89 results about "Execution model" patented technology

The invention discloses a task processing method and device based on a large language model, equipment and a medium, and the method comprises the steps: carrying out the disassembly of a target task through the large language model, so as to obtain a plurality of subtasks and subtask information corresponding to the subtasks; based on the task description, the task category and the task difficulty corresponding to the sub-tasks, determining execution models corresponding to the sub-tasks; executing the plurality of sub-tasks by using an execution model to obtain a plurality of sub-execution results; and integrating the plurality of sub-execution results to obtain an execution result of the target task. The category and difficulty of an input task are identified by using a large model, and a complex task is disassembled into a plurality of sub-tasks. According to the difficulty level of each subtask, determining an execution model which is most matched with the current task in difficulty and scene in a model pool; large models of different parameter volumes are dynamically called, and resource waste is reduced to the maximum extent while task quality is guaranteed.

The invention relates to the technical field of task execution, and provides a workflow parallel computing execution method and system, terminal equipment and a storage medium. The workflow parallel computing execution method comprises the steps that a to-be-executed task is obtained, and after the to-be-executed task is subjected to logic analysis, the to-be-executed task is obtained; dividing a to-be-executed task into a plurality of sub-tasks according to the task type, constructing a task dependency graph, and making an execution model according to the characteristics of the sub-tasks; according to the resource states and the task priorities of the sub-tasks, resource allocation is carried out on the sub-tasks, and an execution sequence is set for the sub-tasks based on the dependency relationship in the task dependency graph; according to the execution mode and the execution sequence, the subtasks enter a parallel execution state, and an execution result of each subtask is obtained; and according to the original logic and data dependence of the to-be-executed task, summarizing, sorting and combining the execution results to obtain a summarized result, and verifying the summarized result to obtain a final execution result.

The invention discloses a method for evaluating the code generation capability of a fine-tuned credential large model, and relates to the technical field of artificial intelligence, and the method comprises the steps: S1, constructing an evaluation data set which is constructed according to a half-and-half rule, S2, executing a model generation test, and based on the evaluation data set, carrying out a model generation test; the method comprises the following steps of S1, adjusting a code output result, S2, calling the adjusted credential large model to generate a corresponding code output result, S3, carrying out multi-dimensional evaluation analysis, S4, forming a comprehensive evaluation result, and calculating the credibility 7 degree of the large model code generation capability for guiding model performance optimization and subsequent iteration improvement; according to the fine-tuned evaluation method for the code generation capability of the credential large model, evaluation deviation caused by a single index is effectively reduced, and the obtained quality credibility is more practical and explanatory and can be used as a decision basis for supporting large model optimization strategy formulation and parameter fine tuning.

The embodiment of the invention provides a data processing method and device and a model optimization method and device, and the data processing method comprises the steps: carrying out the sampling in a plurality of data sets, and determining target sample data according to a sampling result; inputting the target sample data into a plurality of language models for processing respectively, and obtaining target prediction data output by the plurality of language models respectively; performing text annotation processing on each piece of target prediction data by using a preset text annotation model to obtain text annotation information corresponding to each piece of target prediction data; based on the target sample data, the target prediction data and the text annotation information, a sample data set is constructed, and the sample data set is used for executing a model training task and a model verification task.

The invention relates to a model training task execution method, a graphics processing unit and a model training system.The model training task execution method comprises the steps that when attention calculation of model forward propagation is executed on a current text segment, the current text segment is subjected to attention calculation; sequentially loading key values corresponding to all the preamble segments from the central processing unit to carry out attention calculation on the activation values; meanwhile, the key value pair activation value corresponding to the current text segment is unloaded to the central processing unit; when the attention calculation of model back propagation is executed on the current text segment, loading a forward propagation output result and a key value pair activation value corresponding to the current text segment from a central processing unit, and loading the attention activation value corresponding to each preorder text segment in sequence; performing attention calculation on the activation value according to the forward propagation output result and the key value corresponding to the current text segment to obtain a gradient value; unloading the gradient value to a central processing unit; under the condition that GPU resources are limited, the execution efficiency of a text training task of a large language model can be improved.

The embodiment of the invention provides a drainage facility scheduling system and method based on a large language model, and the method comprises the steps: extracting a keyword in a request instruction through a scene analysis unit, generating a standardized task instruction according to the keyword, calling a database to verify the authenticity of the instruction, reading the instruction through a task recognition unit, and obtaining a task type; the task decomposition unit generates a planning path according to a task type, the task decomposition unit generates a plurality of task schemes, decomposes each task scheme to obtain a sub-task set of each scheme, identifies a task chain in each sub-task set, and the resource matching unit matches execution models corresponding to a plurality of sub-tasks and data resources called by the execution models according to the task chains. The analysis unit executes the sub-tasks according to the called data resources and generates execution results of a plurality of task schemes, the strategy unit analyzes the execution results to determine a scheduling execution scheme of the drainage facility, the system scheduling scheme and strategy are formulated based on a large language model, manual formulation is not needed, and the generation efficiency of the drainage system scheduling scheme is improved.

A computer-aided method for generating a computer program is described, which comprises the following steps: a) Providing a functional system model (FM) for the computer program to be created, b) automated execution of a model-to-model transformation (fm2spc) so that a specific system model (SPC) in a domain-specific language (DSL) is generated from the functional system model (FM), c) automated execution of a model-to-text transformation (spc2txt) so that a source code (TXT) for the computer program is created from the specific system model (SPC), d) Implementation of the source code (TXT) into an executable computer program (SUT), e) Generation of a set of executable test cases (TC) based on the functional system model (FM) and / or an optionally underlying original system model (MD), wherein the generation is at least partially automated, as well as f) Verification of the source code (TXT) automatically generated in step c) based on an automated test series of the executable computer program (SUT) using the set of test cases (TC) generated in step e). Furthermore, a computer program product and a development system are specified by carrying out the procedure.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a benchmark test-driven prompt word tuning method, device, equipment and medium, and the method comprises the following steps: analyzing input data to generate a benchmark test set; generating a cue word parameter configuration set based on the benchmark test set; calling a language model interface to process the input text to obtain a model output set; comparing the model output set with expected output to generate a comprehensive performance index set; updating a cue word parameter configuration set according to the comprehensive performance index set; and judging whether a termination condition is met or not, if not, continuing to execute model calling and parameter updating, and if yes, outputting a target cue word configuration and optimization report. According to the method, a structured evaluation and parameter iteration mechanism is introduced, cue word optimization is converted into a systematic process from artificial experience driving, repeated trial and error are reduced, cue word tuning efficiency is improved, and stability and mobility of the model in different tasks are achieved.

The invention relates to computer task processing, in particular to an efficient tagging check task processing system and method, a task decomposition and distribution module is used for decomposing a complex service request into a plurality of sub-tasks and intelligently distributing the sub-tasks; the concurrent execution engine is used for selecting a proper concurrent strategy according to task types and requirements, and performing multi-task concurrent execution based on the concurrent strategy through a concurrent execution model; the asynchronous task management module supports creation, scheduling, execution and result callback of asynchronous tasks, and simplifies asynchronous logic processing by introducing an asynchronous programming mode; the task scheduler is used for realizing asynchronous transmission and queue management of tasks by using the message queue and dynamically scheduling the tasks in the queue; according to the technical scheme provided by the invention, the defects of unreasonable resource allocation, poor task scheduling flexibility and low task execution efficiency in the prior art can be effectively overcome.

The invention relates to a model parallel training method and device for a heterogeneous system and computer equipment, and the method comprises the steps: constructing a plurality of mapping relationships between a to-be-trained model and each piece of computing equipment under each preset condition, and generating a strategy set; the preset conditions comprise the batch size of the training data, the pipeline parallel granularity corresponding to the to-be-trained model and the preset video memory budget of each computing device; traversing each preset condition, and processing different mapping relationships and each candidate strategy in the strategy set through a dynamic programming algorithm to obtain the execution cost of the heterogeneous system under the current optimal model parallel training strategy; and based on the model parallel training strategy corresponding to the minimum execution cost, executing model training through each computing device. By means of the method and device, the problem that efficient parallel training cannot be achieved through multiple computing devices in a heterogeneous system when a fixed processor is adopted for training is solved, efficient model parallel training is achieved, and the resource utilization rate of the computing devices is increased.

Provided in the present disclosure are a task execution method, and a method and apparatus for training a task execution model, which can be applied to the technical fields of artificial intelligence, natural language processing and large models. The method comprises: in response to having received an instruction for executing a target task, on the basis of a plurality of preset keys, performing information extraction on text corresponding to the instruction, so as to obtain a value information set, wherein the types of the preset keys comprise at least one of the following: an instruction object for indicating a subject in the text, an instruction action for indicating a predicate in the text and an instruction parameter for indicating an object in the text, and value information corresponding to the instruction object is used for representing an object required for executing the target task, value information corresponding to the instruction action represents an action required for executing the target task, and value information corresponding to the instruction parameter represents a parameter related to the object or the action; on the basis of the value information set, determining a plurality of sub-tasks used for implementing the target task; and on the basis of an association relationship between the plurality of sub-tasks, sequentially executing the plurality of sub-tasks.

The invention provides a multi-language-oriented source code automatic verification method and system, and relates to the technical field of computers.The method comprises the steps that multi-language source codes are obtained; compiling the multi-language source code through an LLVM compiler to generate an intermediate file; the intermediate file is optimized; analyzing the optimized intermediate file through an IR analyzer to generate a symbolic execution model; performing simulation execution operation on the symbolic execution model; and performing security verification on the symbolic execution model after simulation execution operation to obtain a verification report. Unified verification of multi-language source codes can be achieved, the verification environment building and development cost is greatly reduced, and the overall verification efficiency is improved; and the analysis capability of the intermediate file can be enhanced, the logic constraint relationship in the code can be accurately captured, the comprehensiveness and reliability of the verification work can be practically guaranteed, and diversified software security verification requirements can be met.

The invention discloses a deep learning model multi-NPU end-side one-button deployment system and method.The system comprises a front-end interaction module, a service-side control module, an end-side execution module and a result integration module.The method comprises the steps that front-end configuration is carried out on a user side, the front-end configuration is received on a service side, a unified version container environment is called to execute model conversion, and a unified version container environment is called to execute model conversion; pushing to the end side; receiving a file pushed by the service side at the end side, starting a reasoning script, and obtaining a reasoning result and NPU performance data; and receiving a side deployment report at a user side, and feeding back the report to the user through a front-end interface. By integrating the front-end interaction module, the service-side control module, the container environment management unit, the end-side execution module and the result integration module, automatic model deployment on various NPU devices is realized. One-stop automatic deployment is realized, from front-end configuration to end-side execution, the whole process of result analysis does not need manual intervention, and the efficiency is greatly improved.

An execution method of a machine learning model, comprising: generating output and a begin of sentence (BoS) cache of a BoS token using the machine learning model before or after performing model quantization on the machine learning model to generate a quantized model; and executing inference based on the quantized model, and during the inference, input the next token following the BoS token as a first input token and the BoS cache into the quantized model to generate output and cache of the next token, wherein the next token is based on the output of the Bos token or based on an input content.

An example operation may include one or more of executing an artificial intelligence (AI) model on a query via a software application to generate a formatted data structure, executing a filter on the formatted data structure and conditions of the formatted data structure to determine that the formatted data structure does not satisfy a condition from among the conditions, identifying a prompt that corresponds to the condition, executing the AI model on the formatted data structure and the prompt to generate a modified formatted data structure, executing the filter on the modified formatted data structure and the conditions associated with the formatted data structure to determine that the modified formatted data structure matches the conditions, and in response, deploying a software system via a host platform and executing the modified formatted data structure as part of the software system.

To provide an autonomous testing device for generating an object description described in a time-phase logic expression, and autonomously performing a test using the generated object description related to autonomous performance when a test object is changed in model checking.SOLUTION: An autonomous testing device 100 for executing model testing is provided with a test execution determining unit 90 and a time-phase logic expression generating unit 50. The test execution determining unit 90 determines whether or not to execute a test for a test target by using a test execution status for a target function and product update information that is information indicating the version of the target function. The time-phase logic expression generating unit 50 generates the object description of a time-phase logic expression by using multiplicity information indicating the multiplicity of each stereotype corresponding to each product included in the test object, metamodel information indicating the relationship between each operand name and each stereotype, and pattern information indicating the relationship between a template of the time-phase logic expression and a description rule of the object of the time-phase logic expression.SELECTED DRAWING: Figure 1

The invention provides a training method of a simulation system, an environment simulation method and device, electronic equipment, a computer readable storage medium and a program product. The method comprises the steps of triggering an execution model to execute first state updating through a regulation and control model to obtain a first prediction state of the execution model; triggering the execution model to apply a sample strategy through the regulation and control model, and triggering the execution model to execute second state updating after the sample strategy is applied, so as to obtain a second prediction state of the execution model under the influence of the sample strategy; and on the basis of a first difference between the first prediction state and a first preset state updated by the first state and a second difference between the second prediction state and a second preset state updated by the second state, parameters of the regulation and control model and parameters of the execution model are updated, and an updated regulation and control model and an updated execution model are obtained. According to the invention, the training efficiency and prediction accuracy of the simulation system can be improved.

The invention provides a task execution method and device, equipment, a storage medium and a program product, and the method comprises the steps that a thinking chain model automatically generates instruction information according to task description information and current state information of at least one mode; the instruction execution model obtains instruction information and state information, executes a corresponding instruction and outputs execution information; wherein the state information is updated after the instruction is executed; the matching model determines whether the execution result is matched with the instruction information at least according to the instruction information and the execution information; if the task description information is matched with the task description information, and the updated state information does not conform to the task completion state corresponding to the task description information, the updated state information is used as current state information to be input into the thinking chain model, and the thinking chain model continues to automatically generate next piece of instruction information according to the task description information and the state information; and the state information accords with the task completion state corresponding to the task description information.

This application provides an autonomous game operation AI agent method and device. The method includes: S10, capturing the game screen in real time after the AI ​​agent takes over the control of the game operation; S11, analyzing the game screen and tasks through a large language model (LLM) to generate tactical intentions; S12, generating operation instructions based on the tactical intentions and game screens through a pre-trained action execution model, and executing the operation instructions. During the execution of the above scheme, the game screen is analyzed to monitor whether the task is completed or failed. When the task is completed or failed, the control of the game operation is returned to the user. This fundamentally solves the pain point of players' "operation not keeping up with awareness". It can adapt to the dynamic changes and uncertainties in the game and execute highly complex and non-linear tasks, providing substantial operational assistance to users in playing games.

The invention relates to the technical field of building information models, in particular to a design collaborative modification method and system based on semantic packaging and BCF file transfer, and aims to convert a fuzzy natural language instruction into structured design modification semantics and transfer the design modification semantics through a BCF file, reduce communication cost and misunderstanding and ensure that the instruction is accurately understood; after approval is passed, the system can automatically execute model modification, low efficiency and error-prone performance of manual operation are avoided, and the advantages are more obvious especially for modification involving a large number of components or complex associations; by updating a BCF file and generating an XML log, information such as modification is completely recorded, and quality tracing, problem review and knowledge accumulation are facilitated; standardized semantics and BCF files provide common language and communication basis for designers with different professional backgrounds, and the cross-professional coordination process is simplified; and specific modification execution is automatically completed by the system, so that the dependence of common designers on complex BIM software operation commands is reduced.

The invention provides an autonomous game operation AI proxy method and device, and the method comprises the steps: S10, capturing a game picture in real time after an AI proxy takes over a game operation control right; s11, game pictures and tasks are analyzed through a large language model LLM, and tactical intentions are generated; s12, an operation instruction is generated through a pre-trained action execution model according to the tactical intention and the game picture, the operation instruction is executed, in the execution process of the scheme, whether a task is completed or failed or not is monitored by analyzing the game picture, and when it is monitored that the task is completed or failed, the game operation control right is returned to the user; according to the method and the system, the pain point that the operation of a player cannot keep up with consciousness is fundamentally solved, the method and the system can adapt to dynamic change and uncertainty in a game, highly complex and nonlinear tasks are executed, and substantive operation help is provided for the user to play the game.

The invention provides a software security test method and device, equipment, a medium and a program product, relates to the technical field of information security, and aims to solve the problem of poor test effect of a traditional application security test method in related technologies. The method comprises the following steps: analyzing a source code file of software to be tested to generate an abstract syntax tree; performing taint analysis on the abstract syntax tree to determine at least one code execution path of at least one API endpoint of the to-be-tested software as a taint source and at least one sensitive operation point as a taint sink; performing taint tracking simulation on the at least one code execution path based on a code execution model constructed by the abstract syntax tree to obtain an execution result; and under the condition that the execution result indicates that the to-be-tested software does not pass the security test, determining a problem code based on a vulnerability triggering path corresponding to the execution result. The software security testing method and device can improve the testing effect of software security testing.

The application provides a simulation scheduling method and system, and belongs to the technical field of control simulation. The method comprises the following steps: constructing a Simulink mathematical model and generating an initial model file; converting the initial model file into an RTW file; calling a target language compiler to convert the RTW file into a to-be-called model file; calling a make tool to automatically generate a code and to compile the to-be-called model file to generate an execution model file; and executing closed-loop simulation in a Vxworks system according to the execution model file. The application realizes simulation scheduling distribution of the whole system in the form of auxiliary clock scheduling tasks, realizes simulation scheduling between a Simulink simulation system and a Vxworks operating system, and improves the real-time performance of the Simulink simulation process.

The invention particularly relates to a spacecraft formation flight mission planning method based on cooperation of a large language model and a professional tool set. The method comprises the following steps: constructing a three-layer mission planning architecture of the professional tool set, an LLM and a high-level dynamic execution engine; constructing the professional tool set for spacecraft formation flight; disassembling the user instruction into a meta-task sequence through a complex task decomposition model; and calling the tool set in sequence to execute the meta-task through the task sequence execution model. The invention provides a more efficient and flexible intelligent solution with scene generalization ability for the spacecraft formation flight task planning problem.

The application particularly relates to a spacecraft formation flight task planning method based on cooperation of a large language model and a professional tool set, and comprises the following steps: constructing a three-layer task planning architecture of the professional tool set, the LLM and a high-level dynamic execution engine; constructing the professional tool set for the spacecraft formation flight; decomposing a user instruction into a meta-task sequence through a complex task decomposition model; and calling the tool set to execute the meta-task in sequence through a task sequence execution model. The application provides a more efficient, flexible and intelligent solution with scene generalization capability for the spacecraft formation flight task planning problem.

A computerized method and system for providing and implementing an artificial intelligence (AI) model that includes a plurality of models are provided. The method employs, for example, a computer having at least one processor and at least one memory and comprises receiving a query, identifying a user's context based on the query, and receiving the identified context through at least one artificial intelligence model and generating a response to the query, wherein the generating of a response may comprise controlling at least one router to determine at least one of a plurality of artificial intelligence models as a task execution model for the query, ingesting input data based on the context into the at least one determined task execution model, and sequentially or concurrently operating the at least one task execution model into which the input data has been ingested to generate a response to the query.

Provided are a data processing method and apparatus, and a model optimization method and apparatus. The data processing method comprises: respectively sampling in a plurality of data sets so as to determine target sample data on the basis of sampling results; inputting the target sample data into a plurality of language models for respectively processing to obtain target prediction data respectively outputted by the plurality of language models; using a preset text annotation model to perform text annotation processing on each piece of target prediction data to obtain text annotation information corresponding to each piece of target prediction data; and constructing a sample data group on the basis of the target sample data, the target prediction data, and the text annotation information, wherein the sample data group is used for executing a model training task and a model verification task.

The invention relates to the technical field of parallel computing, in particular to a GPU (Graphics Processing Unit)-based large-scale parallel computing optimization method, which comprises the following steps of: acquiring the minimum GPU resource quantity required by each queuing model task at the current moment in real time, and the load capacity, the batch size and the kernel function quantity of each execution model task in each GPU device; acquiring the operating power, the working frequency and the available resource quantity of each GPU device at each moment in real time; calculating a comprehensive evaluation score; and determining the chain scrambling weight, the parallel scrambling intensity, the frequency reduction performance waste degree, the parallel computing efficiency and the resource allocation priority of each GPU device at the current moment, and carrying out resource allocation on the GPU devices. According to the method, the problems of resource scrambling and frequency reduction caused by GPU resource fragmentation can be reduced, the load of GPU equipment can be balanced, and the efficiency of large-scale GPU cluster parallel computing can be improved.

The invention provides a model script migration method and device, a storage medium and a program product, and relates to the field of data processing. The model script migration method can be applied to a combined processing device. The combined processing device comprises a computing device, an interface device, a processing device and a storage device. The computing device is configured to execute a model script migration task, is mainly implemented as a single-core intelligent processor or a multi-core intelligent processor, is used for executing deep learning or machine learning computation, and can interact with the processing device through the interface device to jointly complete the model script migration task. The interface device is used for transmitting data and control instructions between the computing device and the processing device. As a general processing device, the processing device performs basic control including but not limited to data handling, starting and / or stopping of the computing device, and the like. The storage device is used for storing data to be processed. According to the scheme, the model script migration efficiency can be improved.

A first version of a model specified by a model execution request is executed, producing a first execution result. A second version of the model is selected according to an input data attribute specified by the model execution request. The second version of the model is executed, producing a first execution result. Using a natural language processing engine, responsive to the first execution result and the second execution result differing by more than a threshold amount, a natural language explanation of a difference between the first execution result and the second execution result is constructed.