441 results about "Object code" patented technology

The invention discloses a reasoning optimization method and device for a code generation large model, equipment and a medium, and relates to the technical field of model reasoning, and the method comprises the steps: in the reasoning process of a code generation task of a target code generation large model, executing a multi-granularity uncertainty quantification step in parallel every time a new Token is generated, obtaining a multi-granularity uncertainty score; constructing a state space vector, and utilizing a preset reinforcement learning strategy network to evaluate the selection probability of a plurality of preset reasoning optimization strategies based on a preset smooth decision mechanism so as to determine a target reasoning optimization strategy; if the strategy is a preset reasoning acceleration strategy, optimization processing is carried out through speculation decoding; if the strategy is a preset exploration optimization strategy, performing optimization processing by using a preset multi-path sampling technology and a preset knowledge enhancement technology; if the strategy is the preset fuzzy processing strategy, taking the plurality of candidate outputs as target reasoning outputs for optimization processing; and evaluating the decision effect according to the reasoning result to optimize the preset reinforcement learning strategy network.

The invention provides a code retrieval method and device for AI programming and related equipment, and the method comprises the steps: constructing a double-layer embedded vector model which comprises a basic semantic coding layer for parameter freezing and a plurality of independent lightweight adaptation layers which coexist; monitoring implicit feedback data generated by interaction of a user in the target code library to extract positive and negative sample pairs, and mining difficult negative samples based on a current model to form a training sample set; when the cumulative number of the samples reaches a preset trigger threshold value, optimizing and updating at least one lightweight adaptation layer parameter by utilizing a preset comparison loss function, and applying elastic weight constraint to key parameters of the lightweight adaptation layer parameter to obtain an updated double-layer embedded vector model; and performing multi-dimensional performance evaluation on the updated model, and when an evaluation result meets a preset publishing condition, deploying the model to a code retrieval service. The problems that in the prior art, cross-domain is difficult, the adaptation cost is high, and catastrophic forgetting is likely to happen are effectively solved, and zero-labeling self-evolution of the model is achieved.

The invention provides a multi-agent-based code conversion system and method. The method comprises the following steps: preprocessing an input source code; performing block processing on the source code to obtain source code blocks, obtaining variable information and array size information of the blocks corresponding to the source code, and integrating the variable information and the array size information into complete variable information; fusing the code description file, the code blocks of the source codes and the variable information of the blocks corresponding to the source codes to generate target code blocks; debugging and checking the target code blocks in combination with the source code blocks; the code description file is used for guiding the intelligent agent to establish a jumping type relation in the code conversion process, and it is ensured that the generated target code better meets the actual function requirement of hardware; a block identifier is inserted into a long code to divide the long code into a plurality of blocks, so that the limitation on long code conversion is relieved; codes of different styles can be output by simply adjusting the cue word of the intelligent agent; through multiple times of debugging verification, the conversion quality of the target code is ensured.

The embodiment of the invention discloses a visual fine adjustment method and device of an interface, a storage medium and a terminal. Firstly, a source code and a running state preview interface are synchronously obtained, and it is ensured that operation is based on a real rendering state. Then, when the user edits the interface elements, the system accurately positions the visual elements to the corresponding positions in the source codes during operation through a pre-established bidirectional mapping mechanism. Based on the accurate positioning, the system generates a code modification instruction and updates the target code. And finally, generating a new preview by running the updated code in real time, completing an operation feedback closed loop, and realizing a real-time editing preview effect. According to the method, manual modification of complex codes is avoided through the complete process, the modification precision is ensured while the low operation threshold is kept through intuitive visual operation, and the operation simplicity and control precision are effectively balanced.

The invention discloses an integrated circuit design Verilog code generation method and device based on a large language model, equipment and a medium, and relates to the technical field of computers, and the method comprises the steps that semantic analysis is conducted on an input language of an integrated circuit user side through the large language model, and a target constraint set is determined based on an obtained performance index set and through a regular expression; a candidate architecture scheme is generated by utilizing thinking chain technology reasoning, the candidate architecture scheme is predicted by utilizing an XGBoost regression model, and a target architecture scheme is determined based on an index prediction result and the candidate architecture scheme by utilizing a non-dominated sorting genetic algorithm; and determining an initial Verilog code based on the target architecture scheme and a preset code template library, and performing code style conversion on the initial Verilog code based on a code style feature corresponding to the integrated circuit user side to obtain a target Verilog code. And the efficiency and availability of integrated circuit design are improved.

The embodiment of the invention relates to a test code generation method, device and equipment and a readable storage medium, test codes are iteratively generated through an LLM model, coverage rate improvement is automatically verified, a compilation failure rollback mechanism and environment adaptation are combined, automatic closed-loop control of test code generation and coverage rate improvement is achieved, and the test code generation efficiency is improved. And the manual writing and debugging cost is effectively reduced. And the generated test code is subjected to multiple rounds of compilation verification, so that the grammar correctness and the execution feasibility are ensured. The generation direction is continuously optimized through a dynamic feedback mechanism, and the coverage range of unit testing on target codes is systematically improved. And finally, automatically terminating the process when a preset quality index is reached, and outputting an effective test code set which is fully verified.

The invention provides a front-end code evaluation method and system, electronic equipment and a program product, and the method comprises the steps: obtaining a front-end code, carrying out the compiling and rendering of the front-end code, and obtaining a generated front-end page and a target front-end page; performing code extraction processing on the generated front-end page and the target front-end page, and performing redundant node cleaning and code structure tree generation processing on extracted markup language codes to obtain a generated code tree and a target code tree; performing tree structure and element content error calculation processing on the generated code tree and the target code tree to obtain a content structure error; performing visual structure error calculation processing on the generated front-end page and the target front-end page to obtain a visual structure error; and according to the content structure error and the visual structure error, determining and obtaining an evaluation result for generating the front-end code. The scheme can improve the generation accuracy of the front-end code, and can be widely applied to the technical field of artificial intelligence.

The invention provides a method and device for code analysis, electronic equipment, a computer readable medium and a computer program product. The method comprises the following steps: performing code analysis on a target code to obtain multi-dimensional structured information of the target code; converting the obtained structured information into context information which can be understood by a large language model; performing semantic understanding and intention inference on the context information by using a large language model; and generating analysis result information of the target code based on the semantic understanding and intention inference results. According to the method, through multi-dimensional code analysis, grammar analysis and deep semantic understanding are remarkably improved, and the generation quality of a code analysis document is greatly improved; code logic and design intention are deeply understood by using a large language model, high-quality technical documents are automatically generated, and the intelligent level of code analysis is remarkably improved.

Systems and methods obtain a translation request accompanying source code of source file(s) for conversion. Formatting of the source code of the source file(s) is verified and the source code is converted to target code that includes target file(s). The converting (i) maps programming language functions of the source code to corresponding programming language functions of the target code and (ii) utilizes, through customized prompt template(s), one or more large language models (LLMs) hosted by a cloud platform during the code conversion, where a source application that uses the source code includes TIBCO™ software. The target file(s) are distributed.

The invention discloses a source code reconstruction method and device, equipment and a medium. The method comprises the following steps: acquiring a to-be-processed source code and reconstruction performance description content; determining at least one target code snippet in the to-be-processed source code according to a code reconstruction standard; according to the to-be-processed source code, generating a calling relation and an input and output relation corresponding to the to-be-processed source code; aiming at each target code snippet, reconstructing the target code snippets according to the calling relation and the input and output relation to obtain reconstructed code snippets; and replacing the target code snippets in the to-be-processed source codes with the reconstructed code snippets, wherein the to-be-processed source codes before replacement and the to-be-processed source codes after replacement have the same functions. According to the embodiment of the invention, the source code upgrading efficiency and automation degree of the original system can be improved.

The embodiment of the invention discloses an intelligent fuzzy testing method and device based on a large language model and a medium, and relates to the technical field of fuzzy testing, and the method comprises the steps: obtaining and analyzing a source code of a target code library to construct a structured knowledge graph, and evaluating a function vulnerability risk through the large language model and the structured knowledge graph, generating a test target sorting list; performing multi-dimensional code analysis on the selected test targets in the test target sorting list to generate seed input so as to construct a compilable test driver through the seed input; performing a test driver in an isolated environment, performing a fuzz test using a seed input, collecting a crash report to analyze the crash report, identify a crash cause, and generate a minimization replicator; and generating a candidate patch based on the crash reason, and verifying the validity of the candidate patch through a minimization reproducing device and a structured knowledge graph so as to realize the fuzzy test of the source code of the target code library.

The invention discloses a compiling method for compiling a C language source code into an RISC-V assembly code. The method comprises the following steps: acquiring a C language source code; performing lexical analysis on the C language source code to generate a mark flow; performing syntactic analysis on the mark flow, and constructing an abstract syntax tree; performing semantic analysis on the abstract syntax tree to generate a target abstract syntax tree; constructing a runtime environment of the RISC-V assembly code; generating an intermediate code and a control flow diagram corresponding to the intermediate code according to a rule in the runtime environment and the target abstract syntax tree; generating a target code by using the intermediate code and the control flow diagram corresponding to the intermediate code; according to the technical scheme, the intermediate representation more adaptive to RISC-V custom instruction mapping can be generated, so that the execution efficiency of assembly codes is improved; and meanwhile, by designing a lightweight runtime environment, the performance overhead is further reduced.

The invention provides a large language model operator compiling method oriented to a GPU platform, and belongs to the field of GPU operator compilation. A large language model operator is compiled through a Python interface to obtain a Python code; performing lexical analysis and grammatical analysis on the Python code to generate an abstract syntax tree; mapping the abstract syntax tree into an MMIR extension intermediate representation; tensor layout in the MLIR extension intermediate representation is uniformly converted into a binary linear mapping matrix, and optimization problems of global memory layout, shared memory layout and register block layout are analyzed based on the binary linear mapping matrix; automatically deducing a non-key tensor layout based on a program control flow diagram, generating a global optimal layout scheme, and obtaining an optimized MMIR extension intermediate representation; enabling the optimized MLIR extension intermediate representation to pass through a GGPU (Graphics Graphics Processing Unit) target code generator of the MLIR, and generating a PTX code adaptive to the NVIDIA GPU; the PTX code is loaded to a GPU, and an operator calculation result is output; the invention further provides a compiling system. The problem that the current large language model operator compiling efficiency is low is solved.

The invention discloses a GUI program directional fuzzy test method and system based on a large model agent, and the method comprises the steps: receiving a to-be-tested GUI program source code, constructing an event flow graph for the GUI program source code through static analysis, and carrying out the directional fuzzy test on the GUI program source code based on the event flow graph and a target code region specified by a user, a target association and dependency chain is obtained through reverse event flow analysis; based on the GUI real-time state and constraint information, an initial event sequence conforming to interaction logic is generated through a multi-mode large model; and on the basis of the initial event sequence, taking the event constraint relationship as a core, and carrying out directed evolution on the initial event sequence to obtain a test result. According to the method, deep fusion of the target driving capability of the directional fuzzy test and the complex interaction adaptation capability of the environment perception test is realized, and the problems of insufficient depth and low efficiency in deep vulnerability detection of the GUI program in the prior art are solved.

The invention relates to the technical field of data processing, and discloses a compiling method and system for generating hardware accelerator executable codes facing a multi-mode large language model, a terminal and a storage medium. The method comprises the following steps: constructing a computational graph according to the structure of a to-be-deployed large language model and operators supported by a hardware platform; traversing the computational graph to deduce output tensor type information of each operator, and verifying the type consistency between adjacent operators; based on the verified type information, executing static memory allocation to determine a static storage address of the weight data, and executing dynamic memory allocation to determine a dynamic storage address of the state data; then combining the verified type information and the static and dynamic storage addresses to generate an intermediate representation instruction sequence for configuring a hardware acceleration platform register; and finally, converting the intermediate representation instruction sequence into a target code file which can be directly loaded and executed by a hardware acceleration platform. The compiling efficiency, the deployment reliability and the resource utilization rate are improved.

The embodiment of the invention relates to a code completion method and device based on a code repository, and the method comprises the steps: firstly, obtaining a plurality of identifiers in a to-be-completed target code statement, and obtaining the first k lines of codes at the position of a cursor as target code snippets; then, on the basis of the identifiers, a plurality of code members meeting the grammar requirement of the position where the cursor is located are retrieved in a code repository; next, searching the positions of the identifiers and the code members appearing in a code repository, extracting code snippets serving as contexts based on the positions, and determining a code use example document based on m code snippets ranked in the top of the first similarity with the target code snippets; finally, inputting the target cue word into a pre-trained code completion model, and indicating the code completion model to generate a completion code; the target prompt word comprises the code use example document and a target code snippet.

The invention discloses an automatic construction method of an APK file and a related product. According to the method, firstly, a user construction demand and an APK code corresponding to the user construction demand are obtained, the code comprises application module codes of multiple versions, and the application module codes comprise a bottom layer package code and an HMI code. And then version filtering is carried out on the APK code based on a user construction demand to obtain a target code. And then interface consistency verification is carried out on an interaction interface of the bottom layer package code and the HMI code in the target code. And if the verification is passed, querying a construction workflow matched with the construction demand of the user in the knowledge graph to obtain a target workflow. On this basis, according to the target workflow, bottom layer package codes and HMI codes in the target codes are called and compiled in order, and an APK file matched with the construction requirement of the user is generated. According to the method, the automation level and accuracy of APK construction are greatly improved, and the problems of manual operation errors and version compatibility are reduced.

The present disclosure provides a file execution method and device, wherein the method first detects whether a target code object corresponding to a target file is created in response to an execution instruction for the target file; then, in the case that the target code object is created, the target code object is acquired; finally, an execution container is created under a current thread, and the target code object is executed by using the execution container. The present disclosure creates an execution container under the current thread, executes the target code object directly under the execution container of the current thread, does not need the main thread to allocate tasks, reduces the number of data transmission in the multi-thread concurrent scenario, and reuses the target code object, so that each thread does not need to perform bytecode of the file, and the execution efficiency of the multi-thread concurrent scenario is effectively improved.

The invention belongs to the technical field of software engineering, and discloses a token-level semantic matching-based code annotation generation method and system, and the method comprises the following steps: obtaining a target code and a data set, carrying out the preprocessing and dividing, and obtaining a training set as a corpus; performing word segmentation on the codes in the corpus by using a word segmentation device of a pre-trained large language model, and extracting and storing a context semantic embedding vector of each token in the codes to form a code semantic warehouse; calculating the lexical similarity between the target code and the code in the corpus, and taking the top k high examples as candidate examples; according to the method, a target code and a candidate sample are subjected to token-level semantic matching, the most similar token in another code segment is matched, the accuracy rate and the recall rate are combined to obtain an F1 score, and the sample with the highest score is reused, the context semantic feature of fine granularity is considered in the aspect of retrieval, the code annotation generation precision is improved, extra training is not needed, and the retrieval efficiency is improved. And the method has good applicability to different scenes.

The invention provides a code checking method, electronic equipment and a medium, and relates to the technical field of computers. The method comprises the steps of obtaining a first check result of a target code; the first check result is used for indicating a first code snippet which does not meet each code rule in a pre-constructed code rule set in the target code; determining a second check result according to the operation data of the target code and a preset operation condition; the operation data is used for representing the operation condition of the target code in the execution process; the second check result is used for indicating a second code snippet which does not meet the preset operation condition in the target code; generating a third check result of the target code according to the first check result and the second check result; the third check result is used for indicating a third code snippet in the target code; the third code snippets are code snippets in the first code snippets and the second code snippets. Through the method, the code is checked according to the operation data of the code, and the accuracy of code operation performance detection is improved.

The embodiment of the invention provides a code generation method and device, a storage medium and a program product. The method comprises the steps that first data corresponding to a source code is obtained, the first data comprises a first incidence relation, the first incidence relation indicates a mapping relation between nodes in a first knowledge graph and nodes in a second knowledge graph, the first knowledge graph corresponds to a first version corresponding to the source code, and the second knowledge graph corresponds to a second version corresponding to the source code; the second knowledge graph corresponds to a second version corresponding to the source code, and inputting the first data into the first model to obtain a target code under the second version. According to the method provided by the embodiment of the invention, the accuracy and controllability of version migration can be effectively improved by introducing the first association relationship as structured semantic guidance and inputting the first association relationship into the generation model.

Embodiments of the present application provide a code repair method and device, electronic equipment, storage medium and product. The method comprises: extracting binary features of a packed file, inputting a pre-trained convolutional neural network model to obtain target packing features corresponding to the packed file, performing unpacking processing on the packed file to obtain an unpacked file; performing decompilation processing on the unpacked file to obtain decompiled code, and performing semantic recognition on the decompiled code by a language recognition model to perform symbol recovery processing to obtain symbol recovery code; performing structural processing on the logic of the symbol recovery code by abstract syntax tree technology to obtain structured code; and performing compilation repair processing on the structured code to obtain target code. The above scheme accurately identifies the target packing type by using the pre-trained convolutional neural network model, automatically performs targeted repair according to the target packing type, can accurately repair the influence of the packing tool on the readability of the code, and thus improves the accuracy of code repair.

This application proposes a model-based question-answering method and apparatus for generating context based on grammatical structure. The method includes: selecting multiple initial code fragments related to the question information from a target code library; obtaining a source code file containing any one of the initial code fragments; locating a first code block node corresponding to the initial code fragment from the abstract syntax tree of the source code file; locating all ancestor nodes of the first code block node from the abstract syntax tree and extracting the code structure information of each ancestor node; generating a target code fragment based on the code structure information of the first code block node and all ancestor nodes; concatenating multiple target code fragments to obtain a target context; and inputting the target context into a question-answering model so that the model outputs answer information based on the target context. The embodiments of this application can effectively improve the question-answering performance of the model.

This application discloses code generation methods, model training methods, and related products, applicable to computer technology and AI fields. In this application, the code generation device deploys at least one model or has the ability to invoke at least one model. The device can acquire code to be converted written in a first programming language, perform semantic mapping on the code to be converted to obtain an intermediate representation, and use at least one model and the intermediate representation to obtain target code using a second programming language. This application utilizes at least one model to convert code into another programming language, achieving lexical-level semantic mapping during conversion, ensuring semantic consistency between the generated code and the original language, and enabling efficient and accurate conversion of the code's programming language. This application also provides corresponding model training methods and other related products.

The embodiment of the present application provides a kind of method and device for sharing information, it is related to terminal technical field, can reduce the development difficulty of information shared between equipment, improve development quality.Specific scheme is: determine the code corresponding to the first variable in the source code of application includes keyword, keyword is used to indicate that the first variable is multi-device shared variable;First target code corresponding to auxiliary variable associated with the first variable is generated, and the auxiliary variable is used to support the first variable shared between multiple devices when the application runs on multiple devices.The embodiment of the present application is used in the process of writing application code.

The invention discloses a code compiling device and method, and belongs to the technical field of computers. The device comprises a compiling module used for compiling a test program matched with a cost model for at least one section of candidate code, and the cost model is constructed by depending on the hardware performance of a target platform; the sending module is used for sending the test program to the target platform; the receiving module is used for receiving cost data obtained after the test program runs on the target platform, and the cost data is used for describing the running cost of at least one section of candidate code on the target platform; and the compiling module is also used for determining a target code from the at least one section of candidate code according to the cost data, and compiling the target code into a target machine code matched with the target platform. Therefore, the optimal machine code for the target platform is obtained, and user requirements are met.

The application provides a code modification method and device based on semantic update fragment generation. The code modification method comprises the following steps: generating a semantic update fragment and a first reconstruction code according to original code and modification description information; performing code modification verification on the semantic update fragment according to the code difference between the original code and the first reconstruction code; when the code modification verification is passed, generating a second reconstruction code by modifying the original code according to the semantic update fragment, and determining whether the code modification of the second code modification model is correct according to the code comparison result between the second reconstruction code and the first reconstruction code; taking the original code, the semantic update fragment and the first reconstruction code corresponding to the second code modification model with code modification errors as model training data to perform model training to obtain a target code modification model; and modifying the to-be-modified code by using the target code modification model to obtain a complete code. The application can reduce inference delay and improve code modification efficiency.

The invention discloses a code generation method and device based on a domain model, equipment and a medium, and the method comprises the steps: obtaining a target domain engineering model selected by a target user, and the target domain engineering model comprises an engineering structure model, an engineering component model and an engineering configuration model; analyzing the target domain engineering model based on a code generation engine, and generating a target engineering code corresponding to a target domain; and determining a target code warehouse corresponding to the target user, and pushing the target engineering code to the target code warehouse. Based on the technical scheme, the target domain engineering model is analyzed through the code generation engine, and the target engineering code is automatically generated, so that the preparation time required in the development process is shortened, the project starting efficiency is improved, and the initialization cost is reduced.

The invention relates to the field of software development, and discloses a code consistency detection method and device, computer equipment, a computer readable storage medium and a computer program product. The code consistency detection method comprises the steps of obtaining a target code and a comparison reference code in response to a consistency detection event for the target code; under the condition that a difference code exists between the target code and the comparison reference code, obtaining a plurality of difference dimensions for the target code; querying target information and comparison reference information associated with the difference code from code development information of the target code and the comparison reference code according to each difference dimension; determining a target difference dimension of the target code relative to the comparison reference code from the difference dimensions by comparing the target information with the comparison reference information; and generating a code consistency detection result of the target code based on the target information, the difference code and the target difference dimension. By adopting the method, the code consistency detection efficiency and accuracy can be improved.

This application discloses a method, apparatus, device, and medium for determining the adoption rate of intelligent code generation. The method includes acquiring initial code generated by an intelligent code generation model, comprising multiple first-version code fragments; determining the original attribute tags of the first-version code fragments based on their fragment metadata; encrypting the original attribute tags of the first-version code fragments based on the code version information of the initial code to obtain encrypted attribute tags; acquiring the target code corresponding to the initial code after actual use; determining the adopted target code fragment in the initial code based on the encrypted attribute tags of the second-version code fragment in the target code, the encrypted attribute tags of the first-version code fragment, the first-version code fragment, and the second-version code fragment; and determining the code adoption rate of the initial code based on the target code fragment and the first-version code fragment. This approach improves the accuracy and reliability of determining the sampling rate of intelligent code generation.