77 results about "Limited resources" patented technology

The invention discloses a self-adaptive micro-service resource management method and device, and the method is applied to a resource-limited micro-service system, and comprises the steps: receiving a task, carrying out the modeling of the time sequence data and spatial dependency of a micro-service based on a space-time diagram neural network, and predicting the resource demands of a future time step, including a CPU, a memory and a network bandwidth; classifying the micro-services into resource redundancy, resource insufficiency and resource self-sufficiency according to the predicted resource demand, the current resource allocation and the resource guarantee threshold; according to the resource state and the business importance, the dynamic priority of the micro-services is calculated, the micro-services with redundant resources are sorted in an ascending order, and the micro-services with insufficient resources are sorted in a descending order; and recovering redundant resources from the low-priority micro-services, and allocating the redundant resources to the high-priority micro-services until available resources are exhausted or resource requirements are met, so that the resource utilization rate is optimized, and a service level protocol is guaranteed. According to the method, when resources are limited, the adaptability of the micro-service in different load environments can be improved.

The invention provides a construction progress plan intelligent generation method based on a Monte Carlo tree search algorithm, and particularly relates to a construction progress plan intelligent generation and optimization technology based on a Monte Carlo tree search (MCTS) algorithm. According to the technology, a traditional MCTS algorithm is improved, and rapid generation, real-time dynamic adjustment and resource optimization configuration of a construction progress plan are achieved. Specifically, measures such as dynamic adjustment of an exploration coefficient, nonlinear adjustment of a prior probability and introduction of depth regularization are adopted to improve a traditional upper limit confidence coefficient (UCT) formula. And meanwhile, the total construction period is minimized by combining a dependency relationship between construction steps and constraint conditions of resource finiteness and utilizing a reward function with a negative value of the construction period as a core. Finally, the technology can visually output a construction progress plan in a Gantt chart form, remarkably improves the construction efficiency, reduces the dependence on high-end construction management talents, has a good real-time response capability, and has a wide engineering application prospect.

The invention relates to the technical field of computing resource scheduling, and discloses a computing resource allocation method, device and equipment and a medium, and the method comprises the steps: constructing a global data set in a long-tail scene; a high-dimensional sparse matrix is constructed based on the global data set, the high-dimensional sparse matrix comprises a plurality of data dimensions, feature dimension reduction is performed based on the data dimensions to obtain a feature matrix after dimension reduction, and a feature sub-matrix is generated based on the feature matrix after dimension reduction; evaluating the feature sub-matrix, and determining the priority of the feature sub-matrix in the current scene based on an evaluation result; allocation of computing resources is adjusted based on the priority. According to the method, the vehicle-mounted system can preferentially process high-priority long-tail scenes under the condition that resources are limited, the processing efficiency is high, resource waste and processing delay are avoided, and the efficiency of the whole data processing chain can be optimized, so that the vehicle-mounted system has certain flexibility and coping capacity when facing complex and rare long-tail scenes.

The invention discloses a local reasoning acceleration system and method for a large-scale MoE model under limited resources, and the method comprises the steps: carrying out the MoE model performance modeling on large language model reasoning equipment, and obtaining a module calculation performance model; executing a sliding window type greedy prediction-scheduling strategy: performing expert module routing prediction by utilizing a gating network in the MoE model to obtain activated experts of each layer and an input token of each expert, and dynamically determining the number of layers of the MoE model contained in a single scheduling window by utilizing the module calculation performance model in combination with a minimum sliding window algorithm; and according to the predicted expert module route, performing optimal execution strategy search and MoE model layer scheduling on the activated expert module in the window with the single scheduling window length. According to the method, efficient and stable local large model reasoning deployment is realized through an expert activation prediction mechanism without training cost.

The invention discloses a GPU memory optimization method and system for DNN training under limited resources, and the method comprises the steps: S1, carrying out the traversal selection of a current operator op from an operator sequence O of an input DNN model, and skipping to S2 if the traversal is not completed; otherwise, ending and exiting; s2, executing the current operator op, and recording runtime information of a tensor for executing the current operator op; s3, if the memory usage amount of the GPU is larger than or equal to the preset threshold value, tensors in part of the GPU memory are selected, and the selected tensors are directly expelled or exchanged to the CPU memory to release the GPU memory; and skipping to the step S1. The invention aims to realize GPU memory optimization oriented to DNN training under limited resources, so that the same GPU can support a larger DNN model and a larger training batch, meanwhile, the influence on the training efficiency is reduced, and the overhead during operation is reduced.

According to the maneuvering edge application scene-oriented computing power resource assessment method and system provided by the invention, the task scene description model based on the task sequential relationship is constructed, and the complex relationship among task requirements, subtasks and resources is effectively processed, so that the task resources can be allocated based on a deep reinforcement learning algorithm, and the task resource allocation efficiency is improved. Therefore, the optimal allocation of task resources is realized under the condition that the resources are limited, and the task execution efficiency and the resource utilization rate are improved. Compared with a traditional method, the method is more suitable for dynamically changing task scenes, the resource utilization rate and the task execution efficiency are effectively improved, and system delay and energy consumption are reduced.

The invention discloses a method, a system and a related device for starting an application, which are used for improving the starting speed of the application under the condition of limited resources and meeting the requirement of a payment task. The method comprises the steps of determining a starting trigger event of an application; a pre-configured configuration file is obtained according to the start trigger event, the configuration file comprises a target frequency corresponding to the start trigger event, and the target frequency is greater than a default operation frequency; and starting the application according to the target frequency response to the starting trigger event.

The invention discloses an off-line reasoning task reasoning acceleration method and system for a large language model under limited resources. The system comprises a model construction module, a dynamic batch processing recombination module, a dynamic memory management module and an unloading strategy adjustment module. Based on a lexGen large language model offline inference system, inference tasks are subjected to dynamic batch processing recombination, dynamic memory management design is combined, the inference tasks are managed in an iteration level in a fine-grained mode, memory occupied by the completed inference tasks is released in real time, and uncompleted inference tasks are dynamically arranged. The waste of reasoning resources in the reasoning process is reduced, and the throughput of the off-line reasoning system based on the unloading technology is improved. According to the method, the hardware idle memory is efficiently utilized, and the throughput of off-line reasoning is improved; the influence of unloading on the reasoning process is reduced; and a tensor part in equipment with lower bandwidth is transferred into the part of the memory by utilizing the idle memory, so that the I / O overhead of transmission during reasoning is reduced.

The invention discloses an intelligent transport capacity scheduling method and system, and relates to the technical field of transport capacity scheduling. An intelligent transport capacity dispatching system comprises a transport capacity information integration module and a transport capacity dispatching module. According to the method, the priority index of each order to be distributed is calculated, sorting is performed according to the priorities of all the orders, priority sorting not only considers the emergency degree of the orders, but also comprehensively evaluates factors such as cargo attributes, shipping space capacity and green transportation, and a comprehensive priority index is formed through weighted calculation, so that the accuracy of scheduling decision making is improved; through the group game rule, the transportation party optimizes the bidding strategy according to the priority of the order and the scheduling demand, the flexibility and diversity of scheduling are enhanced, and the transportation party adjusts the bidding weight according to different strategies in the game process, so that the parties can dynamically adjust the scheduling scheme under the condition of limited resources.

The invention belongs to the technical field of mobile group computing. The invention provides a task joint unloading and migration method for air-ground collaborative mobile edge computing. According to the embodiment of the invention, the scenes of resource heterogeneous, resource limited, communication range limited and network dynamic fluctuation under the calculation of the edge calculation of the unmanned aerial vehicle-vehicle networking are considered; the method comprises the following steps: modeling a joint unloading and migration problem of a DAG task, and expressing the problem as a user DAG task average delay minimization problem; according to the method, different characteristics of the UAV and the vehicle are considered, and the task migration triggering conditions are different; according to the method, the situation that the same DAG task may be migrated for multiple times is considered, and the shortest transmission path of task migration, task uploading and calculation result downloading is considered.

A computer-implemented system and a method for automatic electronic order creation are disclosed. The system and method may be configured to: receive a digitized document comprising embedded metadata related to contents of the digitized document; analyze the embedded metadata to extract a boundary parameter; generate a unique reference identifier associated with the digitized document and the boundary parameter; receive a first input data setting one or more target measurements for one or more predefined periods of time, wherein a sum of the one or more target measurements is less than or equal to the boundary parameter; receive a second input data setting configuration data for achieving the one or more target measurements; and generate an electronic order based on the configuration data.

The invention discloses an earthwork stockpiling and deploying method based on a return demand, and the method comprises the steps: introducing decision variables, and representing the types of stockpiling fields, the types of earthwork stockpiling in the stockpiling fields, and the amounts of various earthwork coming and going between each excavation field and each stockpiling field through the decision variables; based on the type of the stockpiling yard, the excavation and return demand, the capacity of the stockpiling yard and the actual value range of each decision variable, establishing association constraints about the decision variables; determining a first measurement target and a second measurement target by minimizing the total transportation distance and minimizing the amount of earth and stone excavated and returned from the mixed stockpiling yard, introducing a trade-off coefficient, and constructing a mixed integer linear programming model in combination with association constraints; and substituting the numerical value to solve the mixed integer linear programming model, and determining an optimized earth-rock stockpiling allocation scheme. According to the method, a classified stockpiling strategy is set, the transportation distance and the backhaul excavation volume are comprehensively considered under the constraint condition, limited resources are reasonably utilized, and a solution with higher efficiency and lower cost is provided.

The invention discloses an optimized deployment method for network mode coexistence in a multi-mode network environment, and belongs to the field of multi-mode intelligent networks. The method comprises the following steps: modeling a deployment problem of a network mode as a 0-1 integer programming problem, and comprehensively considering key constraints such as multi-mode network element resources, link resources and flow forwarding delay so as to ensure user service quality; meanwhile, through joint routing optimization, the connectivity of each network mode is ensured; for the non-convexity of an original problem, the problem is converted into an equivalent 0-1 integer linear programming problem, and a network modal deployment and routing algorithm based on random rounding is designed; according to the algorithm, efficient deployment of the network modals can be realized under the condition of limited resources, and the network performance is ensured. According to the method provided by the invention, the link load is effectively reduced, and the utilization rate of network resources is improved on the premise of ensuring the service quality.

The invention provides a log processing method and device, and the method comprises the steps: responding to a target application program to generate log data, and judging whether the resource quantity required by the current running service of the target application program reaches a first preset threshold value or not; if yes, storing the log data in a memory; if not, storing the log data in a memory, storing the log data in the memory in a disk, and deleting the log data in the memory; judging whether the resource quantity required by the current running service of the target application program during running is smaller than a second preset threshold value or not; if yes, storing the log data currently stored in the memory in a disk; and if not, maintaining the log data currently stored in the memory, and executing the step of generating the log data in response to the target application program and judging whether the resource quantity required when the service currently operated by the target application program is operated reaches a first preset threshold value or not. Under the condition that resources are limited, normal operation of services is ensured, and complete storage of log data is ensured.

The invention provides a project scheduling and resource configuration integrated optimization method considering time-resource balance, and the method comprises the steps: obtaining the information of a to-be-optimized project, and proposing a basic hypothesis according to the information; according to the basic hypothesis, constructing a dual-objective function model with the goal of minimizing the project construction period and the total cost; and solving the dual-objective function model by adopting a self-adaptive large neighborhood search algorithm based on a hypervolume contribution rate index to generate an optimal project scheduling plan and an optimal resource configuration plan. According to the method, on the premise of ensuring that the project is completed on schedule, resource consumption and cost expenditure are effectively reduced, the construction period is shortened, and the execution efficiency and the resource utilization rate are improved, so that the project scheduling and resource configuration effects are improved, and the method is particularly suitable for a complex project management scene pursuing cost benefit maximization under the condition of limited resources.

The application discloses a large language model offline reasoning task reasoning acceleration method and system under limited resources, a model construction module, a dynamic batch processing reorganization module, a dynamic memory management module and an unloading strategy adjustment module; based on the lexGen large language model offline reasoning system, dynamic batch processing reorganization is performed on the reasoning task, dynamic memory management design is combined, reasoning tasks are managed in an iterative level fine-grained manner, occupied memory of completed reasoning tasks is released in real time, and dynamic arrangement is performed on uncompleted reasoning tasks, so that the waste of reasoning resources in the reasoning process is reduced, and the offline reasoning system throughput based on the unloading technology is improved. The application efficiently utilizes the idle memory of hardware, improves the throughput of offline reasoning, reduces the influence of unloading on the reasoning process, and transfers part of tensors in a device with low bandwidth to the idle memory, so that the I / O overhead during transmission is reduced.

The invention discloses a server-free edge function reliability deployment method and system based on reinforcement learning, and the method is oriented to server-free edge calculation, takes the minimization of long-term network cost as a target, and systematically solves the problems of high cold start cost, limited resources, difficult reliability guarantee and the like. A system model including user movement, channel fading, function deployment and a backup mechanism is constructed, four types of costs of cold start, maintenance, transmission and punishment are accurately described and converted into a Markov decision process, and a state-action-award framework is utilized to realize dynamic decision. The PPO algorithm fusing the greedy strategy to assist the action selection is further provided, and a function deployment scheme meeting the reliability constraint is efficiently generated under the condition that resources are limited. According to the method, through fusion of reinforcement learning and a greedy strategy, the problems of feasibility, reliability and cost of function deployment in server-free edge computing are solved, and the method has the capabilities of efficient deployment, adaptive scheduling and long-term performance optimization.

The invention relates to a sketch general configuration method and system based on a data processing unit, and belongs to the technical field of data processing. The method comprises the following steps: (1) identifying resource parameters; (2) selecting a process model; (3) resource constraint input; (4) optimizing target configuration; and (5) parameter configuration recommendation. The invention aims to optimize the communication efficiency of the DPU and the host, reasonably allocate hardware resources, and realize flexible process model selection and efficient state operation configuration, thereby achieving the goals of high accuracy and high throughput of network measurement under limited resources.

The embodiment of the invention discloses an operator adaptive scheduling method and device in an NPU resource shortage environment, and is applied to the technical field of computers, and the method comprises the steps: carrying out the initial priority assignment of each operator; the operators with the in-degree being 0 are added into a ready queue according to the initial priority; presetting a priority adjustment period, if the previous operator is executed in the adjustment period, changing the operator in-degree in the branch into 0, and adding into a ready queue; recalculating the priority of each operator by using a weight adjustment algorithm, and performing adaptive adjustment on the operator ready queue based on an executable degree evaluation model according to the recalculated priority of each operator; if the adjustment period is reached, adding the operator which is being executed into the operator ready queue again, and carrying out self-adaptive adjustment on the operator ready queue; and repeating the above process until the operator ready queue is empty. In this way, ordered execution of various operators under limited resources can be guaranteed, and the resource utilization efficiency is improved.

The present application discloses a large language model training method, system, device and storage medium. The method divides the student model and the large language teacher model and deploys them in a central controller with sufficient resources and on a client with limited resources respectively. This ensures that the local data of the client is only used to train the local network structure, avoiding the risk of local data leakage. In addition, the method trains the network structure with heavy computational load through the central controller, while the client only trains the network structure with light computational load. This adapts to the resource limitations of the client device, allowing the resource-limited client to participate in the model training. Moreover, the knowledge of the teacher model is transferred to the student model through knowledge distillation, ensuring that the data is not leaked and significantly reducing the dependence on the central controller.

The invention discloses a process migration control method and device for flexible assembly task switching, and relates to the technical field of flexible assembly. The method comprises the following steps: acquiring original total data in a first process; taking the minimum sample loss function as an optimization target, and obtaining a distillation candidate subset from the original total data; the sample loss function is used for measuring the retention degree of the distillation candidate subset on the behavior distribution of the original total data; calculating a comprehensive score of each action in the distillation candidate subset through a distribution consistency score and an action border-crossing punishment, filtering out sample data with relatively large comprehensive scores of the actions in combination with a gating mechanism, and storing the sample data meeting the gating mechanism into a small-capacity buffer area; and finally, training the initial strategy network according to the sample data of the small-capacity buffer area and the observation data of the second process to obtain a new strategy network. The method can give consideration to both online speed and stability under the condition of limited resources.

The invention discloses a multi-agent co-localization method based on event triggering under limited resources, which comprises the following steps: executing a distributed co-localization algorithm based on relative measurement by utilizing a kinetic equation and an observation equation of a multi-agent system, and updating an estimation state and a covariance matrix of each agent; traversing k intelligent agents which are neighbor to the intelligent agent i, and calculating a relative covariance matrix between the state estimation value of the intelligent agent i and the state estimation value of the neighbor; when it is determined that a triggering condition is met based on the relative covariance and a set event triggering condition, communication is started between the intelligent agents, after information fusion operation of covariance crossing is carried out, next round of iteration updating operation is carried out on the estimation state and the covariance matrix of each intelligent agent, and otherwise, the next round of iteration updating operation is directly carried out; according to the invention, under the constraint of limited resources, the resource consumption can be effectively reduced, and the multi-agent cooperative positioning precision can be maintained.

The invention relates to the technical field of cloud computing resource management, in particular to an elastic task scheduling method and system based on a large language model, electronic equipment and a medium. The method comprises the steps of firstly obtaining a task mirror image of a user and analyzing a task code through a large language model to generate an initial resource demand feature; performing code embedding vector similarity calculation on the initial resource demand characteristics and historical task codes to generate elastic resource demand distribution characteristics; performing scheduling matching with cluster real-time available resources according to the characteristics; and if the matching fails or the operation fails, correcting the entry point of the task mirror image and the related source code or the configuration file through the large language model so as to re-execute the task scheduling matching. In this way, the technical problem that the resource demand and supply dynamic matching is difficult when an existing container cluster scheduling technology faces a limited resource environment is solved, and the task success rate and the overall cluster resource utilization efficiency are improved.

The invention discloses a data block relevance data storage method for improving network security availability, which comprises the following steps of: firstly, analyzing a historical IO (Input / Output) request, discovering a data block set which is frequently collaboratively accessed, forming a high-relevance data combination, and calculating an IO delay revenue factor, a space cost factor and collaboration access intensity, so as to form a high-relevance data combination; comprehensively evaluating the migration value of each frequent item set data; by constructing a conflict graph model, analyzing a resource competition relationship among different frequent item set data, correcting an expected utility value by using an interference sensing weight, and identifying and avoiding potential resource conflicts; based on the corrected utility value, particle swarm optimization processing is adopted to optimize and select frequent item set data with the highest utility value under the constraint of a high-speed storage space, the frequent item set data are sequentially incorporated into a migration list, and it is ensured that the overall revenue is maximized under limited resources; and according to the priority order of the candidate list, the data block migration is executed by adopting a batch migration strategy, and the storage performance is improved while the system stability is ensured.

This invention provides a game-based CPPS optimal resource allocation method for FDI attacks. The method first obtains the system state estimation results under normal conditions, then simulates the state estimation results after an FDI attack. The offset between the two is used as the participant's payoff. The corresponding payoff amount, the resources required by the attacker to launch the attack, and the resources required by the defender to provide defense are input into the game model. The Nash equilibrium is calculated to obtain the corresponding strategy and expected payoff. Finally, resources are allocated to multiple nodes under limited resources. This allocation method improves defense efficiency while ensuring system security from the defender's perspective, providing feasible suggestions and references for defenders to configure defensive measures. Furthermore, the method can adapt to multiple state estimation methods and different defense methods, showing high flexibility and scalability.

The embodiment of the invention discloses a task processing method and related equipment thereof, tasks can be processed through a solver carrying a symbol expression, the solver can be applied to equipment with limited resources, and various tasks can be completed in more scenes. The method comprises the following steps: when a to-be-processed task needs to be completed, abstracting the task into a mixed integer programming equation, and extracting a plurality of variables of the mixed integer programming equation and a plurality of features of each variable; then, a symbol expression used for indicating the relation among the multiple features of the variables can be obtained, the multiple features of each variable are calculated through the symbol expression, and evaluation values of the multiple variables are obtained; then, in the multiple variables, the variable with the highest evaluation value can be branched, and a branching result is obtained; and finally, the mixed integer programming equation can be solved based on the branch result, so that a solution of the mixed integer programming equation is obtained, and the solution can be used as a task processing result.

The present invention provides a fusion scheduling method and device based on graph data lineage in a full-load scenario, the method comprising: when a source table triggers an update, querying the sub-tables of the source table level by level, treating the multi-level sub-tables triggered this time as the first task batch; obtaining at least one first task batch; performing a first rule check on the sub-tables in the same first task batch before scheduling execution, and the sub-tables that pass the first rule check will be scheduled for execution; for tasks in the pre-execution queue, according to the configuration rules, the N first task batches in the queue are merged to obtain a second task batch; after the task batches in the queue to be executed are completed, the second task batch is pulled for execution. The fusion scheduling method and device based on graph data lineage in a full-load scenario provided by the present invention optimize the task scheduling execution strategy according to lineage information, reduce repeated execution, and ensure the smooth operation of more tasks under limited resources.

The invention discloses a resource allocation method and system for a VR cloud game in a meta universe in an edge environment. The method comprises the following steps: constructing a task execution scene of a VR cloud game in a meta universe in an edge environment, and detailedly describing a task process of the VR cloud game; determining resources needing to be allocated by the VR cloud game scene in the meta universe in the edge environment; determining a heterogeneous user set, wherein the resolutions of the user VR devices in the scene are different; constructing a user experience quality function in the VR cloud game scene in the meta universe in the edge environment; an optimization problem is constructed, constraint conditions are determined, and the average user experience quality is maximized under the condition of limited resources; the number of decision variables in the optimization problem is equal to the number of resources, and due to mutual coupling of the decision variables, an alternating optimization method is adopted for iterative optimization, and an optimal resource allocation scheme is solved; in the alternative optimization process, it is proved that the optimization problem of each decision variable is a convex optimization problem, constraint conditions are inequality constraints, and a KKT condition is adopted for solving. Limited edge server resources are distributed in the static edge environment, so that the average experience quality of users in the system reaches the maximum value.

The present application discloses a method and apparatus for acquiring virtual resources, a storage medium, and a computer device. The method comprises: receiving a time-limited resource acquisition request, wherein the time-limited resource acquisition request includes a target character identifier, a target time-limited resource identifier, and a resource to be exchanged; determining the effective time limit of a target time-limited resource corresponding to the resource to be exchanged, allocating the corresponding target time-limited resource to a target character account, and deducting the resource to be exchanged from the target character account; when the effective time limit of the target time-limited resource is reached, changing the target time-limited resource allocated to the target character account to an exchange voucher matching the resource to be exchanged, wherein the exchange voucher is used to exchange virtual resources. The present application helps to improve the utilization rate of virtual resources within a game, as well as the utilization rate of a game server.

Various example embodiments relate to communication techniques, and more particularly, to apparatuses and methods for facilitating use of limited resources, particularly logical channel identifier (LCID) in a communication network. One aspect may provide an apparatus comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus to: determine that the at least one logical channel identification (LCID) index is to be used for communicating information for supporting communication within the communication network; determining the information and an encoding configuration for allowing the information to be encoded within a field of at least one logical channel identifier index; and encoding the information according to an encoding configuration for transmission in a message comprising at least one logical channel identifier field. It is understood that the principles in accordance with the described aspects may allow for more information to be carried from the UE to the network, for example, using a very limited Msg3 / MsgA transport block (TB) size. The additional information may include additional information related to connection establishment and / or UE state / capability at RRC request.