646 results about "On demand" patented technology

The invention provides an end-side multi-modal large model accelerated reasoning method and system, and the method comprises the steps: carrying out the two-stage screening and rearrangement of visual tokens based on the CLS attention and text-to-visual attention in a visual encoder and pre-filling stage, and constructing a sparse attention and sparse key value cache; in a decoding stage, an important neuron set is judged according to activation gating or historical statistics, only a corresponding feedforward network weight is pulled and calculated, missed weights are loaded on demand through asynchronous I / O, and hot neurons are maintained in a high-speed memory to utilize model sparsity, so that video memory / memory occupancy and calculation overhead are remarkably reduced on an end side; throughput and time delay performance are improved. According to the method, the internal memory and computing resources required by reasoning of the multi-modal large language model are reduced from two dimensions by utilizing the endogenous sparsity of the end-side large language model in input and the model, so that a higher reasoning speed is achieved by utilizing fewer resources on the premise of keeping the size of the model unchanged, and the performance of the whole system is improved.

The invention provides a script preservation method which can be applied to the technical field of artificial intelligence. The script preservation method comprises the following steps: carrying out semantic understanding on demand related files by utilizing a large model, extracting core elements, and generating a structured demand change list; analyzing to obtain an abstract syntax tree of an existing script code, and constructing a mapping relationship between the software automation test element and a script code block; dynamically comparing the structured demand change list with an existing script code structure tree based on a mapping relation between software automation test elements and script code blocks, identifying a script range influenced by demand change, and outputting a change influence matrix; and based on the change influence matrix, generating an automatic script code snippet needing to be updated, and realizing preservation of the automatic script. The invention further provides a script preservation device and equipment, a storage medium and a program product.

The invention relates to the technical field of intelligent terminals, in particular to an intelligent terminal power consumption optimization method based on active and passive identification, which comprises the following steps: passive identification acquires a first signal data set by means of a low-power-consumption wireless communication unit and based on a low-power-consumption wake-up mechanism, active identification period and passive cooperative on-demand wake-up are carried out, and cooperative data acquisition is carried out; carrying out standardized preprocessing on the data, and carrying out weighted fusion on passive electromagnetic and active characteristics according to scene priorities to generate a comprehensive scene vector; querying a predefined strategy mapping table based on the current scene category, obtaining a strategy identifier uniquely associated with the category, and loading a corresponding strategy execution parameter set from a strategy library according to the strategy identifier; analyzing the strategy parameter, generating a time sequence instruction, executing clock gating on the target hardware unit in sequence, and sending instructions by the power supply and the bus arbiter to complete power consumption state switching. The accuracy of scene judgment is improved through an active and passive fusion mechanism, and power consumption regulation and control deviation caused by scene misjudgment is effectively avoided.

The invention discloses a cache maintenance system of a heterogeneous computing system and electronic equipment, and relates to the technical field of data processing, a first directory controller manages the consistency state of a host end, and a second directory controller manages the consistency state of an equipment end. The consistency maintenance operation is firstly efficiently processed by the corresponding directory controller in the storage domain, cross-device frequent coordination communication is reduced, and when cross-domain data access occurs, the two directory controllers cooperate through an established communication coupling mechanism and execute global consistency maintenance as required. The method not only can adapt to a high-burst and high-parallel access mode of a device end, but also can avoid direct conflicts with a cache management mechanism of a host end, and realizes global consistency through distributed collaboration. The technical problem of performance bottleneck of cache consistency in the heterogeneous system is solved, and the technical effect that the overall data access efficiency of the heterogeneous system is improved while the correctness is maintained is achieved.

The invention discloses a live working robot on-line positioning accessory suitable for an overhead line, and relates to the technical field of overhead line maintenance. The structure of the live working robot on-line positioning accessory suitable for the overhead line in the prior art is optimized and improved; the three groups of wheel bodies are tightly attached to the cable to drive the hot-line work robot to move, and when the aging and damage areas of the cable need to be avoided, the shapes of the three groups of wheel bodies are timely adjusted to enable the three groups of wheel bodies to be timely separated from the cable one by one in sequence; the on-line positioning accessory for the hot-line work robot suitable for the overhead line has the technical effects that the on-line positioning accessory is convenient to use, the aging and damaged areas of the cable can be flexibly avoided as required during movement, and the trafficability and the applicability are relatively good.

The present application relates to the technical field of intelligent war game, in particular to an intelligent war game simulation system based on digital twin fusion large model, which comprises a digital twin modeling unit, a large model decision unit, a deduction execution unit and a dynamic interaction unit. The digital twin modeling unit adopts a dynamic precision adaptive modeling mechanism, which can realize dynamic switching of model precision on demand, balance the demand for detailed simulation and the consumption of computing resources, and realize sub-second dynamic synchronization of virtual models and war game scenes through a timing calibration algorithm. Through the multi-modal feature analysis module, spatial adversarial, timing decision and rule trigger features are fused, and the entity performance boundary and scene rule constraint are embedded in the war game rule verification module, which can generate compliant adversarial decision parameters in line with the logic of war game, solving the problem of insufficient model precision adaptation and decision constraint.

The invention discloses a development board Linux system security authentication multi-start method, which comprises the steps of 1, hardware trust root initialization and storage area division, 2, hardware trust root self-inspection and measurement, 3, starting item dynamic loading and multi-stage authentication, 4, security context establishment and system switching, and 5, multi-system isolation and collaboration. Step 6, performing security audit and exception handling; full-link authentication is realized by taking a hardware trust root as an anchor point, malicious mirror startup and inter-system data leakage are effectively resisted in combination with a storage and resource isolation mechanism, scenes such as on-demand dynamic switching, flexible adaptation debugging, upgrading and fault recovery of multiple systems can be realized, and the system reliability is improved by optimizing the lightweight design of an authentication algorithm and resource management. Limited computing power of the development board is adapted, the starting process and abnormal events can be completely recorded, and illegal operation can be traced conveniently.

The invention discloses a PCIe-based CPU, NPU and GPU interconnection cluster management method, and belongs to the technical field of interconnection cluster management. According to the method, an interconnection link is established through a PCIe switch, and bandwidth is dynamically allocated by means of the link to adapt to task requirements; multi-priority data transmission is supported, and bandwidth resources are distributed in order according to priorities; a multi-element operation mode is configured in advance, and physical and virtual computing power resources are distributed according to needs; and an equipment state monitoring mechanism is established, idle equipment enters a sleep mode, and meanwhile, all transmission links are combed and an optimal path is selected. According to the method, fine scheduling of cluster resources is realized, the bandwidth and computing power resource utilization rate is improved, the adaptation capability to different tasks is enhanced, the orderliness and stability of data transmission are guaranteed, the cluster operation energy consumption is reduced, and efficient and stable operation of the interconnected cluster is facilitated.

Embodiments described herein comprise an advanced Software as a Service (SaaS) platform addressing challenges including trust, governance, quality, data accuracy, and transparency, ensuring full visibility of enterprise data warehouse data pipelines. These embodiments seamlessly integrate with cloud-based and legacy data platforms, enhancing visibility through monitoring data quality, tracking metadata changes, and generating exceptions for review. Data quality management allows for custom script creation to validate pipelines and business rules. Cloud-native connectors facilitate integration with one or more cloud-based services streamlining data delivery processes. Personalized notifications aid administrators in responsive action, while data profiling capabilities identify quality issues and provide distribution insights. Actionable data intelligence for privacy, security, and governance is enabled through data classification. Automation features encompass data quality checks, metadata monitoring, profiling, and classification, orchestrated by a flexible scheduler. The architecture leverages a scalable serverless design with independent microservices, dynamically scaling based on demand.

The invention discloses a local key secure storage method and device based on a virtualized cryptographic module, and aims to solve the problem that the security, the cost and the flexibility of the existing scheme are difficult to consider at the same time. According to the method, a high-isolation virtualized cryptographic module (VCM) is created through a CPU hardware virtualization technology, a safe execution environment is constructed, and safe generation, packaging storage and unpackaging operation of a master key and a working key are achieved; the device comprises a hardware layer, a virtualization layer, a VCM and a communication interface, and all the layers are coordinated and are subjected to authority isolation. According to the scheme, the system is supported to start automatic loading or on-demand dynamic loading, and various attacks are resisted in combination with TPM (Trusted Platform Module) enhanced packaging and a strong encryption mechanism. Through verification of a 110kV intelligent substation scene, special hardware does not need to be newly added, the cost is only 0.1% of that of a traditional HSM scheme, the operation response delay meets the real-time requirement, and the method is suitable for key security storage scenes of various types of equipment.

One embodiment is related to a method. The method includes receiving from a storage application server computer, a token request message comprising a user identifier associated with a storage application on a user device, and determining a token. The token is a limited use token. The method includes mapping the token to the user identifier, transmitting, a token response message comprising the token to the storage application server computer. The method includes receiving from a processing network computer, a de-tokenization request message comprising the token, after the processing network computer receives an authorization request message comprising the token from a resource provider computer via a transport computer. The method also includes determining the user identifier using the token, and transmitting the user identifier to the processing network computer.

A telecommunications switch-type platform positioned between a private-branch exchange and a data network dynamically hosts artificial-intelligence agents that interpret, transcribe, translate and transliterate live call traffic while the session is in progress. Machine-readable instructions stored in local memory instantiate, migrate and terminate the agents on demand across dedicated processing resources such as field-programmable gate arrays, application-specific integrated circuits or system-on-chip devices, thereby maintaining end-to-end latency below 250 milliseconds. An integrated call-audio recorder captures bidirectional media streams for secure archiving without interrupting service. The platform exposes a network interface that passes both signalling and media traffic, enabling seamless deployment inside call-centre or enterprise environments and compatibility with softswitch architectures conforming to class-4 Computational Interpretation of Communication using Artificial Intelligence standards. The same functional stack is deliverable as a computer-implemented method and as a non-transitory machine-readable medium storing the instructions executed by the processing resources.

Aspects of the present disclosure are to address on-demand system information transmission. The method includes: receiving from the base station a radio resource control (RRC) message, wherein the RRC message includes one or more sets of configuration parameters for secondary cells; receiving from the base station, a specific Medium Access Control (MAC) Control Element (CE) related to activation of an on-demand Synchronization Signal Block (SSB); and determining based on the specific MAC CE activation of the on-demand SSB in a specific secondary cell. The specific secondary cell is determined based on first information of the specific MAC CE, wherein the first information is associated with a secondary cell index of the specific secondary cell. A specific number of bursts of the on-demand SSBs are transmitted in the secondary cell based on second information of the specific MAC CE.

Various embodiments of the present disclosure provide a method for measurement. The method which may be performed by a terminal device comprises: receiving configuration information from a network node. The configuration information may indicate a measurement object (MO) associated with one or more on demand synchronization signal and physical broadcast channel block (OD-SSB) transmissions. In accordance with an exemplary embodiment, the method may optionally further comprise: performing a measurement for the MO according to the configuration information.

The invention discloses an operation control method and device of an energy storage system based on demand defense. The method comprises the following steps: predicting and obtaining a load change curve, a photovoltaic output change curve, an electricity price change curve and a charge state change curve in a future preset time period based on a collected multi-source data set of a user side; predicting the electricity purchase cost of the user based on the electricity price change curve, the load change curve, the photovoltaic output change curve and the charge state change curve; predicting demand penalty cost based on the load change curve and the photovoltaic output change curve; constructing a target function at least based on the user electricity purchase cost and the demand penalty cost; and determining a day-ahead charging and discharging strategy based on the target function and a predetermined constraint condition, and controlling the operation of the target energy storage system based on the day-ahead charging and discharging strategy. According to the invention, the operation of the energy storage system can be reasonably and accurately controlled.

The invention discloses a management method and system for improving the receiving rate of spare parts purchased in the engineering period of a nuclear power plant, and relates to the technical field of industrial automation and monitoring, and the method comprises the steps: extracting consumption records after commercial transportation for units of the same supplier in a group, generating a list of spare parts purchased in the engineering period, and removing low-failure-rate parts; high-frequency consumables are retained to improve pertinence. And secondly, for spare parts which are not present, not scheduled and wide in coverage, a framework protocol is established with a manufacturer to lock unit prices and agree on demand supply, so that inventory occupation caused by mass transfer is avoided. And finally, executing a four-consistency principle on the on-site spare parts, namely checking an EOMM manual, and finally confirming a list, spare part real objects and on-site installation real objects. When the model or the code is upgraded, the specification is upgraded and solidified in combination with the commitment function of the manufacturer and the EOMM manual; when differences occur, interchangeability argumentation is carried out through technical analysis and an item replacement process, spare parts which are argumented to be available are put in storage and updated to an equipment spare part BOM and a template work order, and verifiability, traceability and efficient receiving of the spare parts are achieved.

The invention relates to the technical field of intelligent display, in particular to an intelligent identification method and system based on energy-saving control, and the method comprises the following steps: S1, collecting external environment parameters through a low-power-consumption sensor disposed in an intelligent identification device, the environment parameters comprising environment illumination intensity, a human body or object approach signal and equipment electric quantity information, and transmitting the environment illumination intensity to an intelligent identification device; the data of the environmental parameters are judged; the system can automatically adjust the display brightness and the refresh frequency according to the environment illumination, the target distance and the electric quantity, refresh according to the demand by combining the content priority, and automatically switch the working state and the dormant state according to the human body approaching, the timed task or the remote instruction; in the aspects of communication and energy management, the communication period and power consumption are adjusted in real time, energy self-balancing is achieved, and the service life of equipment is prolonged; in addition, the system periodically analyzes historical operation data and self-learns and optimizes control parameters, so that energy consumption is continuously reduced in long-term operation, display efficiency is improved, and performance stability is kept.

The invention relates to the technical field of software engineering automation of digital data processing, and discloses a game development automation method based on a multi-agent system. The method is used for solving the problem that a machine-readable cross-product dependence model and consistency verification are lacked in a traditional method. According to the method, based on a task list freezing project context, a constraint packet and a threshold aperture, and based on a demand account book and registration demands, constraints, codes, configuration, resources and other entry abstracts, version information and strong and weak reference relationships; the multiple agents only submit change packets in a temporary storage area by writing tokens and occupying fragmentation write-in gating, and an influence range report binding verification list is generated for each change packet; boundary compliance, reference integrity, compilation verification, determinacy operation verification and acceptance assertion verification are sequentially carried out on the temporary storage area snapshots, and when conflicts occur, rollback or confinement repair is triggered to output traceable and acceptance deliverables based on reproducible evidence judgment.

The invention discloses a browser optimization method and system based on dynamic resource loading and intelligent caching, belongs to the technical field of browser optimization, and aims to solve the technical problems of how to improve browser resource loading efficiency, reduce bandwidth consumption and improve user experience. Comprising the steps of predicting a next page possibly accessed by a user or required resources through a lightweight learning model based on a historical browsing path of the user, and obtaining the resources required by the user; carrying out on-demand dynamic loading on resources required by the user; the loaded resources are cached through a multi-level caching strategy; comparing the resources of the local cache of the browser with the latest version of the server side based on the resource fingerprints, and if the resource versions are inconsistent, downloading the resources of the latest version from the server and updating the local cache of the browser; a priority scheduling and parallel loading strategy is adopted for resources of various levels; and dynamically adjusting a resource loading strategy and resource quality according to the current network state.

The invention discloses a zero-work management system and method based on AI and multiple agents, relates to the technical field of zero-work management, and solves the problems that in the prior art, data acquisition is single in dimension and insufficient in refinement. And the technical problem that the subsequent analysis is lack of accurate and comprehensive data support due to the fact that the collection of the basic data of the zero worker does not cover the core information of the real-time distance from the zero worker to the starting point, the historical favorable comment rate and the like which can be directly associated with the performance efficiency and the service quality is solved. The method comprises the following steps: acquiring zero-worker demand data of a task and zero-worker basic data of a zero worker; calculating to obtain a demand priority sequence based on the zero-work demand data; calculating a historical matching degree evaluation coefficient based on the historical zero work basic data; training an artificial intelligence model based on historical data to obtain a matching degree recognition model; based on the matching degree identification model, identifying the basic data of the zero work collected in real time to obtain a corresponding matching degree evaluation coefficient; and performing zero worker matching based on the demand priority sequence. The technical problem is solved.

An industrial chatbot system and associated chatbot interface support various user experience features designed to improve the chatbot's utility within an industrial context. These features include the ability to invite selected users or teams to participate in an ongoing chatbot thread or conversation, the ability to switch between knowledge domains during a chatbot session, the ability to save prompts for selective resubmission to the system, the ability to organize and share multiple chatbot threads among different teams or team members, constraining submission of prompts in a collaborative multiuser context so that prompts are submitted and processed in an organized manner, and the ability to define prompt-based dashboards that can be stored and deployed on demand or in response to defined conditions.

There is provided a method, system and non-transitory storage medium for serving a large language model (LLM) engine by allocating memory on demand. A virtual memory space with virtual pages is initialized for a tensor. A first request with first input tokens is received, added to the tensor by being allocated to first virtual pages and mapped to first page frames in physical memory. The tensor is provided to the LLM engine for processing. A second request with second input tokens is received, and a first output token is received from the LLM engine. The first output token is added to the tensor by being allocated to a second virtual page contiguous to the first virtual pages and mapped to a second page frame. The second input tokens are allocated to second virtual pages and mapped to second page frames in physical memory. The tensor is provided for further processing.

The invention relates to the technical field of cross-culture demand intelligent mining, and discloses a cross-culture demand intelligent mining method and device, electronic equipment and a storage medium, and the method comprises the steps: introducing a multi-modal demand analysis engine to process an original data set containing multi-language mixed data, according to the scheme, the system can analyze text semantics and culture symbols at the same time, then a dynamic mapping matrix based on demand nodes and function nodes is constructed by using a graph neural network, the market adaptability and accuracy of scheme generation are enhanced, and the optimized target mapping matrix is input into a preset optimization model to obtain an improved scheme. The method has the beneficial effects that feasible candidate product improvement schemes can be efficiently generated and screened out on the premise of considering multi-dimensional constraints, so that the conversion process from demands to the schemes is systematized and automated, the product development cycle is shortened, and the decision cost is reduced.

The invention discloses a wide dynamic underwater sound acquisition method suitable for an underwater glider and a hydrophone, and the method comprises the steps: configuring a unique hydrophone into four parallel band-pass filtering and amplifying circuits, each band-pass filtering and amplifying circuit having different acquisition frequency bands; after the underwater glider enters water, underwater sound collection is carried out based on the only installed hydrophone, and underwater sound signals of four collection frequency bands are collected at the same time through four parallel band-pass filtering and amplifying circuits. Through four paths of differentiated gains and frequency segmentation, the wide dynamic full frequency band can be covered, the dynamic range is improved, a single hydrophone integrates four channels, cables and installation space among multiple devices are saved, and the underwater glider compact layout is adapted; the channels are activated on demand, the total power consumption can be reduced, and the endurance time is prolonged. Under the conditions of low power consumption and limited volume of the underwater glider, wide dynamic range signal acquisition can be realized, and signals can be processed in real time.

Disclosed herein are various embodiments for a cloud resources allocation system. An embodiment operates by determining that an application requests a plurality of instances to execute across one or more processors of a cloud services platform. One or more spare instances are requested to fulfill at least a portion of the plurality of instances, and an allocation of at least a subset of the requested one or more spare instances is received. An execution of the application is directed to the allocated subset of the one or more spare instances, including a first spare instance. A notification is received from the cloud services platform that the first spare instance is to be reallocated to a different process. A subsequent execution of the application is redirected to an on demand instance of the cloud services platform.

The invention is suitable for the technical field of low-voltage electric power internet of things communication, and relates to a multi-service event cache dynamic scheduling and reporting method, a medium and a terminal.The multi-service event cache partition architecture is constructed, a metering unit event area, an expansion unit event area and a multi-service general event area are divided, and it is ensured that cache resources are distributed according to needs; establishing a multi-dimensional priority grading model, and defining a priority rule based on a service type, a PIID (parameter identifier) and an event emergency degree; designing a dynamic scheduling mechanism, monitoring the load of each partition in real time, and dynamically adjusting cache resources and a report queue; the priority reporting process is optimized, the optimal transmission channel is selected by combining the PLC / RF dual-mode link state, and fault retransmission is supported. According to the invention, efficient caching and orderly reporting of multi-service events can be realized, the real-time performance and reliability of key services such as metering, power distribution and new energy can be guaranteed, and the multi-service processing capability and efficiency of a low-voltage transformer area communication module can be improved.

The invention discloses a charging pile intelligent peak-valley period adjustment charging energy storage system and method, and relates to the field of charging piles. Comprising a charging pile region time-phased electrical load presetting module, a multi-power-supply-mode dynamic identification and division module, a to-be-charged vehicle demand analysis module, an execution charging queue intelligent scheduling module, an adjacent pile cooperative monitoring recommendation module and an interaction management platform. A charging pile group power supply mode rule base in peak, valley and tip three-level dynamic time periods is constructed, real-time analysis is performed on demands of to-be-charged vehicles in a charging pile group, an emergency charging demand identification and priority response mechanism is performed, the interiors of the tip and peak time periods are screened, the vehicle demands are responded preferentially, and through power supply load uniform distribution and a basic guarantee charging strategy, the charging pile group power supply mode rule base is established. The charging pile group can serve more low-electric-quantity vehicles in the same time period, meanwhile, the intelligent recommendation function guides the user to the adjacent charging pile with the low load, and the service fairness and the user satisfaction degree are improved.

The invention provides a medical material multistage warehouse intelligent management system and method based on demand prediction, and the system comprises an intelligent recognition module which is used for achieving the quick warehouse-in and warehouse-out operation of medical materials; the stockout pre-taking module is used for receiving and recording the request application when the inventory is insufficient; the demand prediction module is used for adopting a prediction model of multi-granularity time decomposition; the dynamic priority adjustment module is used for constructing an evaluation system; the multi-target optimization distribution module is used for carrying out material distribution; the multi-stage warehouse collaborative scheduling module is used for monitoring the inventory difference of the multi-stage warehouses in real time; the real-time monitoring and early warning module is used for monitoring the storehouse in real time; and the intelligent analysis and optimization module is used for generating a visual report. The intelligent level and the operation efficiency of medical material management are comprehensively improved. According to the system, the group fairness and the resource utilization efficiency can be considered while the quick response of the emergency medical demand is ensured, and the stockout risk and the operation cost are remarkably reduced.

This application discloses a training method, device, electronic device, and storage medium based on a large language model. The method includes pre-configuring training modes, which at least include open-ended training and restricted training. The open-ended training and restricted training can run independently or be switched on demand. Responding to the input of the user to be trained, the method selects the training mode and simultaneously conducts training based on a fallback process, outputting the final training evaluation result. The fallback process ensures the continuity of the large language model training process. This application provides a training method including open-ended training, process tree restricted training, and a fallback process. The open-ended training integrates a knowledge base and a large evaluation model; the process tree restricted training includes configurable nodes and eight core functions; and the fallback process has model monitoring, self-interruption, and dialogue switching capabilities, achieving integrated training with flexible interaction, standardized control, and stable operation.

The application provides an automatic disk IO method based on structural mechanics CAE simulation software, comprising the following steps: S1, constructing a memory-disk cooperative computing framework, and dynamically monitoring a data access state in a simulation solving process; S2, automatically exchanging non-active core data structures to a high-performance disk according to a data access frequency, a life cycle and a memory pressure; S3, loading back to the memory on demand when subsequent calculation needs to access the data exchanged to the high-performance disk; wherein the core data structures comprise a stiffness matrix, shape function data, an intermediate result vector and derivative physical quantities, and the exchange process is transparent to the solver. According to the application, the exchange process is highly transparent to the core computing logic of the solver, is automatically managed by the system, and does not need user intervention, so that efficient utilization of memory resources and stable operation of super-large-scale model simulation are realized.