92 results about "Routing model" patented technology

The invention discloses a distributed model training-oriented routing reconstruction method in a hybrid photoelectric network, and belongs to the field of network communication optimization. The method comprises the following steps: constructing a collaborative routing model based on a current network flow state and topological information, and introducing an adaptive weight mechanism to dynamically adjust the priority of optical channels and electric packet forwarding, thereby realizing efficient collaborative utilization of multi-layer forwarding resources. According to the method, path reconstruction can be rapidly completed according to task characteristics and a network dynamic state, so that the system resource utilization rate is improved, and the maximum task completion time delay is reduced. Compared with a traditional fixed routing and semi-dynamic routing method, the method has the advantages that the average completion time delay and the maximum time delay are improved by more than 45% in the large-flow environment generated by large-scale model training, and the method has good expandability and real-time performance and is particularly suitable for burst flow intensive intelligent computing cluster scenes.

The systems and methods disclosed herein enable dynamic selection of a routing model for generation of an output in response to a provided input (e.g., a prompt for a large-language model). Based on the selected routing model, the data generation platform can evaluate the input and / or other suitable system parameters (e.g., system resource usage) to determine a suitable model for processing the provided input. For example, the routing model can determine a technical application associated with the input and dynamically determine to modify the input prior to generation of the output based on system resource measurement values and / or other suitable information, thereby conferring efficiency, security, and accuracy benefits while preserving system resilience.

Systems, methods, and devices that relate to intelligent query decomposition and parallel routing for specialized model processing are disclosed. In one example aspect, the system receives a query from a user comprising a request relating to a particular domain. The system determines, using a decomposition model, a set of sub-queries based on semantic boundaries, syntactics, tasks, relationships, and rules relating to particular domains. The system inputs the set of sub-queries into a routing model to determine a set of specialized models. For each sub-query, the system routes the sub-query to a respective specialized model, generates an output, and assigns a confidence score. The system detects conflicts among outputs using a conflict detection model configured to identify discrepancies. The system generates an aggregated output by combining outputs according to a weighted aggregation algorithm prioritizing higher confidence scores and conflict resolution rules, then displays the aggregated output.

The invention provides a method, a device and a product for realizing intelligent scheduling and cooperative work of multiple unmanned aerial vehicles. The method comprises the following steps: step S1, receiving a task to be processed in real time and acquiring a real-time operation state of each unmanned aerial vehicle in an unmanned aerial vehicle group; s2, performing intelligent dynamic allocation on the tasks by using a pre-established task value scoring model, an unmanned aerial vehicle task adaptation degree model and a global multi-task dynamic allocation optimization model; the global multi-task dynamic allocation optimization model is a mixed integer programming model established by taking maximization of the total task income as a target under a predetermined constraint condition; and S3, after task allocation is completed, a pre-constructed three-dimensional grid airway model G (V, E) is used to plan an airway for each unmanned aerial vehicle executing the task. According to the technical scheme, the mixed integer programming algorithm and the particle swarm optimization are combined, and the global task allocation optimization model is constructed, so that task allocation can be efficiently solved, and various actual constraints can be met.

The invention discloses an analog circuit performance driving automatic wiring method based on multiple modes, and belongs to the technical field of integrated circuit design. According to the method, key information in circuit layout, physical interconnection and logic netlists is extracted by generating an input cost diagram containing horizontal, vertical and via hole wiring cost information and utilizing a multi-scale routing model fused by multi-modal features. Furthermore, the gradient of the input cost diagram to the circuit performance is calculated through back propagation of the neural network, and a wiring guidance diagram is generated through post-processing to guide subsequent detailed wiring based on an A * search framework and symmetric constraints. Finally, the method can improve the wiring precision, achieves the optimization of the performance of an analog circuit, and provides an efficient and intelligent wiring solution for the automatic design of an integrated circuit.

The application discloses a kind of virtual-real fusion registration satellite internet digital twin deduction evaluation methods, comprising: S1.Satellite internet global network data is acquired, network data is collected and time synchronization interpolation is carried out, and network equipment data matrix is constructed;S2.based on network equipment data matrix, establish satellite network basic routing topology graph;S3.based on satellite network basic routing topology graph, construct by user satellite network dynamic service demand vector, establish satellite network dynamic routing model;S4.based on network dynamic routing model, establish satellite internet digital twin deduction model;S5.based on internet digital twin deduction model, introduce real data and carry out pseudo-real correction, realize the virtual-real fusion digital twin dynamic deduction of all stages;S6.multiple dimensions index and evaluation function are established, and score is calculated.The application can overcome virtual-real mapping distortion, evaluation dimension single and dynamic adaptability insufficient and other problems, complete accurate deduction evaluation to satellite internet.

The invention relates to a computer system utilizing a computer model. A method, a computer storage medium, a program product and an apparatus for training an isolated routing model, and a hybrid inference system are disclosed. The method for training the isolation routing model comprises the following steps: acquiring a training task set; obtaining a first model passing rate and a second model passing rate; generating a soft label based on the first and second model passing rates; generating, by the isolated routing model, a prediction result indicating a probability that the training task is to be routed to the first or second major language model; determining a loss function based on the prediction result and the soft label; and adjusting parameters of the isolation routing model based on the loss function. The first and second model passing rates are determined by sampling a plurality of times a result of performing the training task on the first and second large language models, respectively. According to the method, the isolation routing model can be efficiently trained, so that tasks can be accurately routed among a plurality of large language models to obtain balance between performance and cost.

Disclosed in the present invention are a routing model-based network slice recommendation method and system, a device, and a medium. The method comprises: collecting statistics about real-time traffic to analyze network load and demand, obtaining traffic distribution characteristics of the whole network, and pre-constructing basic network slices on the basis of the traffic distribution characteristics; deploying the constructed basic network slices on a switching node in the form of OpenFlow protocol flow tables; on the basis of bandwidth and time delay classification thresholds, classifying the network slices deployed on the switching node; virtualizing physical resources by means of NFV to obtain associations between various network slices and the physical resources; an SDN controller generating a routing policy on the basis of network load and traffic conditions; deploying the routing policy and obtaining an optimal link from the existing routing policy; selecting, from the various network slices, network slices to be allocated to a user, sorting the network slices to be allocated for recommendation, and sequentially recommending same to the user by means of the optimal link. In the present invention, requirements of users of the same type can be met by means of recommendation priority-based sorting.

The invention discloses a virtual-real fusion registration satellite internet digital twinborn deduction evaluation method, which comprises the following steps of: S1, acquiring global network data of a satellite internet, acquiring the network data, performing time synchronization interpolation, and constructing a network equipment data matrix; s2, based on the network equipment data matrix, establishing a satellite network basic routing topological graph; s3, based on the basic routing topological graph of the satellite network, constructing a dynamic service demand vector of a user to the satellite network, and establishing a dynamic routing model of the satellite network; s4, establishing a satellite internet digital twin deduction model based on the network dynamic routing model; s5, based on an internet digital twinning deduction model, introducing real data to carry out simulation correction, and realizing full-stage virtual-real fusion digital twinning dynamic deduction; and S6, multi-dimensional indexes and evaluation functions are established, and scores are calculated. According to the method, the problems of virtual-real mapping distortion, single evaluation dimension, insufficient dynamic adaptability and the like can be solved, and accurate deduction evaluation of the satellite internet is completed.

The invention provides an intelligent design method and device for a highway channel culvert model, electronic equipment and a storage medium, and relates to the technical field of road-related computer-aided architectural design, and the method comprises the steps: recognizing an intersection point of a highway route model and a surface water system, and calculating a direction vector of a center line of a culvert according to a preset culvert intersection angle; the culvert center line information is fused to the terrain surface to generate a three-dimensional culvert center line, and a highway channel culvert model is generated; whether the distance between the center line of the three-dimensional culvert and the existing structure model is smaller than a preset threshold value or not is judged; if yes, the three-dimensional space position of the highway channel culvert model or the highway route model is adjusted, and the distance between the three-dimensional culvert center line and the existing structure model is recalculated in real time till the distance is not smaller than a preset threshold value. By the adoption of the method and device, the electronic equipment and the storage medium, the technical problems that in the prior art, culvert design layout conflicts are difficult to accurately detect, and modeling efficiency is low can be solved.

The invention belongs to the technical field of intelligent manufacturing and production scheduling, and particularly discloses and provides a cabinet flexible production line intelligent scheduling and material collaborative optimization MES system, which comprises a dynamic process modeling module for constructing a dynamic process route model based on order configuration, BOM and equipment capability; the real-time state sensing module forms a production line operation state data set through multi-source data acquisition; the rolling scheduling optimization module generates an initial production scheduling scheme according to a multi-objective function; the material collaborative distribution module performs material alignment verification based on the schedule and generates a synchronous distribution instruction; the disturbance response rescheduling module triggers local re-optimization for events such as equipment faults. According to the system, a dynamic process route is constructed based on order configuration and equipment capability, multi-target scheduling in a rolling window is driven by fusing multi-source real-time data, accurate cooperation of material delivery and process rhythm is realized, and local rapid rescheduling under disturbance events is supported.

The invention provides a rules engine that manages user requests within an interface integrating multiple AI platforms and AI Models. Upon receiving a query, the engine assigns scores based on factors like management rules, user preferences, and contextual information. Determinative scores, such as those enforcing strict enterprise policies, can override others, leading the engine to block or reroute the query. If no score is determinative, the engine forwards the query and associated prompts to a specialized routing AI Model for contextual analysis. Based on this analysis, the rules engine directs the query to the most appropriate AI Model or defaults to the user-designated AI Model. This system balances user intent with rule enforcement, optimizing query processing across various AI platforms while ensuring compliance with enterprise guidelines.

The invention discloses a text-to-SQL (Structured Query Language) generation and auditing method and device based on retrieval enhancement, and the method comprises the steps: analyzing a user intention by utilizing a generative Schema routing model, and automatically selecting a target database and a related table to form a query range; an example similar to the current question is retrieved from the historical question and answer corpus and serves as a Few-Shot prompt to be provided for the large language model; then, self-adaptively constructing a joint prompt according to the difficulty of the problem, guiding the LLM to generate a preliminary SQL query, and correcting an error through a cyclic feedback mechanism; an LLM-driven SQL script automatic auditing and optimization suggestion module is introduced, comprehensive auditing is conducted on the generated SQL from the aspects of grammar specifications, query performance, structure optimization, safety protection and the like, and optimization suggestions are given out and serve as filtering and feedback links before the SQL falls into a library. According to the method, retrieval enhancement and prompt generation technologies are fully combined, and the generation efficiency, reliability and safety are improved while the correctness of the query result is ensured.

An intelligent router for generative artificial intelligence (GAI) model instances optimizes request routing to reduce latency. The system predicts output lengths using a trained response-length predictor and assesses the state of multiple GAI instances, including prompt and decode distributions. It estimates the workload mixing impact of routing requests to each instance and determines selection probabilities using a machine-learning routing model. The router either assigns the request to the most suitable instance or delays routing if conditions are suboptimal. This approach improves end-to-end latency, Time-To-First-Token (TTFT), and Time-Between-Tokens (TBT) by considering the distinct characteristics of GAI workload phases.

The invention discloses a parking garage entering route construction method and device based on search type and geometric type paths. The parking garage entering route construction method comprises the steps that whether a vehicle receives an effective parking instruction or not and enters a parking range and parking space information or not is judged, and effectiveness verification is carried out; path point sampling is carried out, obstacles are judged, and if the path is not feasible, the process is terminated; candidate paths are generated according to the sampling points and a vehicle and bicycle kinematics model, collision detection is carried out, and effective paths are screened; evaluating the path, selecting an optimal path, splicing the optimal path with the search path, and outputting a final parking path for the vehicle to execute; according to the parking garage entering route construction method, a search type route planning method and a geometric type route planning method are combined, through combination of a vehicle and bicycle kinematics model and a Reed-Shepp geometric route model, splicing optimization of a geometric route and a search route is further carried out, computing resources and consumed time are remarkably reduced, and the parking garage entering route construction method is suitable for large-scale popularization and application. The method is more suitable for an embedded platform chip to improve the calculation performance.

The invention relates to a model determination method and apparatus, and an electronic device. The method comprises the steps of determining a bandwidth parameter of a deployment node where a model in a first model set is located; wherein the model guides to generate a query result based on the prompt information; in response to a user query request, inputting a query text corresponding to the user query request into the routing model to obtain a prediction accuracy rate and a prediction cost of the model for the user query request; and determining a target model for processing the user query request from the first model set based on the bandwidth parameter, the prediction accuracy and the prediction cost.

The invention discloses a method for collaborating multiple models based on an instance aware router. The method comprises the following steps: firstly, constructing a sample-level performance distribution diagram of multiple language models in a target natural language understanding task, and carrying out statistics and analysis on prediction effects of different models on different input samples so as to describe performance differences of the models in different scenes of semantic complexity and reasoning requirements; and forming an'intensity domain 'mapping relation reflecting the model capability boundary. And secondly, on the basis of the sample-level performance distribution characteristics, designing an instance-aware routing instruction tuning mechanism, and automatically learning a corresponding relation between input sample characteristics and model performance through an instruction fine tuning training routing model, thereby realizing intelligent selection and distribution from the sample to an optimal large language model. And then, a multi-model cooperative reasoning mechanism is introduced, different large language models are dynamically called for reasoning aiming at samples with different complexity and feature types according to a decision result of the routing model, and multi-model output is fused through result aggregation strategies such as intersection or union, so that the accuracy and robustness of an overall reasoning result are improved. And finally, a unified input and output template and an anonymization processing strategy are introduced, and the multi-model cooperation process is normalized, so that the generalization ability and fairness of the method in a cross-task and cross-domain application scene are further enhanced.

The invention relates to a dynamic routing based on large model self-interpretation and a training method thereof, and the training method comprises a cold start fine tuning stage: obtaining a pre-trained basic small-parameter large language model and an initialized LoRA module, carrying out the supervision fine tuning training of the basic small-parameter large language model based on a synthetic routing data set, a fine-tuned LoRA module is obtained; and an iterative optimization stage: combining the basic small-parameter large language model and the fine-tuned LoRA module to obtain a dynamic routing model. And based on the iterative optimization data set, optimizing the dynamic routing model through a group relative strategy optimization algorithm, in the optimization process, matching a downstream module according to an output result of the dynamic routing model, and calculating a reward function of the group relative strategy optimization algorithm according to an output result of the downstream module, comprising a result correctness reward function and a format correctness reward function. Compared with the prior art, the method has the advantages of high interpretability, high routing accuracy and the like.

This invention discloses a 2.5D packaged die-to-die simultaneous escape routing method based on a pattern-matching binary tree. The method includes: splitting the interconnect network between multiple dies into multiple sets of simultaneous escape routing models for two-die interconnections; and dividing the signal network in each model into multiple network sets based on the separation characteristics of power and ground networks on signal networks; for each network set, performing fan-out routing planning and layer allocation based on a pattern-matching binary tree, specifically including: setting the routing mode, constructing the pattern-matching binary tree, and backtracking the pattern-matching binary tree; finally, completing the routing based on the fan-out routing planning and layer allocation results. This invention proposes a collaborative planning technology for routing sequence planning and layer allocation in 2.5D IC packaged die-to-die simultaneous escape routing, aiming to solve common routing sequence and layer allocation problems in traditional die-to-die routing. This technology, through optimized routing sequence planning and layer allocation strategies, automates the routing process and significantly improves routing quality and efficiency.

The invention discloses a data multi-level cache collaborative acceleration method and system, and mainly relates to the technical field of data cache acceleration. Comprising the steps of obtaining data access features and inputting a model fusing sliding window statistics and exponential decay to calculate a dynamic heat value; dynamically scheduling the data to a high-speed memory, a distributed sharing or a local file cache level according to the heat value; optimizing a cross-level query path based on the intelligent routing model; the multi-level cache consistency is maintained by adopting an asynchronous version control mechanism; executing hierarchical collaborative cleaning and elastic capacity expansion according to the real-time popularity and the capacity utilization rate; and realizing hot data preloading by utilizing the prediction model. The method has the beneficial effects that through intelligent scheduling and collaborative optimization, the cache hit rate and the system throughput are remarkably improved, the access delay and the storage cost are reduced, and the stability and the data consistency under high concurrency are enhanced.

A hybrid inference system for a coding assistant utilizes a routing model to predict whether output generated by a large language model for a given prompt would be accepted by a user of the coding assistant. The routing model routes the prompt when the routing model indicates that the output generated by the large language model is likely to be accepted. The routing model routes the prompt to a local model when the output generated by the large language model is not likely to be accepted. The routing model is trained on the historical output generated by the large language model for various prompts and the acceptance or rejection of the output by users of the coding assistant.

The invention provides a routing optimization method and system, a medium and electronic equipment. The routing optimization method comprises the following steps: constructing a routing model of the XGPON network; the routing model comprises a plurality of network nodes; performing clustering processing on the plurality of network nodes so as to divide the plurality of network nodes into a plurality of clusters; setting a routing optimization objective function with minimum time delay as a target for each cluster, and setting a constraint condition; and solving the routing optimization objective function with the constraint condition based on a hybrid particle swarm algorithm to obtain an optimization scheme so as to realize routing optimization of the XGPON network based on the optimization scheme, the invention provides a novel routing optimization method based on a hybrid particle swarm algorithm by taking network minimum time delay as a target, and well meets the real-time requirement of network data transmission.

The application discloses a cargo transportation route planning method and device and a storage medium, which can improve the matching number of cargo and vehicle operation routes and improve resource utilization. The cargo transportation route planning method comprises the following steps: based on transportation data and preset model parameters, input data and an initial solution of a route model are constructed; wherein the initial solution comprises at least one single-trip loading and unloading route of cargo between a transfer field and a network point; based on a first preset constraint condition and an optimization target, the initial solution is optimized through the route model to obtain an optimization result; based on the optimization result and a preset shipping condition, an optimized single-trip loading and unloading route is output; and based on a task matching model, the optimized single-trip loading and unloading route is matched with a vehicle original transportation route to obtain a vehicle cargo transportation route.

The invention discloses a fast range routing method based on a key flow, which completes training of a flow identification agent by constructing a reinforcement learning algorithm, realizes judgment of flow and link attributes according to flow data, selects a key flow and a key link in a regional manner, and on the basis of the judgment, performs routing on the key flow and the key link. And inputting a key flow selection result and a key link selection result into a key flow range routing model, and generating routing strategies for the key flow and the non-key flow by adopting a key flow routing generation sub-model and a non-key flow routing generation sub-model respectively. According to the model, the range routing model is adopted to only calculate the key flow, and meanwhile, only the constraint is constructed for the key link, so that the number of model variables and constraints is reduced, the solving speed of the range routing model is remarkably improved, and the expandability of the range routing model is effectively optimized.

A multi-commodity flow-based synchronous escape routing method and apparatus, and a medium. The method comprises: creating a grid pin array (GPA) routing graph on the basis of a chip GPA, and creating an irregular pin array routing graph on the basis of an array having a miss pin; creating escape boundary nodes in synchronous escape routing on the basis of the created GPA routing graph; creating an overall synchronous escape routing graph on the basis of the created escape boundary nodes; on the basis of the created overall synchronous escape routing graph, creating a multi-commodity flow-based synchronous escape routing model; solving the synchronous escape routing model to obtain a routing result, and designing a constraint for grouping and same-layer arrangement in the synchronous escape routing; and optimizing routing paths of the synchronous escape routing model, and accelerating the synchronous escape routing on the basis of an optimization result. By modeling routing constraints and path optimization methods, the present invention supports the simultaneous action of multiple acceleration methods, thereby improving system efficiency.

Disclosed in the present invention is an SDN-architecture-based routing method for guaranteeing network QoS. The method comprises: on the basis of network topology information, performing abstraction to obtain a directed graph; then, on the basis of an improved K-means algorithm, dividing data into mouse flows, elephant flows and mammoth flows; defining a traffic transmission constraint condition; constructing a DQN neural network; training the DQN neural network by using the maximum Q value, which corresponds to an action a having the maximum reward function R, as an objective, so as to obtain a QoS guarantee routing model; obtaining an optimal path weight value; determining a data packet sending path on the basis a Dijkstra algorithm; and finally issuing path information to a switch. The present invention avoids the situations in which data processing performed by using a conventional K-means algorithm is not stable enough and a classification result is not accurate enough. During the training of a DQN neural network, data is extracted from an experience pool according to the priority of data, such that the neural network pays more attention to important data, thereby improving the learning efficiency of the neural network and the performance of the algorithm.

This invention discloses a power communication routing optimization method considering fault risk. The method establishes a primary and backup routing model for the communication system; correlates the importance of load control services in the power system with the information flow, proposing a communication link importance evaluation index; then, combined with a conditional risk measurement model, it proposes a power communication routing optimization method; it treats communication service losses in each fault scenario as risk values, aiming to minimize the losses of important services across all fault scenarios, and considers constraints such as node constraints, bandwidth constraints, and routing constraints to solve for the primary and backup routing paths of the power communication network. This invention flexibly formulates communication routing paths based on changes in the fault area, and the route selection can proactively avoid fault risk areas, effectively reducing load losses during fault processes; it can be used to formulate power communication routing schemes, improving the resilience of power communication systems in responding to faults.

The application discloses a large model routing data synthesis method with diversified expression styles, constructs a user agent to simulate the language style and preference of a real user, generates query questions with diversified styles and clear semantics, inputs the query questions into a candidate language model for reasoning, and generates a benchmark answer by a large language model with superior performance; based on the comparison between the output of the candidate model and the benchmark answer, the content quality is automatically evaluated, the user satisfaction evaluation is simulated in combination with the user agent features, and multi-dimensional routing labels are comprehensively generated; finally, the question, the candidate model output and the routing label are packaged to form a routing data set, which is used for training or evaluating a large model router. The application can automatically synthesize a routing data set covering multiple expression styles, multiple question types and performance differences of multiple models, significantly improves the generalization ability and practicability of a routing model in a real open domain question and answer scene, and does not require high artificial labeling cost.

The present invention discloses an adaptive routing method for a computing network, which is implemented using a computing network, wherein the computing network includes computing nodes, a router network, and a transmission link. The method comprises: obtaining a data packet to be transmitted and network structure information; the data packet to be transmitted includes a data stream, a source address, and a destination address; performing segmentation and partitioning processing on the router network to obtain segmentation result information; the segmentation result information includes a first router set #imgabs0# and a second router set #imgabs1#, respectively determining, based on the source address and destination address in the data packet to be transmitted, a router connected to a computing node corresponding to the source address and a router connected to a computing node corresponding to the destination address; determining whether the router connected to the computing node corresponding to the source address and the router connected to the computing node corresponding to the destination address are in the same group to obtain a first judgment result; if the first judgment result is yes, processing the segmentation result information and the network structure information using an intra-group routing model to obtain intra-group routing information; if the first judgment result is no, processing the segmentation result information and the network structure information using an inter-group routing model to obtain global routing information.