102 results about "Dynamic dispatch" patented technology

The invention relates to the technical field of PC (Personal Computer) terminal AI (Artificial Intelligence) computing, and discloses a method and a system for dynamically scheduling generative AI heterogeneous computing resources of a PC terminal. The system comprises a resource state acquisition module, a scheduling graph generation module, a resource fluctuation entropy analysis module and a scheduling decision engine module. The resource state acquisition module captures running state parameters of a GPU kernel, a CPU thread and a memory block in real time, and generates a resource state feature tensor through normalization processing; the scheduling atlas generation module analyzes and computes the node connection topology, extracts the correlation between the devices, and constructs a multi-dimensional scheduling atlas; the resource fluctuation entropy analysis module separates the load feature vectors, calculates the entropy of each calculation unit and generates a heterogeneous resource entropy matrix; and the scheduling decision engine module jointly analyzes the atlas and the matrix, identifies bottleneck node resource competition characteristics, generates a dynamic scheduling instruction set, adapts to generative AI task requirements, and ensures efficient and stable operation of the task.

The invention discloses a dynamic scheduling method and device for mass call requests and a medium, and relates to the technical field of information management, and the method comprises the steps: receiving mass call request information through an information sending device, carrying out the language processing to generate an analysis data set, and obtaining the request types through deep learning classification; performing multi-objective optimization on the analysis data set to generate a dynamic call queue; constructing a time sequence prediction model based on historical records, predicting a call peak, carrying out resource scheduling analysis, and formulating a scheduling strategy; and generating a feedback result after executing the scheduling strategy, synchronizing the feedback result to the call center, and updating the strategy according to response information to realize intelligent dynamic scheduling. The technical problem that in the prior art, the resource scheduling efficiency of the call center in the mass request processing process is low is solved, and the technical effect of optimizing the resource scheduling efficiency of the call center is achieved.

The invention relates to a large language model and multi-agent collaborative power grid task self-matching and dynamic scheduling method and related equipment. The method comprises the following steps: analyzing and modeling task description information by using a large model to obtain initial structured task information, calibrating the initial structured task information based on a power grid task semantic graph to obtain standard structured task information, and generating a power grid task based on the standard structured task information; performing multi-dimensional vector modeling to obtain a resource object; calculating a matching degree between the power grid task and the resource object, and determining a candidate resource object of the power grid task based on the matching degree; enabling the task agent and the resource agent to perform cooperative scheduling, determining a target resource object of the power grid task, and generating task resource mapping information; and examining the task resource mapping information and sending orders. The task semantic analysis accuracy can be improved, the matching precision of resources and tasks can be improved, the scheduling self-adaption and optimization capability can be improved, and the compliance risk can be effectively avoided.

The invention discloses a tool calling method based on a large language model, and the method comprises the steps: obtaining a question text which comprises task information; obtaining a matched first candidate tool set in the question text and a preset tool library; screening the first candidate tool set on the basis of locally available current hardware resources and / or available current online resources of a cloud application program interface (API) to obtain a second candidate tool set; determining a target tool matched with the task information from a second candidate tool set based on a large language model, and generating a tool calling instruction based on the target tool; selecting a called target tool based on the tool calling instruction, and obtaining an output result of the target tool; based on the question text and the output result, reply text corresponding to the large language model is obtained. According to the method, the tool with a higher matching degree can be obtained in combination with the resource occupation conditions of the tools of multiple platforms, and efficient utilization of resources and dynamic scheduling of task tools are realized.

The invention discloses a hierarchical dynamic scheduling method based on a strategy optimization DDQN algorithm, which is used for solving a dynamic scheduling problem of a reentrant hybrid flow shop with a batch processor, and comprises the following steps: firstly, determining an objective function and a constraint condition of the scheduling problem; secondly, by introducing a self-attention mechanism, providing a layered structure based on the self-attention mechanism DDQN, and constructing two agents, namely a batching agent and a scheduling agent, to respectively solve batching sub-problems and scheduling sub-problems in the problem; besides, in order to cope with the multi-stage batch processing and reentry scheduling characteristics in the problem, a Markov decision process considering the characteristics of the two agents is designed, and the Markov decision process comprises states, actions and rewards. Furthermore, an action selection strategy and a soft start target network updating strategy based on a mask combined with an epsilon-greedy strategy are provided, so that the efficiency and generalization ability are improved. The method has a remarkable effect on solving the dynamic scheduling problem of the reentrant hybrid flow shop with the batch processor.

The invention relates to the technical field of medical informatization and intelligent scheduling, and particularly discloses a ward nursing process prediction and dynamic scheduling method and system.The ward nursing process prediction and dynamic scheduling method comprises the steps that firstly, a digital twinborn simulation environment synchronized with a physical ward in real time is established, and a nursing process is simulated through a multi-agent network; carrying out multi-time sequence deduction based on the environment, generating a prospective scheduling instruction set, and issuing and executing the prospective scheduling instruction set; calculating a prediction mismatch degree index by collecting actual execution data and comparing the actual execution data with a plurality of dimensions of an expected state; when the index exceeds a dynamic threshold value, a meta control layer is triggered to carry out self-adaptive adjustment on simulation environment parameters, and the self-adaptive adjustment comprises reconstruction of an agent behavior logic weight or switching of a scene deduction mode; and deducing and generating an updated instruction set again based on the adjusted environment to form closed-loop optimization.

The invention provides a task analysis execution method and system based on a multi-modal large model, and belongs to the field of computers.The task analysis execution method comprises the steps that S1, multi-modal task data are received, a double-layer cross-modal fusion mechanism is adopted to conduct unified coding fusion on the multi-modal task data, and fusion features are obtained; s2, performing alignment processing on the fusion features based on a multi-modal embedding alignment mechanism to obtain alignment features, and performing structured task analysis based on the alignment features by using a pre-training language model to obtain task description; s3, decomposing the composite task into atomic tasks, and learning an optimal decomposition strategy by using an Actor-Critic network to form an atomic task set; and S4, generating a comprehensive adaptation score through a dynamic scheduling mechanism of collaborative awareness, and performing task distributed execution based on the comprehensive adaptation score. According to the method, the complex task is intelligently decomposed into the atomic task set capable of being executed in parallel for task distributed execution, and the task understanding accuracy and execution efficiency are improved.

The present invention relates to a parallel simulation method, medium and device based on load balancing, which obtains several sub-models by splitting a complex system model, obtains performance index data of each computing node, allocates computing nodes corresponding to each sub-model according to the performance index data, determines the execution order and scheduling mode of the sub-models according to the priority scheduling strategy, dynamic scheduling strategy and resource conflict avoidance strategy, and controls the computing nodes corresponding to each sub-model to perform simulation operations according to the execution order and scheduling mode of the sub-models to obtain simulation results. The method solves the problem that the existing technology cannot improve the simulation speed and resource utilization while ensuring the accuracy of the simulation results.

The present disclosure is directed to a scheduler in a real time operating system for a central processing unit (CPU) that runs a number of, and diverse set of processes in a ready queue, the scheduler when executed by the CPU performs steps including dividing the ready queue of P processes into two lists, list 1 contains n processes while list 2 contains m processes, where n+m=P, averaging total execution times of the processes in each list, generating a first quantum s based on an average of each list, generating a second quantum t based on the average of each list, comparing total execution time of a candidate process which needs to be assigned to the CPU with the quantum values, and assigning the value of a quantum, based on the comparing, that is the quantum value closest to the total execution time of the candidate process, to that process as a time period for running on the CPU.

The invention provides a combined optimized dynamic scheduling method. The method comprises the steps of forming a scheduling task queue based on received order tasks; traversing from the head of the scheduling task queue to the tail of the scheduling task queue, and merging orders of the same item to form a merged scheduling task queue; performing local order sequence change on the merged scheduling tasks, searching a local optimal solution to form an order queue after sequence change, then taking the first task of the order queue after sequence change as a current scheduling task, and performing splitting according to the type of the current scheduling task to obtain an action sequence of the mechanical arm; and inserting the task action sequence into a queue of actions to be executed to form an action queue according to the starting and ending time of the action sequence execution. In addition, the invention also provides electronic equipment and a robot based on the dynamic scheduling method.

The invention relates to a storage dynamic scheduling method and system based on digital twinning and deep learning and a storage medium, and the method comprises the following steps: constructing a digital twinning space synchronized with a physical storage through collecting operation tasks and storage global operation state data in real time; analyzing the data by using a pre-trained anomaly detection model, generating graded early warning information, inputting the graded early warning information into a dynamic scheduling model, and generating a scheduling instruction set subjected to simulation verification; and judging a high-risk area by executing feedback data of a scheduling instruction set, activating a virtual redundant device model in the digital twin space when a high risk is detected, generating a starting instruction, and scheduling a corresponding physical redundant device to access an operation system.

The invention relates to the technical field of high-performance computing and artificial intelligence crossing, and discloses a method for automatically generating a simulation computing task on a supercomputing platform, which comprises the following steps of: receiving a task demand of a user; performing semantic understanding and parameter extraction on the input by using a parameter dynamic analysis engine to generate a simulation parameter table; based on the simulation parameter table and the input format specification of the target simulation software, an executable parallel computing task script is automatically generated through a self-adaptive script generator; according to the real-time resource state and the task requirement of the supercomputing platform, a dynamic scheduling algorithm is adopted to distribute the task script to the optimal computing node; in the task execution process, the running state is monitored in real time, and when abnormity is detected, a fault-tolerant and self-repairing mechanism is triggered; by automatically generating the task script, the user parameter configuration time is saved, and the overall scientific research efficiency is improved by more than 30%; natural language input and zero code configuration are supported, and common scientific researchers can quickly master the method.

This application relates to the technical field of natural language processing, and provides a distributed pipeline-parallel LLM fine-tuning method for heterogeneous GPUs. A plurality of LoRA models are fine-tuned simultaneously based on a multi-job fine-tuning system; each LoRA model is partitioned into a plurality of parts distributed on a corresponding number of GPUs, and the GPUs are sorted. A job configuration module generates a plurality of jobs according to a user request, and divides each job into a plurality of training batches; a dynamic job scheduler generates a scheduling scheme based on a training batch sequence of each job and a dynamic scheduling strategy; and the scheduling scheme is sent to a multi-job training module on each corresponding GPU according to a positive sequence of the GPUs, to train all the LoRA models.

The present invention provides a project resource allocation method and an allocation system based on dynamic scheduling, belonging to the field of intelligent technologies. The allocation method includes: inputting project resources, a task set of a project, and a required resource set corresponding to the task set into a dynamic scheduling model of the project; according to the dynamic scheduling model, determining a set of feasible allocation schemes for the project, where each feasible allocation scheme can enable the available resources of the project resources at any time sequence to meet the required resources corresponding to multiple second tasks in a processing state at this time sequence; and determining the feasible allocation scheme with the minimum completion time in the set of feasible allocation schemes as the target allocation scheme. The method designed by the present invention can, in an environment of resource shortage competition, form different critical paths through different resource allocation schemes, compare all the formed critical paths, and find a path with the earliest completion time as the optimal path, so as to achieve the purpose of making full use of resources and shortening the project cycle.

The invention discloses a steel production line dynamic scheduling system based on data integration analysis, which belongs to the technical field of steel production line plan scheduling, and is characterized in that a dynamic knowledge graph driven by integrated multiple models is constructed, and scattered and heterogeneous data acquired in real time is converged into a data stream; an initial scheduling scheme is produced based on original order data and production line resource data in the'data stream ', simulation deduction is performed based on digital twin driving, an optimized planned production list is obtained, production is executed with the planned production list as a reference, real-time production data and change order data in the real-time'data stream' are received, and the real-time production data and the change order data in the real-time'data stream 'are obtained. Process deviation and order change in the actual production process are dynamically responded, correlation root cause analysis and scheduling decision making are carried out on multiple disturbance factors based on model and map cooperation, dynamic scheduling based on priority and resource constraint is achieved, it is ensured that the actual production condition is consistent with a production plan after scheduling optimization, and production flexibility and adaptability are improved.

The invention relates to an efficient execution optimization method for an embedded operating system based on an MIPS instruction set, and belongs to the field of computer operating systems. The method comprises the following measures: improving the parallelism of instructions: maximizing the parallel execution of the instructions through the intelligent analysis and dynamic scheduling strategy of an operating system; enhancing branch prediction capability: based on process and thread affinity, adding an intelligent scheduling strategy, cache consistency optimization and affinity strategy dynamic adjustment; the context switching efficiency is optimized; and through an improved context switching mechanism, the overhead during task switching of the operating system is reduced. According to the method, the instruction parallelism is enhanced, so that the processor can more efficiently execute a plurality of instructions; the branch prediction accuracy is improved, and the performance loss caused by wrong prediction is reduced; by improving the context switching efficiency, the overhead during task switching is reduced. These measures collectively ensure smoother computing and more efficient management.

The invention provides a dynamic scheduling method and system for a redundant executor based on a foreground theory and a Q learning evolutionary game method. The method comprises the following steps: constructing a strategy space, a profit structure and strategy evolution dynamics of the executor and a reward and punishment mechanism of an enhanced analyzer-regulator; according to the dynamic scheduling method, an executor simulates interaction under a tripartite game framework, iterative loop is carried out, and a strategy space of a mimicry defense system is output as a redundant executor after the loop is ended, the simulation interaction process comprises the following steps: the enhanced analyzer-regulator monitors and obtains objective income of each execution body in real time, calculates subjective perception utility of each execution body according to an income structure of the execution body, and then adjusts the subjective perception utility of the mimicry defense system according to a reward and punishment mechanism of the enhanced analyzer-regulator. And each execution body dynamically updates the strategy space of each execution body according to the adjusted subjective perception utility and strategy evolution.

The invention discloses a dynamic scheduling strategy generation method and device based on a large language model, which utilizes the natural language processing capability of the large language model to generate and perfect a scheduling heuristic algorithm. Through an initialization method based on a big language model, a knowledge extraction method based on the big language model and a sorting selection method based on the big language model, complex scheduling requirements described by natural languages can be met, so that rapid adjustment can be performed without a large amount of manual intervention to adapt to constantly changing parameters; therefore, the flexibility of the scheduling algorithm is enhanced, the calculation burden of the genetic programming method is reduced to the greatest extent, and the dynamic scheduling effect is finally improved.

In some implementations, a scheduling platform may receive task information regarding a set of tasks for execution using a set of computing resources, wherein the task information includes, for the set of tasks, at least one of: a run time parameter, a priority parameter, or a success rate parameter. The scheduling platform may communicate with a computing resource management device to obtain first computing resource information regarding the set of computing resources. The scheduling platform may generate a first assignment of the set of tasks to the set of computing resources. The scheduling platform may transmit assignment information identifying the first assignment. The scheduling platform may receive second computing resource information. The scheduling platform may generate a second assignment of the set of tasks to the set of computing resources. The scheduling platform may transmit second assignment information identifying the second assignment.

The invention relates to a Redis-based asynchronous task security scheduling method and system, and the method comprises the steps: creating an asynchronous task, and storing the asynchronous task to a task queue for scheduling; scheduling the asynchronous tasks through atomic operation executed by the script, preferentially acquiring the overtime asynchronous tasks from the task execution set, or popping up the asynchronous tasks to be executed from the task queue, and recording the acquired asynchronous tasks to the task execution set to track a scheduling state; prolonging the maximum execution time of the asynchronous task by updating the expiration time of the asynchronous task in the task execution set; and after the asynchronous task is completed or failed, deleting the asynchronous task from the task execution set. Compared with the prior art, the scheme has the advantages that the task scheduling stability in a high-concurrency scene is remarkably improved, the task loss risk is reduced, non-blocking access and dynamic scheduling are supported, the scheme is suitable for high-reliability systems such as financial systems and industrial control systems, the system performance is optimized, and the implementation complexity is simplified.

The invention provides a workflow calculation task high-dynamic scheduling method and system. The method comprises the steps that a task scheduling event is monitored; under the condition that the task scheduling event is triggered, a simulation optimization task queue to be scheduled is obtained according to the triggering condition of the triggered task scheduling event, and the simulation optimization task queue comprises a plurality of workflows and the task requirement of each workflow; obtaining a plurality of currently available servers and resource characteristics of each server; according to the task requirement of each workflow and the resource characteristics of each server, a scheduling optimization problem is constructed, an optimization algorithm is selected, and the scheduling optimization problem comprises a constraint condition and an iteration target; according to the scheduling optimization problem, obtaining a coding mode and a decoding mode for an optimization algorithm; and determining a scheduling mode of the simulation optimization task queue to be scheduled by adopting an optimization algorithm, a coding mode and a decoding mode by taking a constraint condition and an iteration target of the scheduling optimization problem as iteration conditions.

The invention discloses a multi-task allocation and scheduling method for a multi-core data stream processor. The method comprises the following steps: step 1, building a system architecture; 2, task collection and sparseness calculation; step 3, dynamic priority calculation; step 4, implementing a load prediction mechanism; step 5, task scheduling and distribution; and step 6, evaluating and optimizing the system. By introducing a dynamic priority adjustment mechanism, the system can respond to changes of task characteristics and kernel states in real time, the scheduling efficiency is remarkably improved, and the dynamic adaptability ensures that key tasks can preferentially obtain processing resources, so that the processing delay is reduced; through a load prediction mechanism, the load among the cores is effectively balanced, and the situation that some cores are overloaded and other cores are idle is avoided, so that the overall resource utilization rate of the system is improved, dynamic scheduling and load optimization are integrated, the system can keep efficient operation in a complex data stream processing scene, and the throughput and real-time response capability of data processing are improved.

The invention discloses a method and system for generating a software development scheme, and belongs to the technical field of software development, and the method comprises the steps: collecting system resource occupation data and thread operation data during multi-user concurrent operation, segmenting the system resource occupation data into resource fragments according to time dimensions, positioning resource competition points, and generating conflict mark data; constructing a thread execution graph based on the conflict mark data, identifying mutually exclusive resource points, constructing a double-layer distributed queue, setting a resource state monitor to dynamically allocate thread priorities, and generating scheduling processing codes; monitoring the running state of the scheduling processing code, and splitting the thread into a plurality of sub-threads for parallel execution when detecting that the thread is blocked; and collecting an execution result, recording the execution path with the shortest response time and the highest resource utilization rate as a scheduling template, and generating a software development scheme. According to the method, the utilization efficiency of system resources is improved through dynamic scheduling and parallel processing, and the thread blocking probability is reduced.

This invention discloses a robust optimization-based dynamic scheduling method for remanufacturing workshops, comprising a pre-scheduling stage and a dynamic scheduling stage. In the pre-scheduling stage, a discrete scenario set is used to describe the arrival and processing times of jobs in the remanufacturing workshop. A robust optimization objective function is established based on minimizing the completion time of the pre-scheduling scheme and the difference in completion time under different scenarios. A biogeographical optimization algorithm is used to output the pre-scheduling scheme. In the dynamic scheduling stage, if a disturbance event occurs, the pre-scheduling scheme is used as the input scheme, and a job right-shifting and rescheduling strategy is executed to obtain the dynamic scheduling scheme. If the dynamic scheduling scheme is equal to or does not dominate the input scheme, the dynamic scheduling scheme is output; otherwise, a dynamic scheduling objective function is established by minimizing efficiency and robustness indices, and a biogeographical optimization algorithm is used to output the dynamic scheduling scheme. This invention considers the problems of multiple uncertainties and the impact of disturbance events, resulting in a better scheduling scheme.

The invention discloses an account checking task scheduling management method and system based on a dynamic priority, and relates to the technical field of task scheduling management, and the method comprises the steps: aggregating a plurality of original account checking tasks according to a task dependence topology; setting a plurality of basic priorities of the plurality of original account checking tasks; outputting a plurality of dynamic priorities; performing hierarchical correction connection of the hierarchical task tree to obtain a target correction task tree; segmenting the target correction task tree; analyzing time sequence scheduling of the Q task execution containers, and outputting an elastic rotation container scheduling chain; and performing dynamic scheduling and account checking. The technical problems that in the prior art, account checking task scheduling lacks dynamic priority consideration, task execution efficiency is low, resource allocation is unreasonable, account checking work consumes long time, and resource waste is serious are solved, dynamic and efficient scheduling of the account checking tasks based on the priority is achieved, the task execution efficiency is improved, and user experience is improved. And the technical effect of resource allocation is optimized.

The application discloses a kind of based on performance perception's heterogeneous computing task dynamic scheduling system and method.The system is deployed in digital ecological platform complex system, through the collaborative work of performance knowledge base, real-time performance monitor, scheduling strategy re-evaluation intelligent contract and task allocator, the efficient dynamic scheduling of computing task is realized;Through computing task characteristic fingerprint analysis and AMCU adaptability score model, the computing task with matrix operation operation dominant and higher adaptability score is preferentially scheduled to AMCU computing node, and the efficiency and reliability of heterogeneous computing task scheduling are significantly improved.The application realizes the global performance optimization and trusted scheduling of various heterogeneous computing resources in platform by real-time sensing and dynamic scheduling.

The application provides a method for automatically decomposing and dynamically scheduling quantum tasks under a quantum super-fusion architecture. The method comprises: obtaining a task script corresponding to a quantum task, and performing parsing processing on the task script to obtain circuit information, the circuit information including quantum bit quantity and circuit gate depth information; generating a plurality of sub-task scripts and a sub-task execution chain according to the circuit information and the task script, the plurality of sub-task scripts including a compiled task script, a simulated task script and a post-processing task script; for any one sub-task script, determining a job scheduling script corresponding to the sub-task script according to the circuit information; associating and binding a plurality of job scheduling scripts according to the sub-task execution chain to generate a target execution script, and submitting the target execution script through a preset native interface, thereby improving the quantum task execution efficiency and resource utilization.

The invention discloses a thread pool dynamic scheduling method and system, and the method comprises the steps: dividing the time of a time period into a plurality of time blocks, merging the time blocks with the average time consumption exceeding a set first threshold value, and taking the merged time blocks as an estimated slow time period of a task A in a next time period; in response to an execution starting instruction of the task A, judging whether the current time is in an estimated slow time period of the task, if so, calculating comprehensive average time consumption based on average time consumption of the estimated slow time period and total time consumption of the task in the estimated slow time period, comparing the comprehensive average time consumption with a set second threshold value, and determining a thread pool allocated to the task; and if not, calculating the latest execution time consumption based on the average value of the latest set m times of execution time consumption of the task, comparing the latest execution time consumption with a set second threshold value, and determining the thread pool allocated to the task. According to the method, the thread pool distributed to the task A can be adjusted in advance, the system throughput can be improved, and the service stability is improved.

The invention discloses a fluid pulsation accelerator architecture for reinforcement learning of a large language model, and belongs to the technical field of artificial intelligence hardware acceleration. The system aims at solving the problems that when a current universal processor executes an RLHF working load, an instruction driving normal form is not matched with data flow calculation, fixed parallel granularity cannot adapt to a dynamic load, and the efficiency is low due to the fact that a framework does not sense data statistical characteristics. The core of the system is a liquid systolic array calculation fabric capable of being dynamically reconstructed, execution is triggered through data flow, and kernel boundaries and instruction overhead are eliminated. According to the system, a global lookup table subsystem is fused, and nonlinear calculation is optimized through merging query and parallel lookup by using data distribution prior; and a dynamic scheduling unit is configured, and elastic parallelism is realized by adopting a resource allocation algorithm supporting work stealing. According to the method, the throughput and the energy efficiency of RLHF training and a large language model reasoning stage can be remarkably improved, and a new design normal form is provided for a next-generation AI special computing architecture.