Software operation detection method and device, equipment, medium and product

Abstract

The invention discloses a software operation detection method, device and equipment, a medium and a product, relates to the technical field of software testing and large models, and can be applied to the field of financial science and technology. Acquiring software operation information of the target software; according to the software operation information, detecting the target software based on a software inspection model to obtain a software detection result; wherein the software inspection model is obtained by performing pre-training and fine tuning on a large model. By the adoption of the technical scheme, software inspection is conducted through the large model, problems encountered in the software operation process can be rapidly positioned, solving measures are generated, and it is ensured that the software problems are rapidly solved in time.

Description

Technical Field

[0001] The present invention relates to the field of software testing and large model technology, can be applied to the field of financial technology, and in particular to a software operation detection method, device, equipment, medium and product. Background Art

[0002] The current financial industry is increasingly dependent on digitalization. If the software fails due to abnormalities, it will cause business interruptions and affect the normal operation of the business, resulting in huge losses.

[0003] Existing detection of software running processes often requires manual acquisition of logs, checking the cause of the error, or writing some regular matching solutions to extract keywords for error analysis. The overall software detection solution is inefficient, has poor accuracy, and requires manual supervision, and does not have the ability to actively generate solutions. Summary of the invention

[0004] The present invention provides a software operation detection method, device, equipment, medium and product. By performing software inspection through a large model, problems encountered in software operation can be quickly located, and solutions can be generated to ensure that software problems are solved promptly and quickly.

[0005] According to one aspect of the present invention, a software operation detection method is provided, comprising:

[0006] Obtain software running information of the target software;

[0007] According to the software running information, the target software is detected based on a software inspection model to obtain a software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

[0008] According to another aspect of the present invention, there is provided a software operation detection device, comprising:

[0009] An operation information acquisition module is used to obtain software operation information of the target software;

[0010] The detection module is used to detect the target software based on the software inspection model according to the software operation information to obtain the software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

[0011] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0012] at least one processor; and

[0013] a memory communicatively connected to the at least one processor; wherein,

[0014] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the software operation detection method described in any embodiment of the present invention.

[0015] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the software operation detection method described in any embodiment of the present invention when executed.

[0016] According to another aspect of the present invention, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the software operation detection method according to any embodiment of the present invention is implemented.

[0017] The technical solution of the embodiment of the present invention performs software inspection through a large model, which solves the problem of low problem analysis efficiency and inability to automatically recover after an error or interruption occurs during the operation of the software system. It can quickly locate problems encountered in software operation and generate solutions to ensure that software problems are resolved promptly and quickly.

[0018] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 is a flowchart of a software operation detection method provided according to an embodiment of the present invention;

[0021] Figure 2 is a software detection flow chart provided according to an embodiment of the present invention;

[0022] Figure 3 is a flowchart of a software operation detection method provided according to an embodiment of the present invention;

[0023] Figure 4 is a model fine-tuning flow chart provided according to an embodiment of the present invention;

[0024] Figure 5is a structural schematic diagram of a software operation detection device provided according to an embodiment of the present invention;

[0025] Figure 6 It is a structural schematic diagram of an electronic device for implementing the software operation detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0028] In addition, it should be noted that the collection, storage, use, processing, transmission, provision and disclosure of data involved in the technical solution of the present invention are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0029] Figure 1 1 is a flowchart of a software operation detection method provided according to an embodiment of the present invention. The present invention is applicable to the case of software operation detection. The method can be executed by a software operation detection device provided by an embodiment of the present invention. The software operation detection device can be implemented in the form of hardware and / or software and can be integrated into an electronic device that carries the software operation detection function, such as a server. Figure 1 As shown, the method includes:

[0030] S110: Obtain software running information of the target software.

[0031] Among them, the target software is the software that needs to be evaluated for operation; the software operation information is the operation status data of each operation process of the software during operation, which can be the operation status data such as which function the target software is currently running, the current running memory, CPU usage and network usage; it should be noted that the present invention does not specifically limit the target software, and it can be application software that processes any type of business; for example, the Industrial and Commercial Bank of China application software in a financial scenario.

[0032] Specifically, the running status data of the target software during its running process is obtained, which may include information such as which function the target software is running and the corresponding running memory.

[0033] In an optional manner of the present invention, a log file of the target software is obtained, and software running information of the target software is extracted from the log file.

[0034] Specifically, the log file of the target software can be obtained through the log management tool, and the printed log is redirected to a specific file, and the context information in the log file is intercepted according to a certain period. It should be noted that the log management tool can be ELK Stack, Splunk and Graylog, and relevant technical personnel can configure the corresponding log management tool to obtain logs according to current needs, and the embodiment of the present invention does not specifically limit this.

[0035] S120. According to the software running information, the target software is detected based on the software inspection model to obtain the software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

[0036] Among them, the software inspection model is obtained by pre-training and fine-tuning the generative big model. The generative big model is a machine learning model that can generate new and relevant outputs based on input data; the software detection result is the detection result of the software, which can be running well or running abnormally.

[0037] Specifically, the acquired software operation information is input into the software inspection model for detection to obtain the corresponding software detection results, which may be whether there is an abnormality in the current operation, memory usage, etc.

[0038] Optional, Figure 2 A software detection flow chart provided according to an embodiment of the present invention; according to the software running information, the target software is detected based on the software inspection model to obtain the software detection result, including:

[0039] S121. Determine whether the target software is running according to the software running information.

[0040] Based on the software running information, whether the target software is running can be determined based on whether the log is updated.

[0041] S122: If the software is in operation, concatenate prompt words on the software operation information to obtain a first prompt word; perform detection and reasoning on the first prompt word based on the software inspection model to obtain a software detection result.

[0042] Among them, the prompt word is a word used for prompts, which can be the name or key information corresponding to the running information. For example, if the current target software is running, it has been running for 2 hours, the CPU usage is 30%, the memory occupancy is 200MB, and the network connection is normal, then the first prompt word after splicing may be: "Running - 2 hours - CPU 30% - Memory 200MB - Network connection normal".

[0043] Specifically, if it is determined that the target software is in the process of running, part of the information is intercepted from the log file, and the corresponding prompt word is spliced ​​with the intercepted software running information to obtain a first prompt word, and the first prompt word is used as the input of the software inspection model, so that the software inspection model performs detection and reasoning on the first prompt word to obtain the software detection result. It should be noted that the splicing method can splice the intercepted information into a complete sentence by setting a placeholder, which is set by relevant technicians according to application requirements, and the embodiment of the present invention does not specifically limit this.

[0044] In an optional embodiment of the present invention, if the target software is running, after obtaining the software detection results, it also includes: determining whether operation alarm information appears in the software detection results; if so, pushing the operation alarm information and re-obtaining the log; if not, re-obtaining the log.

[0045] The alarm information is that some abnormalities exist in the current target software, which will cause abnormal software operation at some time in the future; for example, whether the memory continues to increase, and whether some alarms of insufficient video memory appear.

[0046] Specifically, when an alarm message appears in the software detection results, the alarm message can be quickly sent to relevant support personnel by calling the email API interface, which can support them to quickly discover abnormalities and prepare countermeasures; if no logs affecting operation are found in the software detection results, log acquisition and analysis can continue.

[0047] It is understandable that when an abnormality occurs in the running status of the target software, relevant operation alarm information can be sent so that technical support personnel can quickly repair the abnormality, achieve rapid warning of abnormal problems, and quickly deal with abnormal problems, further ensuring the continuity of software services.

[0048] S123, if the operation stops, intercept the software operation information before the operation stops, and perform prompt word splicing on the software operation information to obtain a second prompt word; perform detection and reasoning on the second prompt word based on the software inspection model to obtain a software detection result.

[0049] If it is found that the target software has stopped running, a section of information before the stop is intercepted from the acquired log file information, and a specific prompt word is spliced ​​to obtain a second prompt word, and the second prompt word is used as the input of the software inspection model so that the software inspection model can perform detection and reasoning on the second prompt word to obtain the software detection result.

[0050] In an optional implementation of the present invention, after obtaining the software detection result, the software detection result is further sent to relevant technical support personnel by calling an email API interface.

[0051] In an optional embodiment of the present invention, if the target software runs abnormally, the software detection result includes a software solution, and the software detection result is sent to support personnel. If the solution is passed, the solution is activated to process the target software.

[0052] Among them, the software solution may include the execution operation corresponding to the analyzed software problem; the analyzed problem category may include software memory usage problems and other operation information problems.

[0053] Exemplarily, the software inspection model detects and infers the second prompt word. If it finds that there is insufficient hard disk space or hard disk damage, it sends relevant error information to remind support personnel to quickly contact relevant department personnel to jointly restore software operation; if there is insufficient memory or other problems that can be restarted and recovered, it submits solutions such as restart, applies for rapid resumption of production first, and saves relevant log files for relevant personnel to modify errors.

[0054] It is understandable that when an exception occurs during the process of stopping the target software, an alarm message and corresponding solution can be sent to the technical support personnel so that the technical support personnel can quickly repair the exception, realize rapid warning of abnormal problems, and quickly deal with abnormal problems, thereby further ensuring the continuity of software services.

[0055] The embodiment of the present invention obtains software running information of the target software; according to the software running information, the target software is detected based on a software inspection model to obtain a software detection result; and the software inspection model is obtained by pre-training and fine-tuning a large model; the above-mentioned technical solution performs software inspection based on a large model to quickly locate problems encountered in software operation, and uses the log analysis capability of the large model to analyze abnormal logs in real time and generate solutions; it effectively solves the problem of low problem analysis efficiency and inability to automatically recover if an error or interruption occurs during the operation of the software system, and analyzes the causes of the software operation interruption, and then generates solutions to ensure that software problems can be solved promptly and quickly.

[0056] Figure 3is a flowchart of a software operation detection method provided according to an embodiment of the present invention. Based on the above embodiment, the embodiment of the present invention supplements the construction method of the software inspection model. It should be noted that for the part not described in detail in the embodiment of the present invention, please refer to the relevant description of other embodiments, such as Figure 3 As shown, the method includes:

[0057] S210: Acquire training text data from the Internet, and encode the training text data to obtain training text features.

[0058] The training text data is high-quality text data used for model training; the training text data is obtained from the Internet, and the training text is encoded according to a specific encoding method to obtain training text features. It should be noted that relevant technicians can select a suitable encoding method according to the specific type and characteristics of the data, and the embodiment of the present invention does not specifically limit the encoding method.

[0059] It can be understood that converting raw text data into digital information that can be understood by computers as training text features provides strong support for subsequent model training.

[0060] In an optional method of the present invention, before obtaining the training text data, the text data is also preprocessed; hundreds of TB or even PB-level data are often obtained from the Internet, and after a large amount of illegal, advertising, and low-information dirty data is removed through data cleaning, 3-5TB-level high-quality text data is obtained as training text data. It can be understood that by cleaning the Internet data, the high quality of the training text data is ensured, and the accuracy of model training is further improved.

[0061] S220. Pre-train the text modality large model using training text features to obtain a pre-trained model.

[0062] The text modality large model can be a Transformer model, which uses a self-attention mechanism to capture dependencies in sequence data; the text modality large model is pre-trained using training text features to obtain a pre-trained model.

[0063] For example, pre-training is to process the collected text and use the above information as input to predict the next word. For example, if "financial technology is developing rapidly" is input into the text modal large model, the "different" in "financial technology is developing rapidly" can be predicted. Similarly, the model can complete the text prediction and continuation tasks well by fitting a large amount of data.

[0064] S230, obtaining business data; business data includes customer consultation data and customer service answer data.

[0065] Obtain business data, including customer consultation data and customer service answer data; customer consultation data generally refers to consultations or inquiries initiated by customers through various channels; customer service answer data refers to the responses and solutions provided by customer service personnel to customer consultation issues.

[0066] S240. Fine-tune the pre-trained model using business data to obtain a software inspection model.

[0067] Fine-tune the pre-trained model based on business data to obtain a software inspection model. Fine-tune the pre-trained model using customer consultation data and customer service answer data.

[0068] It is understandable that the pre-trained model can be fine-tuned through business data in the form of question-answer pairs to enhance the command compliance and specific task capabilities of the software inspection model.

[0069] Optional, Figure 4 A model fine-tuning flow chart is provided according to an embodiment of the present invention; the pre-trained model is fine-tuned using business data to obtain a software inspection model, including:

[0070] S241. Construct fine-tuning data based on business data and prompt words.

[0071] Fine-tuning data is constructed based on customer consultation data, customer service answer data and prompt words. For example, in a financial scenario, the prompt word can be "You are an expert in risk management. Please carefully analyze the following content and determine whether there is any financial risk", which is then spliced ​​with the consultation and customer service data in the specific business data as fine-tuning data.

[0072] S242. Annotate the fine-tuning data based on the generative model to obtain fine-tuning label data.

[0073] Among them, the generative model can be a text generation type model, and relevant technical personnel can choose it according to specific task requirements. Any generative model that can annotate text data can be implemented, and the embodiment of the present invention does not specifically limit this.

[0074] The fine-tuning data is input into the generative model for data annotation to obtain fine-tuning label data. The labels in the fine-tuning label data may include whether the data contains risks and the corresponding reasons. Exemplarily, the fine-tuning label data can be designed as 5k data to facilitate subsequent model fine-tuning, so that model data fine-tuning training is feasible.

[0075] S243. Fine-tune the pre-trained model using fine-tuning label data to obtain a software inspection model.

[0076] The fine-tuning label data is used as the model input to fine-tune the pre-trained model to obtain the software inspection model. It should be noted that the specific fine-tuning process can be referred to by relevant technicians in the industry for common fine-tuning means, and the embodiment of the present invention does not specifically limit this.

[0077] The fine-tuning of the software inspection model is based on question-answering training based on fine-tuning label data to ensure that the model can independently complete the corresponding specific tasks. For example, in financial scenarios, the fine-tuned software inspection model is used to answer specific questions, which can be:

[0078] {

[0079] "instruction":"Reference document: XXX.\n\n\n\"\"\"\nYou are a bank branch employee. You will encounter many problems in your daily business. Please refer to the above document, think carefully step by step and answer the following questions. Please answer as complete, credible and high-quality as possible;\n\"\"\"\"\"\"\n\n\nQuestion: What work do overseas institutions need to do when exiting the purchased system? \n\n\"\"\"\n\n\nAnswer:",

[0080] "output":"When an overseas institution needs to exit the outsourced system, XXX"

[0081] }

[0082] Among them, "instruction" is the model input, and "output" is the model output. Separators can be set between each data information.

[0083] It is understandable that model training completes output through historical input, and model fine-tuning is to let the model generate output by learning instructions; and the business data is labeled through the generative model to obtain a more accurate model suitable for the business scenario.

[0084] Optionally, fine-tuning the pre-trained model using fine-tuning label data to obtain a software inspection model, including: fine-tuning the pre-trained model using fine-tuning label data to obtain a fine-tuning result; the fine-tuning result includes a forward result and a reverse result; respectively calculating the difference between the forward result and the reverse result to obtain a forward difference and a reverse difference; determining the training loss based on the forward difference and the reverse difference, and using the training loss to fine-tune the pre-trained model to obtain a software inspection model.

[0085] Among them, the positive result is a good answer, and the negative result is a bad answer; the data format of the fine-tuning result is {"instruction":"XXXX","[pos_output":"XXXX","neg_output":"XX X"}, defining pos_output as a positive result and neg_output as a negative result.

[0086] Specifically, by obtaining the positive and negative results in the fine-tuning results, the scores of two different answers are further calculated as the positive score difference and the reverse score difference. The positive score difference is higher and the negative score difference is lower. The training loss is determined according to the positive score difference and the reverse score difference, and the pre-trained model is fine-tuned based on the training loss to obtain the software inspection model.

[0087] It is understandable that the training model assigns different differences to positive and negative results according to the reward and punishment mechanism to obtain training losses, which can generate results that meet reward preferences, ensure that the output is consistent with values, adjust model parameters, and achieve large model value alignment, making it more stable and reliable, and further improving model performance, accuracy, and generalization capabilities.

[0088] S250: Obtain software running information of the target software.

[0089] S260. According to the software running information, the target software is detected based on a software inspection model to obtain a software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

[0090] The embodiment of the present invention obtains an accurate model suitable for business scenario analysis by pre-training and fine-tuning the software inspection model, so that the software inspection model can perform detection according to the software operation information; and assigns different differences to the fine-tuning results to obtain training losses, so that it can generate results that meet the reward preferences, ensure that the output is consistent with the values, and achieve large model value alignment, thereby being more stable and reliable.

[0091] Optionally, before using the training text features to pre-train the large text modality model, it also includes: dividing the network layers of the text modality model according to the parameter quantity of the text modality model to obtain at least two network groups; distributing the at least two network groups to different graphics cards respectively to pre-train the text modality model.

[0092] According to the parameter quantity of the text modal model, the network layer of the text modal model is divided according to a certain parameter quantity to obtain at least two network groups, and the at least two network groups are distributed to different graphics cards to complete the subsequent pre-training of the text modal model.

[0093] In an optional method of the present invention, the number of network layers of the text modality model is evenly divided. For example, if the text modality model has 80 layers and needs to be divided into 32 graphics cards, 3 layers are placed on each graphics card on average. The number of network layers of the text modality model can also be divided according to the parameter amount of each layer. For example, if the embedding layer is relatively large, the graphics card where the layer is located will have fewer network layers. For models with particularly large parameter amounts, a single graphics card may not be able to accommodate one layer, and one layer will be divided into two parts according to the attention head. For example, the transformer has multiple attention heads and can be divided according to the head. It should be noted that in the embodiment of the present invention, the network layers of the model are divided. The relevant technical personnel can divide according to the actual model parameter amount to ensure the maximum utilization of graphics card resources. The embodiment of the present invention does not specifically limit this.

[0094] It is understandable that since the number of parameters of large models reaches hundreds of billions or trillions, the capacity of a single card is generally only 40GB-96GB, and it is impossible to load all of them onto one card during model training. By splitting the network layer of the model, the training efficiency can be improved and the model training can be accelerated.

[0095] Figure 5 1 is a schematic diagram of the structure of a software operation detection device provided according to an embodiment of the present invention. The present invention is applicable to the case of software operation detection. The software operation detection device can be implemented in the form of hardware and / or software, and can be integrated into an electronic device that carries the software operation detection function, such as a server. Figure 5 As shown, the software operation detection device 300 includes an operation information acquisition module 310 and a detection module 320 .

[0096] The operation information acquisition module 310 is used to acquire the software operation information of the target software;

[0097] The detection module 320 is used to detect the target software based on the software inspection model according to the software operation information to obtain the software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning the large model.

[0098] The embodiment of the present invention obtains software running information of the target software; according to the software running information, the target software is detected based on a software inspection model to obtain a software detection result; and the software inspection model is obtained by pre-training and fine-tuning a large model; the above-mentioned technical solution performs software inspection based on a large model to quickly locate problems encountered in software operation, and uses the log analysis capability of the large model to analyze abnormal logs in real time and generate solutions; it effectively solves the problem of low problem analysis efficiency and inability to automatically recover if an error or interruption occurs during the operation of the software system, and analyzes the causes of the software operation interruption, and then generates solutions to ensure that software problems can be solved promptly and quickly.

[0099] Optionally, the detection module 320 is specifically used to determine whether the target software is running according to the software running information; if it is running, the software running information is spliced ​​with prompt words to obtain a first prompt word; the first prompt word is detected and inferred based on the software inspection model to obtain a software detection result; if it stops running, the software running information before the running stops is intercepted, and the software running information is spliced ​​with prompt words to obtain a second prompt word; the second prompt word is detected and inferred based on the software inspection model to obtain a software detection result.

[0100] On the basis of the above technical solutions, the software operation detection device 300 further includes a model construction module, specifically including: a training text feature determination unit, a pre-training model determination unit, a business data acquisition unit and a model acquisition unit;

[0101] A training text feature determination unit is used to obtain training text data from the Internet and encode the training text data to obtain training text features;

[0102] A pre-training model determination unit, used to pre-train the text modality large model using training text features to obtain a pre-training model;

[0103] A business data acquisition unit, used to acquire business data; business data includes customer consultation data and customer service answer data;

[0104] The model acquisition unit is used to fine-tune the pre-trained model using business data to obtain a software inspection model.

[0105] Optionally, the model acquisition unit includes a fine-tuning data acquisition subunit, a fine-tuning label data acquisition subunit and a model acquisition subunit;

[0106] A fine-tuning data acquisition subunit is used to construct fine-tuning data based on business data and prompt words;

[0107] A fine-tuning label data acquisition subunit is used to annotate the fine-tuning data based on the generative model to obtain fine-tuning label data;

[0108] The model acquisition subunit is used to fine-tune the pre-trained model using fine-tuning label data to obtain a software inspection model.

[0109] Optionally, the model acquisition subunit is specifically used to fine-tune the pre-trained model using fine-tuning label data to obtain a fine-tuning result; the fine-tuning result includes a forward result and a reverse result; the difference between the forward result and the reverse result is calculated respectively to obtain a forward difference and a reverse difference; according to the forward difference and the reverse difference, the training loss is determined, and the training loss is used to fine-tune the pre-trained model to obtain a software inspection model.

[0110] Optionally, the model building module also includes a segmentation unit, which is used to segment the network layers of the text modal model according to the parameter quantity of the text modal model to obtain at least two network groups; and distribute the at least two network groups to different graphics cards respectively to pre-train the text modal model.

[0111] The software operation detection device provided in the embodiment of the present invention can execute the software operation detection method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0112] According to an embodiment of the present invention, the present invention also provides an electronic device, a readable storage medium and a computer program product.

[0113] Figure 6 1 is a schematic diagram of the structure of an electronic device 10 that implements the software operation detection method of an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) 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 invention described and / or required herein.

[0114] like Figure 6 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0115] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0116] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a software operation detection method.

[0117] In some embodiments, the software operation detection method can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the software operation detection method described above can be performed. Alternatively, in other embodiments, the processor 11 can be configured to execute the software operation detection method in any other appropriate manner (for example, by means of firmware).

[0118] 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), load 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.

[0119] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program 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.

[0120] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. 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.

[0121] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device 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 a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0122] The systems and techniques described herein may be implemented in a computing system that includes backend 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 frontend 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 backend components, middleware components, or frontend 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), a blockchain network, and the Internet.

[0123] A computing 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 client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0124] 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 the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0125] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A software operation detection method, characterized in that: include: Obtain software running information of the target software; According to the software running information, the target software is detected based on a software inspection model to obtain a software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

2. The method according to claim 1, characterized in that According to the software running information, the target software is detected based on the software inspection model to obtain a software detection result, including: According to the software running information, determining whether the target software is running; If the software is in operation, the prompt words are spliced ​​on the software operation information to obtain a first prompt word; the first prompt word is detected and inferred based on the software inspection model to obtain a software detection result; If the software stops running, the software running information before the running stops is intercepted, and the prompt words of the software running information are spliced ​​to obtain a second prompt word; the second prompt word is detected and inferred based on the software inspection model to obtain a software detection result.

3. The method according to claim 1 or 2, characterized in that: The software inspection model is obtained in the following way: Acquire training text data from the Internet, and encode the training text data to obtain training text features; Pre-training the text modality large model using the training text features to obtain a pre-trained model; Acquire business data; the business data includes customer consultation data and customer service answer data; The business data is used to fine-tune the pre-trained model to obtain the software inspection model.

4. The method according to claim 3, characterized in that The fine-tuning of the pre-trained model by using the business data to obtain the software inspection model includes: Construct fine-tuning data according to the business data and prompt words; Annotating the fine-tuning data based on a generative model to obtain fine-tuning label data; The pre-trained model is fine-tuned using the fine-tuning label data to obtain the software inspection model.

5. The method according to claim 4, characterized in that The fine-tuning of the pre-trained model by using the fine-tuning label data to obtain the software inspection model includes: Fine-tune the pre-trained model using the fine-tuning label data to obtain a fine-tuning result; the fine-tuning result includes a positive result and a negative result; Calculate the difference between the forward result and the reverse result respectively to obtain a forward difference and a reverse difference; The training loss is determined according to the forward difference and the reverse difference, and the pre-training model is fine-tuned using the training loss to obtain the software inspection model.

6. The method according to claim 3, characterized in that Also includes: According to the parameter quantity of the text modality model, the network layers of the text modality model are divided to obtain at least two network groups; Distribute at least two network groups to different graphics cards respectively to pre-train the text modality model.

7. A software operation detection device, characterized in that: include: An operation information acquisition module is used to obtain software operation information of the target software; The detection module is used to detect the target software based on the software inspection model according to the software operation information to obtain the software detection result; wherein the software inspection model is obtained by pre-training and fine-tuning a large model.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the software operation detection method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the software operation detection method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements the software operation detection method according to any one of claims 1 to 6.

Images