the data center refers to the process of updating and upgrading the firmware on the server hardware devices in the data center. Before performing the server firmware upgrade, the following potential risks need to be identified and addressed:

[0039] Whether it is necessary to upgrade through an intermediate version. For example, when directly upgrading from version A to version C, whether it is necessary to upgrade to version B first;

[0040] Whether the business will be interrupted during the upgrade process and whether there are incompatible configuration items after the upgrade;

[0041] Whether preprocessing is required before the upgrade. For example, whether it is necessary to turn on a specific configuration switch;

[0042] Whether there are alarms that need to be resolved first, which may affect the upgrade process;

[0043] The compatibility relationship of the firmware. For example, whether the BIOS version needs to be strongly compatible with a specific BMC (Baseboard Management Controller) version;

[0044] And other exception handling.

[0045] The above handling matters related to before the upgrade ensure that possible problems can be foreseen and resolved during the upgrade process, thus ensuring the smooth progress of the firmware upgrade.

[0046] In the field of machine learning, large models refer to large neural network models that have millions or even billions of parameters and are obtained through in-depth training with a large amount of labeled data. These models can capture and learn complex patterns and relationships in the data.

[0047] In the embodiments of the present application, a large model is pre-trained using a training set to enable the large model to adapt to the risk identification task before the device firmware upgrade. Among them, the training set includes input examples and output examples. Then, a description text of the device upgrade information input by the user is generated according to the template of the input example and used as a prompt to be input into the large model, so that the large model outputs the upgrade risk information of the device firmware in the format of the output example. Since this risk identification process does not require manual participation, the efficiency and accuracy of the system for identifying the risk of device firmware upgrade can be improved.

[0048] Exemplarily, Figure 1 FIG. 1 shows an architecture diagram of a device firmware upgrade risk identification system provided by an embodiment of the present application.

[0049] As Figure 1 shown, the device firmware upgrade risk identification system 100 includes a device management module 110, a training set management module 120, and a large model module 130, and there are Figure 1 shown connection and interface call relationships among the modules. Each module can be implemented by any computing unit, server, device, device cluster, etc. with computing and processing capabilities, and they can be deployed on the same or different computing devices with communication connections.

[0050] Based on these modules, the device firmware upgrade risk identification system 100 constructs a training set to pre-train the large model, generates a prompt according to the device upgrade information input by the user, and inputs the prompt into the large model to perform device firmware upgrade risk identification.

[0051] Specifically, before using the large model module 130 to execute the device firmware upgrade risk identification task, the training set management module 120 is used to construct a training set, which involves annotating the upgrade risks that may be encountered during the device firmware upgrade process. For example, by querying the common storage area, obtaining device firmware upgrade guidance information including device firmware upgrade manuals or other relevant materials, identifying potential device firmware upgrade risk points and annotating them. The annotation methods include manual annotation, machine annotation, or a combination of the two. Then, multiple example samples are constructed according to the annotation data, and a training set is constructed according to the multiple example samples.

[0052] The large model module 130 is used to pre-train a pre-obtained basic large model using the training set constructed by the training set management module 120 to obtain a large model that can better adapt to the device firmware upgrade risk identification task. The basic large model can be a basic neural network model, such as a BERT model, a ResNet model, a GPT series model, etc.

[0053] In the process of performing the firmware upgrade risk identification task, first, the device management module 110 needs to obtain the upgrade information of the target device by using the user interaction platform. In this process, the device management module 110 can also perform device management operations with the target device to obtain device information related to the device firmware upgrade. It is also used to generate a prompt word based on the obtained upgrade information and send the prompt word and device information to the large model module 130.

[0054] Furthermore, the large model module 130 performs reasoning based on the prompt word and device information and outputs the upgrade risk information of the target firmware to be upgraded for the target device.

[0055] Finally, the device management software 110 sorts out the upgrade risk information identified by the large model module 130 and sends a list of risk information for the target device to perform firmware upgrade to the user, so that the user can perform risk investigation and handling before the device performs firmware upgrade.

[0056] Thus, by combining data annotation, large model pre-training, generating a prompt word according to user input, and the large model performing the device firmware upgrade risk identification task according to the prompt word and device information, the automation and intelligent identification of the device firmware upgrade risk in the data center are realized.

[0057] Exemplarily, Figure 2 shows a flowchart of a device firmware upgrade risk identification method provided by an embodiment of the present application. The device firmware upgrade risk identification method can be implemented by Figure 1 the device management module 110 shown, and mainly includes the following steps:

[0058] Step S201, obtain the upgrade information of the target device. The upgrade information includes the device information of the target device, the firmware type, the current version information, and the upgrade target version information of the target firmware to be upgraded in the target device.

[0059] In one embodiment, the device management software 110 obtains the upgrade information of the target device through Figure 1 The target device is a hardware device to be upgraded with firmware, and the target firmware is the firmware to be upgraded in the target device. The upgrade information of the target device includes the device information of the target device, such as the device manufacturer, device type, etc., and also includes the firmware type of the target firmware, such as BIOS or iBMC, etc., the current version information of the target firmware, and the upgrade target version information.

[0060] In one implementation, the device management software 110 can provide a user interaction interface to obtain upgrade information of a target device. The user interaction interface represents different devices according to different IP addresses, represents each firmware included in the device according to different firmware types, and displays the device manufacturer, device type of different devices, and the current versions of each firmware. The user can specify the target device by selecting the IP address with reference to the device manufacturer and device type, specify the target firmware by selecting the firmware type, and input the target version for upgrading the target firmware in the target dialog box with reference to the current version of the target firmware.

[0061] Thus, the device management software 110 can obtain the upgrade information of the target device according to the user's selections on the user interaction interface to identify the upgrade risks of the target firmware of the target device.

[0062] It can be understood that if the user specifies multiple target firmwares for the target device, multiple upgrade information of the target device can be obtained, and each target firmware has corresponding upgrade information. In addition, the user can also specify multiple target devices in the system and specify multiple target firmwares for each target device to identify the upgrade risks of multiple firmwares of multiple devices during each device firmware upgrade risk identification process, thereby improving the efficiency of the system in identifying device firmware upgrade risks.

[0063] If the user specifies multiple target firmwares for the target device, the device management software 110 also aggregates the information input by the user through the user interaction interface to obtain a list of upgrade information of the target device.

[0064] Optionally, the device management software 110 also performs device management operations on the target device by calling the device management interface with the target device, obtains the current device alarm information of the target device to more accurately judge the upgrade risks of the device firmware. For example, during the upgrade process of some device firmwares, it is necessary to judge whether there are alarms that need to be resolved first, and these alarms may affect the upgrade process. Therefore, it is necessary to assist in identifying the upgrade risks according to the current alarm information of the device. The alarm level or alarm range of the device alarm information that needs to be obtained can be set according to actual upgrade risk identification experience, such as critical alarms, hardware alarms, etc.

[0065] Exemplarily, taking the server with the IP address of Address 1 as an example, Table 1 lists multiple upgrade information of this server aggregated by the device management software 110 and the current device alarm information.

[0066] Table 1

[0067] Server Type Server Manufacturer Firmware Type Current Version Upgrade Target Version Server Alarm Type 1 XXX iBMC Current Version 1 Target Version 1 Alarm Information 1 Type 1 XXX BIOS Current Version 2 Target Version 2 Alarm Information 1

[0068] As shown in Table 1, the user selects the firmware iBMC and BIOS of the server with the IP address of Address 1 for device upgrade. Among them, each firmware has corresponding upgrade information. For example, the first piece of upgrade information in Table 2 includes: server type: Type 1, server manufacturer: XXX, firmware type: iBMC, current version: Current Version 1, upgrade target version: Target Version 2, and it is detected that the server currently has the server alarm described in "Alarm Information 1".

[0069] Step S202, generate a description text of the upgrade information according to the input template, and determine the prompt word. The input template is the same as the template of the input example. The input example is used to construct an example sample, the example sample is used to construct a training set, and the training set is used to pre-train the large model.

[0070] In one embodiment, the device management software 110 generates corresponding prompt words according to each piece of upgrade information shown in Table 2. The prompt word is the description text of the upgrade information and is used to input into the pre-trained large model to identify the upgrade risk of the target firmware of the target device.

[0071] The large model is pre-trained based on the basic large model using the training set. The training set includes multiple example samples, and each example sample includes an input example and an output example.

[0072] Exemplarily, use the template used in the input example as the input template to generate the prompt word of the upgrade information, ensuring that the generated text format exactly matches the input example. The input template includes a specific, predefined structure for guiding how to organize and fill in the data.

[0073] For example, the input template may be: {device manufacturer}, {device type}, {firmware type}, upgrade from {current version} to {upgrade target version}{risk information of a single category},

[0074] For the first piece of upgrade information shown in Table 2, the description text generated according to this input template is: XXX, Type 1, iBMC, upgrade from Current Version 1 to Target Version 1,

[0075] Among them, the item of "risk information of a single category" is in the default state.

[0076] Based on this, the prompt word corresponding to the first piece of upgrade information shown in Table 2 includes the following description text:

[0077] Text content:

[0078] [XXX, Type 1, iBMC, upgrade from Current Version 1 to Target Version 1]

[0079] The text content is the description text of the upgrade information, which is the data that the large model needs to process.

[0080] Exemplarily, the prompting words further include an output format description to improve the accuracy and consistency of the large model output. The output format description includes multiple fields related to the upgrade information and their field descriptions. Among the multiple fields, there are a risk category field and a risk description field. The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information.

[0081] For example, the prompting words may include the following description text and output format description:

[0082] Text content:

[0083] [XXX, type 1, iBMC, upgrade from the current version 1 to the target version 1]

[0084] Please provide the output in JSON format according to the following text. The output may include the following structured key-value pairs:

[0085] - "Manufacturer": Equipment manufacturer

[0086] - "ServerCategory": Equipment type

[0087] - "FirmwareType": Firmware type

[0088] - "SourceVersion": Current version

[0089] - "TargetVersion": Upgrade target version

[0090] - "RiskType": Upgrade risk type

[0091] - "RiskDescription": Risk description

[0092] As shown above, the output format description includes several entity category fields related to the upgrade information: "Manufacturer", "ServerCategory", "SourceVersion", etc., which are used to indicate the entity information in the upgrade information, and their corresponding field descriptions: equipment manufacturer, equipment type, current version, etc. It can be understood that the output format also includes the field "RiskType", which corresponds to the field description: upgrade risk type, and includes the field "RiskDescription", which corresponds to the field description: risk description, to further describe the risk information corresponding to the upgrade information.

[0093] Exemplarily, the prompt also includes at least one example sample, which includes an input example and an output example. Such a prompt is used to guide the large model to generate an output similar in format to the output example by processing text similar in format to the input example.

[0094] Among them, the output example includes multiple fields and their field values. The multiple fields include a risk category field and a risk description field. The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information. Each output example includes upgrade risk information of one category. The classification method of the category is predefined, including severe alarm risk, compatibility risk, and upgrade step risk. The multiple fields also include several entity category fields, which are used to indicate the entity information in the input example.

[0095] For example, the prompt may include the following description content and example samples:

[0096] Please give the output in JSON format according to the following text.

[0097] Text content:

[0098] [XXX, Type 1, iBMC, upgrade from the current version 1 to the target version 1]

[0099] Input example: [YYY, Type 3, BIOS, upgrade from the current version 3 to the target version 3, upgrade risk information 1]

[0100] Output example:

[0101] {

[0102] ″Manufacturer″: ″YYY″,

[0103] ″ServerCategory″: ″Type 3″,

[0104] ″FirmwareType″: ″BIOS″,

[0105] ″SourceVersion″: ″current version 3″,

[0106] ″TargetVersion″: ″target version 3″

[0107] ″RiskType″: ″Compatibility″ / / Compatibility risk

[0108] ″RiskDescription″: ″upgrade risk information 1″

[0109] }

[0110] Among them, the output examples include several entity category fields related to upgrade information, such as "Manufacturer", "ServerCategory", "SourceVersion", etc., and their corresponding field values: "YYY", "Type 3", "Current Version 3", etc. The output examples also include the field "RiskType", which corresponds to the field value "Compatibility" and is used to indicate the risk type corresponding to the upgrade risk information described by the risk information of {a single category} in this input example. In addition, it includes the field "RiskDescription", which corresponds to the field value "Upgrade Risk Information 1" to further describe the upgrade risk information.

[0111] Thus, by providing an output format description in the prompt or including a sample of input examples and output examples, it helps to provide a clear and consistent information output representation for the large model in data processing tasks.

[0112] Exemplarily, the prompt text can also only include the description text of the upgrade information. By using the pre-trained large model, the large model learns the processing ability of the text content with the template used in the input example, and based on the text content, it infers and outputs the upgrade risk information corresponding to the upgrade information. This will not be elaborated here.

[0113] Exemplarily, corresponding description texts are generated for multiple upgrade information of the target device and provided to the large model as prompts.

[0114] Continuing with the example in Table 1, Table 2 lists the corresponding prompts generated by the device management software 110 for multiple upgrade information.

[0115] Table 2

[0116]

[0117] Step S203, input the prompt into the large model to obtain at least one model output of the upgrade information. The model output indicates the upgrade risk information of a single category of the target firmware. The categories indicated by each of the at least one model output are different from each other.

[0118] Exemplarily, the user specifies multiple target firmwares of multiple target devices. The device management software 110 generates corresponding prompts for the multiple target firmwares of the multiple target devices. The device management software 110 also classifies the multiple generated prompts according to the target device or the target firmware of the target device and inputs them into the large model, or inputs them into the large model simultaneously, to obtain multiple model outputs. A single model output indicates the upgrade risk information of a single category of a single target firmware of a single target device.

[0119] It can be understood that for a single target firmware of a single target device, its upgrade process may include at least one category of upgrade risks, so there can be at least one model output corresponding to it. The categories of the upgrade risk information indicated by each model output are different from each other, and the classification method of the categories is predefined. By using these categories with predefined classification methods in the above-mentioned multiple example samples, the trained large model can be enabled to have the ability to output categories with predefined classification methods.

[0120] Exemplarily, after inputting the prompt word and the device alarm information into the large model, at least one model output of the upgrade information is obtained. At this time, for the upgrade risk information with the category of "severe alarm risk", the large model will output a more accurate field value of the "RiskDescription" field.

[0121] For example, if only the prompt word is input into the large model, the field value of the "RiskDescription" field may be output as: It is necessary to ensure that there is no severe alarm in the BIOS, otherwise the upgrade will fail. At this time, the user needs to first determine whether there is severe-level device alarm information for the target device corresponding to the prompt word to further identify the upgrade risk of this category for the corresponding target device, and then conduct an upgrade risk investigation and handling according to the identification result. If the prompt word and the device alarm information are input into the large model together, the field value of the "RiskDescription" field may be output as: There are unresolved server alarms. At this time, since the field value of the "RiskDescription" field output by the large model is more accurate, the user can directly conduct an upgrade risk investigation and handling of this category for the corresponding target device according to the output of the large model.

[0122] Thus, the efficiency and accuracy of the system for identifying the upgrade risks of device firmware can be further improved.

[0123] Exemplarily, the large model can perform entity extraction from the upgrade information described in the prompt. An entity refers to a specific object, concept, or thing mentioned in the prompt, such as the firmware version number in the prompt. The large model can further predict the categories of these entities. An entity category refers to the features associated with the entity. For example, if the entity is a firmware version number, the entity category may be the current version or the target upgrade version. Further still, the large model can convert the identified entities and predicted categories into structured key-value pairs. This typically involves mapping the identified entities to the fields corresponding to the preset entity categories, such as "Manufacturer", "ServerCategory", "FirmwareType", "SourceVersion", "TargetVersion", and the entity is the field value of these fields. Finally, the large model can organize these key-value pairs into a formatted output, such as a JSON object, for further processing and analysis.

[0124] Thus, in this way, the large model can be effectively used for tasks such as automated firmware upgrade risk identification, improving the efficiency and accuracy of data processing.

[0125] Finally, the device management software 110 classifies the model outputs corresponding to multiple target devices according to the firmware type of the target firmware of the target device, and obtains a list of upgrade risk information for each of the multiple target devices.

[0126] Continuing the examples in Table 1 and Table 2, Table 3 lists the list of upgrade risk information for the server to perform firmware upgrade.

[0127] Table 3

[0128]

[0129] As can be seen from Table 3, for a single target firmware iBMC, the upgrade process from the current version 1 to the target version 1 includes two categories of upgrade risk information, so there are two corresponding outputs of the large model, corresponding to the field value of the "RiskType" field being: upgrade step risk, and the field value of the "RiskDescription" field being "Cannot directly upgrade from the current version 1 to the target version 1, need to upgrade to version M first", and corresponding to the field value of the "RiskType" field being: severe warning risk, and the field value of the "RiskDescription" field being "There are unresolved server alarms".

[0130] Finally, send the list of upgrade risk information for each of the above-mentioned multiple servers to the user, so that the user can perform risk investigation and handling before upgrading the server firmware by connecting to the corresponding server through the IP address.

[0131] Accordingly, the embodiments of the present application use a training set to pre-train a large model to make the large model adapt to the risk identification task before the device firmware upgrade. Among them, the training set includes input examples and output examples. Then, a description text of the device firmware upgrade information input by the user is generated according to the template of the input example and used as a prompt word to input into the large model, so that the large model outputs the upgrade risk information of the device firmware in the format of the output example. Since this risk identification process does not require manual participation, the efficiency and accuracy of the system for identifying the risk of device firmware upgrade can be improved.

[0132] Exemplarily, Figure 3 Figure 3 shows a flowchart of a large model training method provided by an embodiment of the present application. As Figure 3 shown, the training method includes the following steps:

[0133] Step S301, obtain a plurality of example samples; the example samples include input examples and output examples. The input example is a description text of the upgrade guidance information of the first device. The upgrade guidance information includes the device information of the first device, the firmware type, the current version information, the upgrade target version information of the first firmware to be upgraded in the first device, and the upgrade risk information of a single category. The output example includes a plurality of fields and their field values. The plurality of fields include a risk category field and a risk description field. The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information; the classification method of the category is predefined in advance.

[0134] Exemplarily, the first device is a hardware device to be upgraded with firmware, and the first firmware is the firmware to be upgraded in the first device. Use Figure 1 the shown training set management module 120 to connect to the device manufacturer's website of the first device in real time or regularly, and obtain the firmware upgrade guidebook of the first device released by the device manufacturer. These firmware upgrade guidebooks may include technical documents, user manuals, online forum texts, etc.

[0135] Perform data cleaning and sorting on the firmware upgrade guidebook, remove irrelevant content, format errors, and duplicate information, and obtain multiple upgrade guidance information of the first device.

[0136] The upgrade guidance information includes the device information of the first device, the firmware type, the current version information, the upgrade target version information of the first firmware, and the upgrade risk information of a single category.

[0137] In one embodiment, use Figure 1 the shown training set management module 120 to perform annotation according to the obtained multiple upgrade guidance information to construct a training set for pre-training the large model.

[0138] Exemplarily, processing a single piece of upgrade guidance information using the training set management module 120 includes: First, generating a description text of the upgrade guidance information according to the above input template to obtain an input example. Second, predefined entity categories to be extracted are defined so that the extracted entities can be filled into the corresponding entity category fields. Predefined upgrade risk categories to be output are defined so that the labeled upgrade risk information can be filled into the corresponding upgrade risk category fields. According to the entity extraction algorithm, several entities in the input example are extracted and filled into several entity category fields, and the upgrade risk information of a single category manually input from the interaction interface is filled into the risk category field and the risk description field, thereby obtaining an output example.

[0139] Thus, the output example includes multiple fields and their field values. The multiple fields include a risk category field and a risk description field. The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information. The predefined upgrade risk category fields include severe alarm risk, compatibility risk, and upgrade step risk.

[0140] The output example also includes several entity category fields, which are used to indicate the entity information in the input example. The predefined several entity category fields include device manufacturer, device type, firmware type, current firmware version, and target firmware upgrade version. Thereby, the readability of the output of the large model can be improved.

[0141] Specifically, develop or use existing entity recognition algorithms as entity extraction algorithms, which can recognize entities in the upgrade guidance information. This may involve machine learning models such as conditional random field (CRF), long short-term memory (LSTM), or pre-trained language models such as BERT. Use the entity extraction algorithm to process multiple pieces of upgrade guidance information, recognize and extract entities. Define an input template similar to the above content, fill the extracted entities into the input template to generate an input example. And fill the extracted entities into the corresponding entity category fields as the field values corresponding to the respective several entity category fields in the output example.

[0142] Next, the training set management module 120 provides a user interaction interface to allow manual input of risk information. This interface can be a command line, a web form, or any other interactive input method. The upgrade risk information input by the user is received through the interaction interface and filled into the corresponding risk category and risk description fields, that is, as the field values of the risk category field and the risk description field. Thereby, an output example is obtained. Then, all the extracted and received information is integrated to form a structured output example.

[0143] Thus, by manually annotating risk categories and risk descriptions, the accuracy of input examples can be improved.

[0144] Finally, by processing multiple upgrade guidance messages separately, a high-quality training set can be constructed.

[0145] For example, the device firmware upgrade guidance message includes: server type: type 3, server manufacturer: YYY, firmware type of a target firmware in the server: BIOS, current version of this firmware: current version 3, upgrade target version: target version 3, and there is a firmware upgrade risk: upgrade risk information 1.

[0146] Performing the above processing according to this upgrade guidance message, input examples and output examples in Json format are obtained. For example, the input example obtained is:

[0147] [YYY, type 3, BIOS, upgrade from current version 3 to target version 3, upgrade risk information 1]

[0148] The output example obtained is:

[0149] {

[0150] ″Manufacturer″: ″YYY″,

[0151] ″ServerCategory″: ″type 3″,

[0152] ″FirmwareType″: ″BIOS″,

[0153] ″SourceVersion″: ″current version 3″,

[0154] ″TargetVersion″: ″target version 3″

[0155] ″RiskType″: ″Compatibility″ / / Compatibility risk

[0156] ″RiskDescription″: ″upgrade risk information 1″

[0157] }

[0158] The above input examples and output examples form an example sample, which can help the large model learn how to generate appropriate outputs according to the given inputs. This learning method enables the large model to improve its performance in specific applications.

[0159] Step S302, training the large model using the training set constructed from multiple example samples.

[0160] Exemplarily, obtain a pre-trained large foundation model, such as models like BERT, GPT, DeepSeek, etc. Use the training set to train the large foundation model to obtain a large model that can adapt to the task of identifying risks in device firmware upgrades.

[0161] In the application of the large model, the pre-trained large model can be used to identify and generate structured information similar to that included in the output examples to support the system in making decisions and managing risks. This process enables the large model to more accurately identify and predict entities and attributes in the field of device firmware upgrade risk identification.

[0162] Specifically, use the training set to pre-train the large foundation model to obtain a large model, enabling the large model to effectively process descriptive texts similar to the input examples, automatically extract key information, and generate structured outputs similar to the output examples. This not only improves the efficiency of data processing but also helps to enhance the automation and intelligence levels of the device firmware upgrade risk identification process.

[0163] Exemplarily, Figure 4 shows a schematic diagram of a process for identifying risks in server firmware upgrades provided by an embodiment of the present application. As Figure 4 shown, this process is implemented by the user, the training set management module, the large model module, and the server management module. Among them, the training set management module, the large model module, and the server management module respectively complete the functions of the training set management module 120, the large model module 130, and the device management module 110 in Figure 1 . The implementation process includes the following steps:

[0164] Step S401, import the server firmware upgrade guide.

[0165] Exemplarily, the user imports the server firmware upgrade guide into the training set management module.

[0166] Step S402, process according to the server firmware upgrade guide to construct a training set.

[0167] Step S403, send the training set and perform pre-training of the large foundation model to obtain a large model.

[0168] Exemplarily, after the training set management module constructs the training set, it inputs the training set into the large foundation model obtained by the large model module for pre-training to obtain a large model.

[0169] Further, after the user determines that the system has completed the training of the large model, enter step S404.

[0170] Step S404, select the server with the firmware to be upgraded and perform server firmware upgrade risk identification.

[0171] Exemplarily, the user specifies the server whose firmware is to be upgraded through the user interaction interface provided by the server management module, and starts to execute the server firmware upgrade risk identification process.

[0172] Step S405: Summarize the upgrade information list of the servers whose firmware is to be upgraded.

[0173] Exemplarily, the server management module summarizes the information according to the user's selection to obtain the upgrade information list.

[0174] Step S406: Construct prompt words according to the upgrade information list and send them down.

[0175] Exemplarily, the server management module constructs multiple prompt words according to the upgrade information list and inputs them into the large model.

[0176] Step S407: Return the upgrade risk information.

[0177] Exemplarily, the large model returns the upgrade risk information to the server management module.

[0178] Step S408: Return the server firmware upgrade risk list.

[0179] Exemplarily, after summarizing the upgrade risk information returned by the large model, the server management module obtains the upgrade risk list of the server and sends it to the user, so that the user can perform risk investigation and handling before the server firmware upgrade according to the upgrade risk list of the server.

[0180] For the detailed execution processes of the above steps S401 to S408, refer to the descriptions in steps S201 to S203 and steps S301 to S302, which will not be elaborated here.

[0181] It can be understood that in the embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0182] Exemplarily, the embodiments of the present application also provide a computing device 1000. As Figure 5 shown, the computing device 1000 includes: a bus 1002, a processor 1004, a memory 1006, and a communication interface 1008. The processor 1004, the memory 1006, and the communication interface 1008 communicate with each other through the bus 1002. It should be understood that the present application does not limit the number of processors and memories in the computing device 1000.

[0183] The bus 1002 can be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 it is represented by a single line in Figure 5 , but this does not mean that there is only one bus or one type of bus. The bus 1004 can include a path for transmitting information between various components of the computing device 1000 (for example, the memory 1006, the processor 1004, and the communication interface 1008).

[0184] The processor 1004 can include any one or more of a central processing unit, a graphics processing unit (GPU), a microprocessor (MP), a digital signal processor (DSP), a baseboard management controller, and other processors.

[0185] The memory 1006 can include volatile memory, such as random access memory (RAM). The processor 1004 can also include non-volatile memory, such as read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0186] The communication interface 1008 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the computing device 1000 or a cluster composed of multiple computing devices 1000 and other devices or communication networks.

[0187] The computing device 1000 includes internal network devices, or the computing device 1000 is externally connected to multiple network devices. The internal network devices communicate with the processor 1004, the memory 1006, and the communication interface 1008 through the bus 1002, and the external network devices communicate with the computing device 1000 through interfaces such as Ethernet, Fibre Channel, and InfiniBand.

[0188] The memory 1006 stores executable program code / instructions, and the processor 1004 executes the executable program code / instructions to implement Figure 2 or Figure 3The method flow shown therein is used to implement all or part of the steps of the method in the above embodiments. In other words, a program / instruction for executing all or part of the steps of the method in the above embodiments is stored on the memory 1006.

[0189] An embodiment of the present application provides a computing device, including: a memory and a processor; the memory and the processor are coupled; the memory is used to store a program; the processor is used to execute the program stored in the memory, and when the program stored in the memory is executed, the processor is used to execute the method in the above embodiments.

[0190] Based on the method in the above embodiments, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor is caused to execute the method in the above embodiments.

[0191] Based on the method in the above embodiments, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor is caused to execute the method in the above embodiments.

[0192] The method steps in the embodiments of the present application can be implemented in a hardware manner or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0193] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

[0194] It can be understood that the various numerical numbers involved in the embodiments of the present application are only for the convenience of description and are not used to limit the scope of the embodiments of the present application.

Claims

1. A method for identifying risks in device firmware upgrades, characterized in that, The method includes: Obtaining upgrade information of a target device; the upgrade information includes device information of the target device, firmware type, current version information, and upgrade target version information of a target firmware to be upgraded in the target device; Generating a description text of the upgrade information according to an input template and determining a prompt; the input template is the same as the template of an input example; the input example is used to construct an example sample, the example sample is used to construct a training set, and the training set is used to pre-train a large model; Inputting the prompt into the large model to obtain at least one model output of the upgrade information; the model output indicates upgrade risk information of a single category of the target firmware; the categories indicated by at least one model output are different from each other.

2. The method according to claim 1, wherein The method further includes: Obtaining current device alarm information of the target device; The step of inputting the prompt into the large model to obtain at least one model output of the upgrade information includes: Inputting the device alarm information and the prompt into the large model to obtain at least one model output of the upgrade information.

3. The method according to claim 1, wherein The example sample further includes an output example, the output example includes multiple fields and their field values, and the multiple fields include a risk category field and a risk description field; The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information.

4. The method according to claim 3, wherein The prompt further includes at least one of the example samples.

5. The method according to claim 1, wherein The prompt further includes an output format description, and the output format description includes multiple fields related to the upgrade information and their field descriptions, and the multiple fields include a risk category field and a risk description field; The risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information.

6. The method according to claim 1, wherein The step of obtaining upgrade information of the target device includes: Obtaining multiple upgrade information of the target device; The method further includes: Classifying the model output corresponding to each upgrade information according to the firmware type of the target firmware to obtain a list of upgrade risk information of the target device.

7. A large model training method, characterized in that, The method includes: Obtaining multiple example samples; the example sample includes an input example and an output example, the input example is a description text of upgrade guidance information of a first device, and the upgrade guidance information includes device information of the first device, firmware type, current version information, upgrade target version information, and upgrade risk information of a single category of a first firmware to be upgraded in the first device; the output example includes multiple fields and their field values, the multiple fields include a risk category field and a risk description field, the risk category field is used to indicate the category corresponding to the upgrade risk information, and the risk description field is used to describe the upgrade risk information; the classification method of the categories is predefined; Training the large model with a training set constructed by multiple example samples.

8. The method according to claim 7, wherein The multiple fields further include a number of entity category fields; the number of entity category fields are used to indicate entity information in the input example; the number of entity category fields include equipment manufacturer, equipment type, firmware type, current firmware version, and target firmware upgrade version.

9. The method according to claim 7, wherein The upgrade guidance information of the first device is determined by the firmware upgrade guide of the first device, and the firmware upgrade guide is released by the device manufacturer of the first device.

10. A computing device, characterized in that, including: at least one memory for storing programs; at least one processor for executing the programs stored in the memory, and when the programs stored in the memory are executed, the processor is used to execute the method according to any one of claims 1-9.

Images