514 results about "Execution time" patented technology

The invention discloses a process automation execution method based on a large language model, and belongs to the technical field of artificial intelligence and process automation. User intention is analyzed through multi-modal input, and a structured task definition is constructed; the semantic reasoning layer is used for performing task layering, complexity evaluation and sorting optimization; the task execution layer completes subtask scheduling and execution; and the feedback and optimization layer performs performance evaluation and model updating based on execution data to realize closed loop and continuous optimization of the process, so that the technical problems of dynamically analyzing unstructured instructions, automatically optimizing a complex task dependency relationship and adapting to business changes in real time by a process automation tool are solved; according to the method, end-to-end conversion from an unstructured instruction to a structured task is realized, a subtask execution path is dynamically optimized, cross-platform tool calling is supported, the existing system integration cost of an enterprise is reduced, visual display task decomposition logic and prediction and execution time consumption comparison are provided, and the system credibility is enhanced.

The invention provides a timed task execution optimization method and system, and aims to solve the problems of inflexible static priority scheduling, unintelligent resource allocation and the like in the prior art. According to the method, a distributed timed task intelligent scheduling framework is constructed and comprises five core components including a task management console, a scheduling decision engine, a resource monitoring agent, a task execution cluster and metadata storage. Task basic information is obtained through task feature extraction, a dynamic priority score is calculated based on decision factors such as SLA urgency, resource matching degree and service weight, and task priority ranking is achieved. And performing intelligent task allocation by adopting a BestFit algorithm, and allocating the task to the optimal execution node. The system monitors the CPU utilization rate, the memory occupation and the IO waiting time of the nodes in real time, and when the CPU utilization rate exceeds 85%, an elastic resource allocation strategy is triggered. And when the task execution time exceeds the pre-estimated duration, performing task splitting and rescheduling based on a dynamic fragmentation algorithm.

The invention relates to a heterogeneous computing-based task scheduling method and device, a storage medium and equipment. The method comprises the following steps of: obtaining an inference task of a large language model; in the execution process of the reasoning task, at least based on the memory access intensity and the current reasoning stage, memory access intensive operators and calculation intensive operators in the reasoning task are recognized; allocating the memory access intensive operator to a first computing unit for executing a memory intensive task, and allocating a computing intensive operator to a second computing unit for executing a computing intensive task; obtaining estimated execution time of the two calculation units for executing the corresponding tasks and data transmission time between the two calculation units; according to the pre-estimated execution time and the data transmission time, the task starting moments of the first calculation unit and the second calculation unit are determined with the purpose of minimizing the overall execution delay of the reasoning task; and controlling the first calculation unit and the second calculation unit to asynchronously execute the corresponding tasks in parallel according to the task starting time.

The invention discloses a use case testing method and device, electronic equipment and a medium, and relates to the technical field of computers. The method comprises the steps that multiple to-be-tested use cases are acquired; determining a comprehensive weight of each to-be-tested case according to the priority information, the function modification information, the historical test fault information and the execution duration corresponding to each to-be-tested case; according to the comprehensive weight corresponding to each to-be-tested case, determining a sorting sequence of the to-be-tested cases; and grouping the to-be-tested cases according to the sorting sequence, and distributing each group of to-be-tested cases to test environments with different test efficiencies for testing. According to the application, the comprehensive weight of each to-be-tested case during testing is determined according to the priority information, the function modification information, the historical test fault information and the execution duration of different to-be-tested cases, and the to-be-tested cases are distributed to test environments with different test efficiencies for testing based on the comprehensive weight, so that the test period is shortened, and the test efficiency is improved. The test efficiency is improved.

The invention provides a timed task execution optimization method and system. The method comprises the following steps: forming a timed task execution optimization system architecture; determining a corresponding calculation factor to obtain a dynamic priority score; predicting the to-be-predicted task data by using the task prediction model to obtain a task execution time prediction value; analyzing the to-be-executed task queue, calculating the average waiting time, determining a multiple factor according to the task priority and the system load state, and identifying the task with the task execution time prediction value larger than the dynamic threshold value in the to-be-executed task queue as a long task; performing automatic fragmentation processing on the long task, generating a fragmentation result, determining a scheduling sequence based on the dynamic priority score and the task execution time prediction value, and generating a scheduling decision; and performing resource availability evaluation on the resource use data to obtain a node resource evaluation result, and allocating execution resources for the selected task. According to the method, the problems of efficiency and reliability of the timed task execution system in a high-load scene are solved.

The invention provides a model training method. The method comprises the steps of collecting information of a first model; and making a re-calculation strategy according to the communication time of the first model in the communication stage and the execution time required by the plurality of operators in the first model. The information of the first model comprises execution time respectively required by a plurality of operators in the first model, the recalculation strategy comprises at least one recalculation operator in the plurality of operators and opportunity for executing a recalculation process by the at least one recalculation operator, the recalculation operator is an operator used for executing the recalculation process in the plurality of operators, and the opportunity for executing the recalculation process by the at least one recalculation operator in the plurality of operators is the opportunity for executing the recalculation process by the at least one recalculation operator. The opportunity at which the at least one recalculation operator performs the recalculation process includes performing the recalculation process in parallel with the communication phase of the first model. The recalculation process is parallel to the communication process of the model, so that the model training time is shortened, and the throughput of model training is improved.

The invention provides a large language model application workload scheduling method, system and device, and relates to the technical field of model load scheduling, and the method comprises the steps: modeling a composite large language model application into a directed acyclic graph comprising a conventional stage, an LLM stage and a dynamic stage; modeling execution correlation among stages in the directed acyclic graph through a Bayesian network, dynamically predicting duration distribution of uncompleted stages, and calibrating a duration estimated value of an LLM stage in combination with a real-time batch processing size of an LLM executor; the uncertainty reduction amount of each ready stage is quantitatively scheduled based on information entropy; an epsilon-greedy strategy is adopted, and a JCT priority queue and an uncertainty reduction priority queue are combined to allocate scheduling resources; and assigning the task to a corresponding executor for execution, and repeating the above process until all operations are completed. The technical problem that the scheduling technology in the prior art is difficult to effectively deal with the execution time uncertainty and the structure uncertainty of the composite LLM application is solved.

The embodiment of the invention provides a compilation optimization method and electronic equipment, the method is executed by the electronic equipment, and the method comprises the steps that under the condition that starting of compilation optimization of an application is triggered, a plurality of decision parameters are obtained, and the decision parameters are related to execution time of compilation optimization; based on the multiple decision parameters, a target compiling optimization scheme adopted for compiling optimization is determined, and the target compiling optimization scheme comprises a full-amount compiling optimization scheme or an incremental compiling optimization scheme; and performing compilation optimization based on the target compilation optimization scheme. Therefore, when the electronic equipment starts compiling optimization of the application, the multiple decision parameters are obtained to comprehensively evaluate the target compiling optimization scheme adopted for compiling optimization, and the compiling optimization scheme with short execution time is selected to execute the compiling optimization task, so that the compiling time is optimized, and the running performance of the electronic equipment is improved.

The invention discloses a method and a system for asynchronous execution of JavaScript at a server side, and relates to the technical field of computer software, the method comprises the following steps: constructing a multi-language runtime context pool on the server side based on GraalVM, and pre-compiling a script and a core class; when the script is loaded, asynchronous execution strategy parameters are injected into the corresponding isolation context; calculating the fragmentation number of the overall task submitted by the script, and carrying out overall task fragmentation; mapping the fragments to the corresponding isolation assembly lines according to the fragment execution categories, constructing feature vectors according to the fragments in the isolation assembly lines, calculating fragment priorities, and executing in sequence; after fragment execution is completed, execution time delay is collected, historical time delay prediction is updated, and when conditions are met, isolation retry is carried out, and compensation fragments are generated. Through pre-compilation, asynchronous execution strategy, adaptive fragmentation and priority scheduling, and compensation fragmentation generation, fault self-recovery and service continuity guarantee with low time delay response, high scheduling flexibility and controllability and automation are realized.

The invention discloses a heterogeneous processor-oriented deep neural network reasoning task dynamic scheduling method and system, and belongs to the technical field of computer systems. The method comprises the steps that a DNN model is converted into a computational graph in a user space, and the computational graph is divided into a plurality of fine-grained micro-tasks through key path analysis; monitoring resource states of the CPU and the GPU in a kernel space in real time; dynamically selecting an optimal processor for each microtask in a kernel layer based on a cost model comprising execution time, queue length and utilization rate; and executing the micro-task on the corresponding processor according to a scheduling result, and triggering cross-processor task dynamic migration when detecting that the load is unbalanced, including context storage, data transmission and execution recovery. The system adopts a user space and kernel space collaborative architecture to realize the method. Through fine-grained division, kernel-level real-time scheduling and a dynamic migration mechanism, the reasoning efficiency, the resource utilization rate and the system stability of the heterogeneous system under a complex load are effectively improved.

The invention belongs to the technical field of computers, and particularly relates to a resource allocation method and device, equipment and a storage medium. The method comprises the following steps: S1, acquiring a target resource occupation task and real-time state data of a plurality of clusters; s2, calculating a cluster resource utilization rate based on the total resource configuration amount and the resource usage amount; calculating a residual resource proportion of the cluster based on the resource residual amount and the resource demand parameter; calculating a load sensitivity coefficient of the target resource occupation task based on the task priority, the resource demand parameter, an execution duration threshold and a task dependency identifier; and S3, screening the clusters according to preset screening conditions corresponding to different task guarantee levels, determining the clusters meeting the preset screening conditions as candidate clusters, monitoring in real time, triggering resource migration based on quantitative migration conditions, and forming a closed-loop process of data acquisition-index calculation-screening distribution-preemption coordination-monitoring migration.

The invention discloses a production line unit task allocation method based on reinforcement learning, which comprises the following steps: extracting a specific task instruction of each production line unit from a final task allocation matrix, verifying the feasibility of the instruction under the constraint of completion time through a simulation execution module, determining an instruction set passing verification, and performing task allocation on the instruction set; wherein the instruction set is obtained by fusing the solution of the multi-target conflict; according to the verified instruction set, production line feedback data such as actual execution time and energy consumption records are obtained, deviation is analyzed from the feedback data, if it is judged that the deviation is larger than a preset threshold value, a self-adaptive adjustment mechanism is triggered, and a corrected distribution strategy is obtained; and the optimized resource allocation indexes are extracted from the corrected allocation strategy, the indexes are pushed to the production line equipment through the real-time distribution system, the execution monitoring cycle after pushing is determined, and the monitoring cycle is obtained by continuously tracking the low-efficiency risk.

The invention relates to a real-time hybrid test control and test system and method based on a Linux-RT system and application, and belongs to the technical field of real-time hybrid tests of structural engineering. Comprising the following steps of: realizing deterministic conversion of states by adopting a real-time system manager of a singleton mode and an atomic state variable; according to the real-time executor based on the SCHEDFIFO scheduling strategy, affinity binding of a high-priority thread and a CPU is set, and periodic task scheduling of microsecond-level precision is achieved; a lambda functional programming data processing pipeline is adopted, and efficient data transmission of a filter chain is achieved through a zero copy mechanism; multi-protocol communication interfaces of UDP (User Datagram Protocol), EtherCAT (Ethernet Control Automation Technology) and WebSocket are supported, and a uniform abstraction layer is provided to realize low-delay data exchange between a numerical value and a physical substructure; and a real-time performance monitoring module adopting an exponential moving average algorithm provides accurate statistics of execution time and jitter.

The invention relates to an analog quantity selection method and system based on a Boolean switch, in particular to the field of industrial automation, and aims to directly realize signal mapping by adopting a pure hardware combinational logic circuit and thoroughly eliminate instruction execution time, so that the state change of a Boolean switch signal can be instantaneously reflected on analog quantity output; according to the method, real zero-delay switching is achieved, meanwhile, program adjusting algorithms in any form are avoided, disturbance phenomena such as overshoot or undershoot are completely eradicated from the principle, and high stability and accuracy of output signals are ensured. Finally, according to the technology, the control instruction can be accurately and synchronously issued without delay, and the response speed and the synchronization precision of the whole control system in a high dynamic response industrial scene are remarkably improved.

The invention discloses a task scheduling method and device for attention calculation, a medium, equipment and a product. The method comprises the steps that a loading task of a query block is decomposed into N1 loading subtasks, and a calculation task of a jth attention score block is decomposed into N2 calculation subtasks; decomposing an updating task of a middle accumulation block corresponding to the attention output block in the current iteration round into N3 parallel updating sub-tasks; asynchronously submitting all the subtasks to different task slots of N4 synchronization channels; and performing polling operation on the task slot in each synchronization channel to execute the sub-tasks stored in the task slot, and controlling an execution time sequence among the sub-tasks with the dependency relationship in each synchronization channel by configuring a number threshold value of the uncompleted sub-tasks in the waiting instruction. According to the method, the covering capability and the parallel computing capability of hardware can be improved, and the attention computing efficiency is further improved.

The invention provides refrigeration equipment, relates to the technical field of household appliances, and is used for realizing operation of an artificial intelligence algorithm while a refrigeration function of a refrigerator is not influenced under the condition that resources of a main control chip of the refrigerator are limited. The refrigeration equipment comprises a box body, a refrigeration system, a heater, a memory and a controller, wherein a target model is deployed in the memory; the controller is configured to split a target algorithm task corresponding to the target model into a plurality of algorithm sub-tasks; executing at least one algorithm sub-task in the plurality of algorithm sub-tasks in the residual execution duration of at least one execution cycle of the plurality of execution cycles; wherein the remaining execution duration is the remaining duration after all the core tasks are executed in each execution period, and the core tasks are used for achieving normal operation of the refrigeration equipment; and running the plurality of execution cycles until the plurality of algorithm sub-tasks end running, and obtaining a running result of the target model.

The invention discloses an integrated design management system with an automatic registration communication agent engine and a method thereof, and relates to the technical field of software integrated design simulation. The method has a heterogeneous integration breakthrough, the five communication modes are matched with the software without the API / command line / file, different heterogeneous software can support one communication mode, and the integration cost is remarkably reduced; dynamic expansibility is achieved, a newly added software module only needs to deploy a proxy engine and configure registration, and zero-shutdown upgrading of the platform is achieved; high concurrency reliability is achieved, resource conflicts are solved for an instance mechanism, and parallel task efficiency is improved; and full-process automation is achieved, and the cross-software design process execution time is obviously shortened.

The invention relates to the technical field of hardware utilization rate calculation, and discloses a method and device for determining the hardware utilization rate, equipment, a medium and a program product, and the method for determining the hardware utilization rate comprises the following steps: counting first total execution time and second total execution time of tasks executed on hardware, updating the first total execution time based on the first dynamic compensation time, and updating the second total execution time based on the second dynamic compensation time; the first dynamic compensation time and the second dynamic compensation time are used for compensating the execution time of the task executed on the hardware; and determining a hardware utilization rate in a specified time period by utilizing the updated first total execution time, the updated second total execution time and the specified time period. Compared with the prior art, the accuracy of determining the hardware utilization rate in the specified time period can be improved.

This prevents the system from inferring that the setting has been changed. [Solution] The performance control means executes a first performance related to a specific benefit when a specific benefit is granted, does not execute the first performance when a specific benefit is not granted as a result of the operation in the second operation mode, does not execute the first performance when the stop switch is operated in the second operation mode during a non-advantageous section and a specific benefit is granted in the next game, performs performances through multiple layers including a first layer and a second layer which is a higher layer than the first layer, executes the first performance in the first layer and executes a second performance which is different from the first performance in the second layer, and makes the execution timing of the first performance and the execution timing of the second performance different.

The invention relates to a vulnerability mining method and device of an industrial control protocol, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a historical asynchronous event of a node, calculating a time difference value between a preset time point and a completion time point of the historical asynchronous event, taking a node corresponding to the time difference value not within a preset duration range as a conflict node, taking a current moment state of the conflict node as a state interaction starting point, and performing state interaction on the conflict node; respectively constructing an original state path and a conflict path, and generating a fusion structure chart; and extracting dependency field information in the fusion structure chart, comparing the dependency field information of the state nodes to obtain the state nodes with differences, calculating the response interval duration of each state node according to the execution starting time and the achievement time of the state nodes, and generating an abnormal response time sequence fragment set under the condition that a sudden change phenomenon exists. By adopting the method, the accuracy of vulnerability mining can be improved.

The application discloses a kind of cluster job scheduling method and system, the computing job submitted by user is modeled as directed acyclic graph DAG, to convert real-time scheduling problem into the allocation problem between resource and DAG.The application includes two scheduling modes of normal scheduling and special state scheduling, to realize efficiency target and fair target respectively.The normal scheduling mode is based on graph neural network, job information is collected from local to whole, and these information is input to reinforcement learning model, to give scheduling decision of minimizing average job completion time.The special state scheduling mode detects long tail job according to the statistical law of job execution time in real time, and carries out special optimization for long tail job, to reduce long tail job completion time by preferentially allocating resources for it.The scheduling system is converted between the two scheduling modes, to achieve the trade-off between efficiency and fairness.

In an industrial processes, a properly instrumented line facilitates capture of data including detected steps, application input, and execution graph transitions, that permit the creation of empirical models of process timing. In this context, a process controlled by individual applications, e.g., at manufacturing workstations, provides a proxy for overall process timing by dividing a workflow into a number of discrete steps completed at each workstation, and further into any number of sub-steps, each controlled by a user and explicitly completed, e.g., by user interactions with widgets or other controls of the application. These applications provide a useful framework for modeling execution timing by providing an initial, implicit model for workflow (based on application control logic) that also facilitates automated detection of process sub-steps based on execution flow, as well as detection and measurement of the contributions of individual widgets and / or combinations of widgets to the process timing. By gathering data in this manner, mixed statistical distributions can be applied based on individual timing data for each possible sub-step, widget, process step, and the like performed with each application.

The invention discloses a GPU cluster resource allocation method and device, equipment and a storage medium, and relates to the technical field of GPU cluster resource allocation. According to the method, a hardware performance description vector is generated by fusing static hardware parameters and dynamic micro-benchmark test data, the limitation that the GPU performance is represented only by depending on the static parameters is broken through, and the actual operation capacity of different architecture GPUs in a heterogeneous cluster is accurately matched; task feature vectors are generated by extracting task calculation features and resource demands, and quantitative description of reasoning task resource consumption features is achieved; information of hardware, tasks and load dimensions is integrated through a machine learning model, performance degradation characteristics during multi-task parallel are effectively captured, and the accuracy of execution time prediction is improved; scheduling decision operation including candidate node screening, comprehensive cost evaluation and resource reservation backfilling is executed in combination with the predicted execution time, and the task execution efficiency and cluster load balancing are both considered.

Techniques for facilitating efficient polling using machine learning are disclosed. A system uses historical data associated with execution of tasks to train a machine learning model to predict execution times. After receiving a request for execution of a task, the system provides a polling configuration to the requesting device that includes a polling frequency based on a prediction for when the task execution will be completed. This prediction is generated by the machine learning model.

Embodiments of the invention provide a model input request scheduling method and device, a storage medium and a program product. The method comprises the steps of receiving an input request of a preset language model; determining prediction time consumption of the input request in a pre-filling stage in a preset language model processing process; determining the latest execution time of the pre-filling stage of the input request according to the receiving time of the input request, the predicted consumed time of the pre-filling stage and the target first lexical element delay; and according to the latest execution time of the pre-filling stage, sending the input request to a service node of a preset language model for processing. According to the embodiment of the invention, the latest execution time of the pre-filling stage is determined based on the predicted time consumption of the pre-filling stage of the input request to schedule the input request, and the system throughput and the calculation load balance are optimized while the service level target is met, so that the resource utilization efficiency and the service quality guarantee capability of the service node of the preset language model can be improved.

The invention relates to the technical field of heterogeneous computing, in particular to a prediction model dynamic scheduling system and method based on a heterogeneous computing platform. Running the prediction model in the edge host, and abstracting a calculation-intensive operator in the model into a candidate unloading task; and constructing a time cost model containing host execution time, FPGA execution time and transmission overhead, and dynamically updating in combination with execution statistics. And the scheduling module compares the predicted completion time of the local and unloading paths according to a time cost model, and selects to execute a prediction task on the host or the FPGA. A page locking DMA buffer area is arranged in a system memory, and the FPGA directly reads and writes historical data; and the FPGA side schedules tasks according to the priority and the deadline, and returns to a traditional closed-loop correction time cost model. Compared with the prior art, the method can dynamically adapt to illumination conditions and system load changes, reduces the average reasoning time delay and high quantile tail time delay of tasks, and improves the performance and stability of edge side photovoltaic power prediction.

The invention provides a method for executing a dynamic quantum circuit on a superconducting quantum computing device, a superconducting quantum computing system and a computer readable storage medium, and the method comprises the following steps: dividing the dynamic quantum circuit into at least one main circuit and at least one control sub-circuit, the main circuit is a circuit which is not influenced by a measurement result in the dynamic quantum circuit during execution, and the control sub-circuit is a circuit which is controlled by depending on the measurement result in the dynamic quantum circuit during execution; and respectively carrying out routing compiling on the at least one main line and the at least one control sub-line so as to sequentially execute the at least one main line after routing compiling and the control sub-line which is determined to be required to be executed after routing compiling on the superconducting quantum computing equipment according to the execution time sequence of the dynamic quantum line.

The invention provides a surface code-based quantum circuit fault-tolerant compiling method and a surface code-based quantum circuit fault-tolerant compiling device. The method comprises the following steps: acquiring an input set and a quantum circuit; aiming at each initial parameter combination, according to the corresponding bit block number parameter and the corresponding factory number parameter, estimating execution time required for executing the quantum circuit; for each initial parameter combination, generating line space-time cost information of the quantum line according to the corresponding bit block number parameter, the corresponding factory number parameter and the corresponding execution time; determining a target parameter combination from the plurality of initial parameter combinations based on the respective line space-time cost information of the plurality of initial parameter combinations; and according to the target parameter combination, mapping the logic bit of the quantum circuit to the physical space of the chip, and scheduling the quantum operation in the mapped physical space to obtain a physical execution sequence.

The invention provides a task flow dynamic optimization method and system based on data feedback, and the method comprises the steps: obtaining process feedback data, determining the process execution time consumption and a processing result through a production monitoring system log, recognizing the current processing capability of a process, and obtaining the to-be-processed workpiece amount of a downstream assembly node and the frequency fluctuation amplitude of an upstream process output workpiece; acquiring the queuing workpiece quantity and the real-time load quantity of each process corresponding to the rearranged process processing execution sequence, analyzing to obtain a process processing capacity matching degree index, and determining a global workpiece circulation stability evaluation result; and obtaining total number change data of workpieces completed in each process from the dynamic optimization implementation scheme, judging the system processing throughput improvement degree, and determining the process load balancing state of workpiece circulation.

The invention provides a service deployment method and a service deployment system based on Kubernetes. The method comprises the following steps: receiving a deployment instruction carrying an overall planning script package, wherein the overall planning script package comprises an overall planning file, a global configuration file and a plurality of sub-script packages; analyzing the overall arrangement file to determine the execution sequential logic of each sub-script package, and extracting a target configuration parameter from the global configuration file; the sub-script packages are scheduled in sequence, sub-script package description files in the sub-script packages are read, and instantiation deployment parameters are generated in combination with local parameters; analyzing an instruction definition file in the sub-script package to identify an extension instruction; and executing the instruction according to the instantiated parameters, and completing container deployment, database initialization and configuration issuing of business services by issuing an atomic operation instruction to a Kubernetes cluster. According to the method, a complex deployment process is abstracted into a unified arrangement and scheduling process, and the efficiency and reliability of service deployment are remarkably improved.