Construction method and equipment of automobile software defect repair evaluation model based on LLAMA2 and medium
By building a LLAMA2-based automotive software defect repair evaluation model, the problem of inaccurate automotive software vulnerability repair evaluation in the existing technology is solved, low-cost and efficient vulnerability evaluation and repair are achieved, and the specific environmental needs of automotive software are adapted to.
Patent Information
- Application Number
- CN202411533343.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-31
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing automotive software vulnerability repair evaluation methods rely on CVSS scores and cannot accurately measure the harm of automotive software vulnerabilities, resulting in high cost and difficult repairs, and are not suitable for the needs of the automotive environment.
Building a automotive software defect repair evaluation model based on LLAMA2. By collecting pre-training data in a predetermined format, using word segmentation device Tokenizer to expand word results, and fine-tuning the instruction of the LLAMA2 basic model through the LoRA method, combining with LangChain integration, the Recommendations for automatic evaluation and repair of automotive software vulnerabilities.
It reduces the cost and difficulty of repairing automotive software defects, improves safety work efficiency, provides reasonable vulnerability assessment and repair solutions, and adapts to the specific environmental needs of automotive software.
Smart Images

Figure CN119067222A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of computer simulation, and in particular relates to a method, device and medium for constructing an automobile software defect repair evaluation model based on LLAMA2. Background Art
[0002] At present, a large number of component vulnerabilities are often scanned in the security verification link after automobile assembly and the operation link after automobile mass production. Because the structure of automobile software is complex, it is very difficult to truly understand and identify the risks of automobile security vulnerabilities. Repairing automobile vulnerabilities is also a very costly task. The existing method of classifying whether automobile vulnerabilities should be repaired is to check the CVSS score and level. When the scanned or detected vulnerabilities are classified as high-risk or above according to the CVSS calculation, they are repaired. Other vulnerabilities are temporarily suspended due to the high cost of repair work.
[0003] However, the evaluation criteria for automobile vulnerability repairs are based only on the CVSS scores corresponding to CVE vulnerabilities. The automotive environment is different from the traditional IT software environment, which will result in the CVSS vulnerability scores in the IT field not being applicable to the automotive software field. The CVSS score cannot accurately measure the harm of the vulnerability in the car. In other words, the automotive software repair evaluation method is not suitable for production needs. Summary of the invention
[0004] Aiming at the existing problems. The present invention provides a method, device and medium for constructing an automobile software defect repair evaluation model based on LLAMA2. The present invention collects pre-trained data in a predetermined format; performs word segmentation processing on the pre-trained data in a predetermined format according to a word segmenter Tokenizer; expands the word segmentation results; pre-trains the LLAMA2 basic model according to the expanded word segmentation results, and uses the LoRA method to use the fine-tuning data to fine-tune the instructions of the LLAMA2 basic model; obtains an automobile software defect repair evaluation model; performs LangChain integration, so that the automobile software defect repair evaluation model can complete automatic answers to specific documents; the automobile software defect repair evaluation model outputs answers to questions about vulnerabilities based on complete input questions, makes reasonable evaluations of automobile software defects, reasonably reduces the cost of automobile software defect repair, and also reduces the difficulty threshold of automobile software defect repair work, and improves safety work efficiency.
[0005] The technical solution adopted by the present invention to solve the above technical problems is: The method includes: A method for constructing an automotive software defect repair evaluation model based on LLAMA2 includes the following steps: S1. Collect pre-trained data in a predetermined format; the predetermined format includes: CVE vulnerability number, CVE vulnerability description, CVE vulnerability CVSS score, open source component where the vulnerability occurs, the scope of the open source component where the vulnerability exists, and the POC program and patch corresponding to the vulnerability.
[0006] S2. For the pre-trained data in the predetermined format, perform word segmentation processing according to a Tokenizer; The word segmenter training comprises the following steps: Use the SentencePiece tool training command to train the Tokenizer, including: S20, select subword mode for training by default; S21, select the BPE encoding method and select the amount of vocabulary data to be generated; S22, select a training data file, clean the text data in the training data file and convert it into UTF-8 encoding; S23, setting the character ratio that the model should cover; S3, expand the word segmentation results; expand the words in the word segmentation results into synonyms and near-synonyms; S4. Pre-train the LLAMA2 basic model based on the expanded word segmentation results. S5. Use the LoRA (Low-Rank Adaptation) method to fine-tune the instructions of the LLAMA2 basic model using the fine-tuning data; including: S51, obtaining fine-tuning data, where the fine-tuning data includes input words of automobile vulnerability data and corresponding output labels; S52, LoRA parameter initialization: LoRA parameters are introduced on the last layer of the LLAMA2 basic model, where the parameters are second-order matrices, and their rank is 1 / 10 of the weight matrix of the LLAMA2 basic model; LoRA parameters are obtained by random initialization; S53, training LoRA parameters: use fine-tuning data to train LoRA parameters. During the training process, the weights of the LLAMA2 basic model remain unchanged, and only the LoRA parameters are updated. The update includes using the gradient descent method; S54, model fusion: In the inference stage, the trained LoRA parameters are fused with the LLAMA2 basic model weights. The LoRA parameters are combined with the last layer of LLAMA2 basic model weights through matrix addition or multiplication to obtain the automotive software defect repair evaluation model; S6. Perform LangChain integration to obtain an automotive software defect repair assessment model to complete automatic answers for specific documents; the LangChain integration includes: S61, text acquisition, including obtaining text-based vulnerability problem input; S62, text segmentation, including inputting the vulnerability question for segmentation processing or executing steps S2-S4; S63, text vectorization, including, using embedding technology to convert text-based vulnerability issues into vulnerability issue vectors; S64, vector similarity query, including querying other vulnerability problem vectors whose similarity with the vulnerability problem vector exceeds 90%; S65. Design a prompt input template, and based on the vulnerability question vector and other vulnerability question vectors with a similarity of more than 90% with the vulnerability question vector, select the final question vector and the prompt input template to generate a complete input question for the automotive software defect repair assessment model; S7. The automotive software defect repair assessment model outputs answers to questions about vulnerabilities based on complete input.
[0007] The present invention also provides an electronic device, comprising: at least one processor, and a memory communicatively coupled to at least one of the processors; The memory stores instructions that can be executed by at least one of the processors, and the instructions are executed by at least one of the processors to enable at least one of the processors to execute any one of the methods for constructing an automotive software defect repair assessment model based on LLAMA2.
[0008] The present invention also provides a computer-readable storage medium, on which computer instructions are stored. The computer instructions can be executed by one or more processors to implement any method for constructing an automotive software defect repair evaluation model based on LLAMA2.
[0009] The beneficial effects of the present invention are as follows: (1) By building a large model for evaluating automotive software defect repair, the input information of the large model includes the CVE number of the automotive vulnerability, the software operating system and hardware platform where the automotive vulnerability is located. The large model will give a decision on whether to repair the vulnerability, as well as the repair method and mitigation measures based on the user's input.
[0010] (2) Make a reasonable assessment of automotive software defects and reasonably reduce the cost of repairing automotive software defects. At the same time, it also lowers the difficulty threshold of repairing automotive software defects and improves safety work efficiency.
[0011] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above description and other purposes, features and advantages of the present invention more obvious and easy to understand, the preferred embodiments are specifically cited and described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 A flowchart of a method for constructing an automotive software defect repair assessment model based on LLAMA2; Figure 2 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of another electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0013] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0014] In the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be connected, detachably connected, or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0015] Example 1 A method for constructing an automotive software defect repair evaluation model based on LLAMA2 includes the following steps: S1. Collect pre-training data in a predetermined format; the predetermined format includes: (1) CVE vulnerability number, CVE vulnerability description, CVE vulnerability CVSS score, open source component where the vulnerability occurs, the scope of the open source component where the vulnerability exists, and the POC program and patch corresponding to the vulnerability. Specifically includes the following fields: CVE ID, vulnerability description, CVSS vulnerability level, affected components and version information, vulnerability link URL, and POC platform.
[0016] (2) The CWE number corresponding to the CVE vulnerability; specifically includes the following fields: CVE ID, CWE ID, CWE description.
[0017] (3) The scope of the hardware platform where the CVE vulnerability occurs; specifically includes the following fields: CVE ID, affected components and version information, hardware platform information where the vulnerability occurs, and the operating system used by the hardware platform.
[0018] (4) The components of the vehicle; specifically, the following fields are included: vehicle model information, component information included in the vehicle model, operating system used by the components, and hardware platform used by the components.
[0019] S2. For the pre-trained data in the predetermined format, perform word segmentation processing according to a Tokenizer; The word segmenter training comprises the following steps: Use the SentencePiece tool training command to train the Tokenizer, including S20, select subword mode for training by default; S21, select the BPE encoding method and select the amount of vocabulary data to be generated; S22, select a training data file, clean the text data in the training data file and convert it into UTF-8 encoding; S23, setting the character ratio that the model should cover; For example: spm_train -input=carbug_data.txt -model_prefix=carbug_model -vocab_size=2000 -model_type=bpe -character_coverage=0.9995 In this command: carbug_data.txt is a text data file; carbug_model is the generated model prefix; SentencePiece will generate two files: carbug.model and carbug.vocab.
[0020] 2000 is the generated vocabulary size.
[0021] 0.9995 is the character coverage, which means the model will try to cover 99.95% of the characters in the text; S3, expand the word segmentation results; expand the words in the word segmentation results into synonyms and near-synonyms; S4. Pre-train the LLAMA2 basic model based on the expanded word segmentation results. S5. Use the LoRA (Low-Rank Adaptation) method to fine-tune the instructions of the LLAMA2 basic model using the fine-tuning data, including: S51, obtaining fine-tuning data, wherein the fine-tuning data includes input words of automobile vulnerability data and corresponding output labels; after the fine-tuning data is collected, the data must be cleaned, preprocessed and standardized to ensure data quality and consistency.
[0022] S52, LoRA parameter initialization: LoRA parameters are introduced on the last layer of the LLAMA2 basic model, where the parameters are second-order matrices, and their rank is 1 / 10 of the weight matrix of the LLAMA2 basic model; LoRA parameters are obtained by random initialization; S53, training LoRA parameters: use fine-tuning data to train LoRA parameters. During the training process, the weights of the LLAMA2 basic model remain unchanged, and only the LoRA parameters are updated. The update includes using the gradient descent method; S54, model fusion: In the inference stage, the trained LoRA parameters are fused with the LLAMA2 basic model weights; the LoRA parameters are combined with the last layer of LLAMA2 basic model weights through matrix addition or multiplication to obtain the automotive software defect repair evaluation model; That is, it is achieved by introducing low-rank parameter adjustments to specific layers of the pre-trained LLAMA2 base model, so that the LLAMA2 base model can be adjusted according to the instructions of the fine-tuning task.
[0023] The data used for fine-tuning is stored in JSON format, and the fine-tuning uses the Stanford Alpace template: [{“instruction”:””, “input”:”” ,”output”:””},…] Instruction corresponds to the description of the execution task; The input corresponds to three questions, and the output corresponds to the expert's answer to each of the three questions: (1) Verify and identify CVE vulnerabilities and their impacts in automotive components; (2) Verify whether the CVE vulnerabilities in automotive parts need to be repaired and the repair method to be used if necessary; (3) Verify mitigation methods for CVE vulnerabilities found in automotive components.
[0024] S6. Perform LangChain integration to obtain an automotive software defect repair assessment model to complete automatic answers for specific documents; the LangChain integration includes: S61, text acquisition, including obtaining text-based vulnerability problem input; S62, text segmentation, including inputting the vulnerability problem for word segmentation processing or executing steps S2-S4; through the above processing, the vulnerability problem is segmented and the vocabulary is expanded, which expands the coverage and comprehensiveness of the vulnerability problem, and also makes the model's answer to the question more targeted.
[0025] S63, text vectorization, including, using embedding technology to convert text-based vulnerability issues into vulnerability issue vectors; S64, vector similarity query, including querying other vulnerability problem vectors whose similarity with the vulnerability problem vector exceeds 90%; S65. Design a prompt input template, and based on the vulnerability question vector and other vulnerability question vectors with a similarity of more than 90% with the vulnerability question vector, select the final question vector and the prompt input template to generate a complete input question for the automotive software defect repair assessment model; S7. The automotive software defect repair assessment model outputs answers to questions about vulnerabilities based on the complete input questions.
[0026] Among them, the LLAMA2 model is a relatively lightweight model that can run on mobile devices. The automotive software defect repair assessment model generated based on the LLAMA2 model is deployed in the mobile terminal, which is convenient for adapting to various usage scenarios, or can be used on Gradio's web-based question-and-answer platform to assist multiple people in their work.
[0027] The automotive software defect repair assessment model can perform repair assessment on the vulnerabilities found in the car, that is, whether they need to be repaired, and can give reasons for not needing to repair them, and can also provide methods or mitigation measures for repair.
[0028] Example 2 The complete input question design includes outputting answers to questions about vulnerabilities Design prompt templates: First, determine the questions that the model answers or the content that it generates. Design specific prompt input templates to give some initial context or questions. For example, for an automotive software bug fix assessment model, use a prompt similar to “Hello, I have a question about automotive vulnerabilities:”
[0029] Combine prompt with user input: Combine the user's actual input with the prompt template to form a complete input. For example, if the user enters "How to fix XXXbug?", the complete input may be "Hello, I have a question about car vulnerabilities: How to fix XXXbug". This will provide clear prompts for the model to guide it to generate relevant answers about mobile phone choices.
[0030] Pass input to the automotive software defect repair assessment model: Pass the designed complete input (including prompt and user input) to the automotive software defect repair assessment model for reasoning or generation.
[0031] The advantages of the present invention are: (1) By building a large model for evaluating automotive software defect repair, the input information of the large model includes the CVE number of the automotive vulnerability, the software operating system and hardware platform where the automotive vulnerability is located. The large model will give a decision on whether to repair the vulnerability, as well as the repair method and mitigation measures based on the user's input.
[0032] (2) Make a reasonable assessment of automotive software defects, that is, whether to repair them, whether to patch the source code project, or to repair them through mitigation measures such as shutting down services or enabling gateway isolation. Reasonably reduce the cost of repairing automotive software defects, while also lowering the difficulty threshold of automotive software defect repair work and improving safety work efficiency.
[0033] Example 3 like Figure 2 As shown, an embodiment of the present application provides an electronic device, including: at least one processor; and a memory communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that at least one processor can perform the above method. At least one processor in the electronic device can perform the above method, and thus has at least the same advantages as the above method.
[0034] Optionally, the electronic device also includes interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on a memory to display graphical information of a GUI (Graphical User Interface) on an external input / output device (such as a display device coupled to an interface). In other embodiments, if necessary, multiple processors can be used together with multiple memories, and / or multiple buses can be used together with multiple memories. Similarly, multiple electronic devices can be connected (for example, as a server array, a group of blade servers, or a multi-processor system), and each device provides some necessary operations. Figure 2 A processor 301 is taken as an example.
[0035] The memory 302, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the method for constructing an automotive software defect repair evaluation model based on LLAMA2 in the embodiment of the present application. The processor 301 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 302, that is, implements the above-mentioned method for constructing an automotive software defect repair evaluation model based on LLAMA2.
[0036] The memory 302 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 302 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some examples, the memory 302 may further include a memory remotely arranged relative to the processor 301, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0037] like Figure 3 As shown, the electronic device may further include: an input device 303 and an output device 304. The processor 301, the memory 302, the input device 303 and the output device 304 may be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.
[0038] The input device 303 can receive input digital or character information, and the output device 304 can include a display device, an auxiliary lighting device (e.g., LED), and a tactile feedback device (e.g., a vibration motor), etc. The display device can include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, and a plasma display. In some embodiments, the display device can be a touch screen.
[0039] Example 4 This embodiment provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable the computer to execute the above method. The computer instructions on the computer-readable storage medium are used to enable the computer to execute the above method, and thus have at least the same advantages as the above method.
[0040] The computer-readable storage medium in this embodiment can adopt any combination of one or more computer-readable media. The medium can be a computer-readable signal medium or a computer-readable storage medium. The medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of the medium (a non-exhaustive list) include: an electrical connection with one or more wires, 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 above. In this document, the medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.
[0041] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.
[0042] The program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF (Radio Frequency), etc., or any suitable combination of the above.
[0043] Computer program code for performing the operation of the present application can be written in one or more programming languages or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).
[0044] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A method for constructing an automotive software defect repair evaluation model based on LLAMA2, characterized in that: The following steps are involved: S1. Collect pre-training data in a predetermined format; S2, performing word segmentation processing on the pre-trained data in the predetermined format according to the word segmenter Tokenizer; S3, expand the word segmentation results; expand the words in the word segmentation results into synonyms and near-synonyms; S4. Pre-train the LLAMA2 basic model based on the expanded word segmentation results. S5. Use the LoRA method to fine-tune the instructions of the LLAMA2 basic model using the fine-tuning data; obtain an automobile software defect repair evaluation model; S6. Integrate LangChain to enable the automotive software defect repair assessment model to automatically answer specific documents; S7. The automotive software defect repair assessment model outputs answers to questions about vulnerabilities based on the complete input questions.
2. According to claim 1, a method for constructing an automotive software defect repair evaluation model based on LLAMA2 is characterized in that: The predetermined format includes: CVE vulnerability number, CVE vulnerability description, CVE vulnerability CVSS score, open source component where the vulnerability occurs, the scope of the open source component where the vulnerability exists, and the POC program and patch corresponding to the vulnerability.
3. The method for constructing an automotive software defect repair evaluation model based on LLAMA2 according to claim 1, characterized in that: The word segmenter training comprises the following steps: Use the SentencePiece tool training command to train the Tokenizer, including S20, select subword mode for training by default; S21, select the BPE encoding method and select the amount of vocabulary data to be generated; S22, select a training data file, clean the text data in the training data file and convert it into UTF-8 encoding; S23. Set the character ratio that the model should cover.
4. The method for constructing an automotive software defect repair evaluation model based on LLAMA2 according to claim 1, characterized in that: Using the LoRA method to use the fine-tuning data to fine-tune the instructions of the LLAMA2 base model; include, S51, obtaining fine-tuning data, wherein the fine-tuning data includes input words of automobile vulnerability data and corresponding output labels; S52, LoRA parameter initialization: LoRA parameters are introduced on the last layer of the LLAMA2 basic model, where the parameters are second-order matrices, and their rank is 1 / 10 of the weight matrix of the LLAMA2 basic model; LoRA parameters are obtained by random initialization; S53, training LoRA parameters: use fine-tuning data to train LoRA parameters. During the training process, the weights of the LLAMA2 basic model remain unchanged, and only the LoRA parameters are updated. The update includes using a gradient descent method; S54, model fusion: In the inference stage, the trained LoRA parameters are fused with the LLAMA2 basic model weights; the LoRA parameters are combined with the last layer of LLAMA2 basic model weights through matrix addition or multiplication to obtain the automotive software defect repair evaluation model.
5. The method for constructing an automotive software defect repair evaluation model based on LLAMA2 according to claim 1, characterized in that: The LangChain integration includes: S61, text acquisition, including obtaining text-based vulnerability problem input; S62, text segmentation, including, inputting the vulnerability problem and performing segmentation processing; S63, text vectorization, including, using embedding technology to convert text-based vulnerability issues into vulnerability issue vectors; S64, vector similarity query, including querying other vulnerability problem vectors whose similarity with the vulnerability problem vector exceeds 90%; S65. Design a prompt input template. Based on the vulnerability question vector and other vulnerability question vectors with a similarity of more than 90% to the vulnerability question vector, select the final question vector and combine it with the prompt input template to generate a complete input question for the automotive software defect repair assessment model.
6. The method for constructing an automotive software defect repair evaluation model based on LLAMA2 according to claim 5, characterized in that: The LangChain integration also includes: The text segmentation includes inputting the vulnerability problem and executing steps S2-S4.
7. An electronic device, characterized in that: include: at least one processor, and a memory communicatively coupled to at least one of the processors; The memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute the method for constructing an automotive software defect repair assessment model based on LLAMA2 as described in any one of claims 1-6.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, which can be executed by one or more processors to implement the method for constructing an automotive software defect repair assessment model based on LLAMA2 as described in any one of claims 1-6.
Citation Information
Patent Citations
Drug design method based on autoregressive model
CN116779060A
Banking industry large language model training method
CN117217289A
Question answering method and device based on vector database
CN118093808A
Large model data token processing method and device and large model training method
CN118364917A
Large model fine tuning method and device based on supervised fine tuning, equipment and medium
CN118445397A
