60 results about "Model migration" patented technology

The invention discloses a comprehensive traffic low-altitude global intelligent sensing method based on autonomous evolution, and belongs to the crossing field of intelligent traffic and computer vision. The method comprises the following steps: collecting multi-source sensing data and extracting a structured causal variable; a dynamic causal graph is constructed, and attribution analysis of perception errors is realized through intervention learning; realizing cross-scene model migration and few-sample self-adaption based on the knowledge graph; carrying out adaptive reasoning by adopting an environment-aware dynamic graph neural network; and realizing a causal-driven autonomous evolution closed loop based on multi-source feedback. According to the method, the crossing of the perception system from'perception-optimization 'to'understanding-evolution' in a low-altitude complex traffic environment is realized, the perception precision, the scene adaptability and the system interpretability are remarkably improved, and the manual operation and maintenance cost is reduced.

The invention provides an automatic operation and maintenance method and system suitable for a closed system. The technical problem that a traditional operation and maintenance scheme is difficult to apply due to network isolation and sample scarcity in closed environments such as finance and energy is solved. The method comprises the following steps: generating a unique and traceable migration identifier for all operation and maintenance data, models and reasoning results through a data and migration management module; quantitatively calculating a transferability score between the source domain and the target domain through a transferability evaluation module; an optimal model migration strategy is dynamically selected according to the mobility score, and efficient self-adaption of the model is achieved; closed-loop operation and maintenance are executed through a multi-agent cooperation system comprising detection, diagnosis and repair agents, and self-learning and self-optimization of the system are realized through small sample active learning and a knowledge base evolution mechanism. According to the method, rapid construction, continuous evolution and whole-process traceability of the artificial intelligence operation and maintenance capability in the closed system are realized, and the method has remarkable innovativeness and industrial application value.

The invention provides a corresponding method for solid engine digital twin modeling and design optimization aiming at the problems of lack of an integrated parametric modeling framework, weak proxy model mapping capability, lack of a cross-working-condition / cross-model migration mechanism and the like in the background technology. The invention provides a digital twin modeling and optimizing method for a solid engine. The method comprises the following steps: step 1, setting parameters and inputting a model; 2, parametric modeling and simulation calculation are carried out; 3, configuring input and output of the proxy model; 4, extracting hidden variable features; 5, training a mixed structure agent model; step 6, parameter optimization based on a calibration mechanism; and 7, optimizing a result and analyzing.

The invention discloses a cross-tax finance and taxation inspection method and device and a medium, and the method comprises the steps: obtaining multi-modal data according to a preset business path, carrying out the evaluation calculation of the multi-modal data, carrying out the processing of the multi-modal data according to a preset multi-modal cross verification mechanism, and carrying out the evaluation calculation of the multi-modal data; constructing a finance and taxation knowledge graph according to the set of entities and relationships, and determining priori knowledge according to the finance and taxation knowledge graph so as to determine a heterogeneous graph risk pre-training model in combination with the priori knowledge and a preset attention network; training the heterogeneous graph risk pre-training model according to a pre-sampled sample, and performing feature space alignment and model migration on the trained model to obtain a cross-tax migration model; and performing multi-tax risk collaborative identification through the cross-tax migration model to determine a risk point, performing motivation analysis to determine a motivation probability of the risk point, and performing finance and tax inspection decision according to the motivation probability so as to determine a finance and tax inspection report.

The application discloses a table question and answer task ability enhancement processing method of a large language model. For table data, a TABLE-UAM network including an encoding part and a decoding part is constructed; different table data sets are sequentially input into the TABLE-UAM network for two-stage training, the first stage is difference reconstruction training, and the second stage is combined with multiple large language models for training; the trained TABLE-UAM network and the large language model are spliced, and used for processing input question text and table data to output answers. The newly-built TABLE-UAM network structure can efficiently encode and feature extract table data, effectively perceive table row and column dependency features, multi-table dependency features and global relationship features; and the two-stage training method can improve the table analysis task ability, significantly enhance the cross-model migration ability, and has strong generalization and practical value.

The invention relates to the technical field of automobile in-loop simulation, in particular to an automatic emergency braking in-loop simulation method and system, and the method comprises the steps: constructing a virtual simulation environment comprising a plurality of traffic scenes and a vehicle dynamics model; an AEB decision model is constructed based on a preset reinforcement learning algorithm, a state space of the AEB decision model at least comprises a vehicle state, a target object state, a driving area state, a conflict risk field and a traffic consciousness code, and an action space of the AEB decision model at least comprises an intention action and a control action; and AEB controller hardware is accessed to the virtual simulation environment, and real-time interaction and closed-loop training of the AEB controller and the virtual simulation environment are realized. The method can solve the problems of difficult model migration, poor adaptability, lack of online learning ability and the like in the prior art, and has good migration ability and adaptability.

This application provides a model migration method, device, storage medium, and program product, particularly relating to the field of artificial intelligence chip technology. The method includes: reading the original model under the original framework; extracting the original definitions of each module and the original configuration data structure of the original model; converting the original definitions of each module into target definitions under the target framework, and simultaneously converting the original configuration data structure into a target configuration data structure under the target framework. Next, based on the model structure definition template of the target framework and the target definitions of each module, the model structure definition under the target framework is automatically generated, achieving automated model structure conversion; finally, based on the model structure definition and target configuration data structure under the target framework, the target model under the target framework is obtained, thus automating model migration and effectively improving the efficiency of model migration.

The invention discloses a table question and answer task capability enhancement processing method of a large language model. Aiming at the table data, constructing a TABLE-UAM network comprising a coding part and a decoding part; the different table data sets are sequentially input into the TABLE-UAM network for two-stage training in sequence, the first stage is reconstruction difference training, and the second stage is combined with multiple large language model training; and splicing the trained TABLE-UAM network and the large language model for processing the input question text and table data and outputting an answer. According to the method, a newly established TABLE-UAM network structure can efficiently perform coding and feature extraction on table data, and the table row and column dependency feature, the multi-table dependency feature and the global relationship feature are effectively perceived; moreover, the two-stage training method can improve the table analysis task capability, remarkably enhances the cross-model migration capability, and is higher in generalization and practical value.

The invention discloses a motor cross-domain optimization agent model transfer learning method and system and a storage medium, and relates to the technical field of motor design, and the method comprises the following steps: S1, pre-training a source domain agent model based on a first data set; s2, performing rapid prediction in the new target domain parameter space, and screening out bridge design points in an intelligent sampling mode; s3, only performing high-fidelity simulation on the bridge design points, and constructing a target domain data set; s4, loading a source domain agent model, freezing a front-end layer, finely adjusting a rear-end layer by using a target domain data set, and migrating into a target domain agent model; and S5, replacing high-fidelity simulation with the target domain agent model, and finding out optimal design parameters of the target domain in combination with a global optimization algorithm. According to the agent model transfer learning method and system for motor cross-domain optimization and the storage medium, the number of expensive simulation times needed in a new design domain is greatly reduced, the calculation cost is remarkably reduced, and the optimization efficiency is improved.

The application relates to the field of artificial intelligence and hardware acceleration technology, in particular to a model migration method and system, an electronic device and a computer storage medium, the method comprising the following steps: determining a mapping operator and a replacement operator of a to-be-migrated model based on hardware characteristic description information of a target platform, wherein the to-be-migrated model is a model to be migrated to the target platform; obtaining a task type of a calculation task, determining a network structure based on the task type and the hardware characteristic description information, wherein the calculation task is a to-be-executed task input by the to-be-migrated model, and the network structure is used for adapting the to-be-migrated model to the target platform; and deploying the mapping operator and the replacement operator to the target platform through the network structure and the hardware characteristic description information, so as to obtain an executable model adapted to the target platform. The application solves the technical problem of insufficient operator adaptation caused by hardware architecture difference when a model is migrated to an NPU.

The application provides an automated operation method and system suitable for a closed system. It aims to solve the technical problem that traditional operation schemes are difficult to apply due to network isolation and sample scarcity in closed environments such as finance and energy. The method includes: generating a unique and traceable migration identifier for all operation data, models and inference results through a data and migration governance module; quantitatively calculating the transferability score between the source domain and the target domain through a transferability evaluation module; dynamically selecting the optimal model migration strategy according to the transferability score to achieve efficient self-adaptation of the model; through a multi-agent collaboration system containing detection, diagnosis and repair agents, a closed-loop operation is performed, and the system is self-learning and self-optimizing through small sample active learning and knowledge base evolution mechanism. The application realizes the rapid construction, continuous evolution and whole-process traceability of artificial intelligence operation capability in a closed system, and has significant innovation and industrial application value.

This invention discloses a method, system, device, storage medium, and product for deploying large-scale model inference. The method involves dividing the vocabulary of a large-scale model into blocks to obtain several plaintext blocks; generating an initialization vector based on the MAC address of the server to be deployed; randomly generating a first key; encrypting the plaintext blocks using a ciphertext block chaining mode based on the initialization vector and the first key to obtain several ciphertext blocks; encrypting the first key using a public key to obtain a second key; and sending the second key and the ciphertext blocks to the server to be deployed, enabling the server to decrypt the second key and the ciphertext blocks using its private key to obtain the vocabulary of the large-scale model and perform inference deployment. Using this invention, the security of large-scale models can be improved, effectively avoiding the leakage of source code during large-scale model migration and deployment, and preventing data from being tampered with or stolen during transmission.

The invention discloses a diffusion model migration method based on prediction residual guidance and a related device, and relates to the technical field of diffusion model knowledge migration, and the method comprises the steps: taking noise data as input at each denoising time step, and carrying out denoising on the noise data; determining a first prediction noise, a second prediction noise and a third prediction noise by using the basic model, the adaptive model and the target model respectively, calculating a deviation between the first prediction noise and the second prediction noise to obtain a prediction residual error, performing weighted summation on the third prediction noise and the prediction residual error to obtain a guide prediction noise, and outputting the guide prediction noise. And on the basis of the noise data and the guide prediction noise, noise data of the next denoising time step is calculated until the last denoising time step is reached, and an output result of the target model is obtained, and the output result is a picture, a video, a voice or a text. According to the method, the knowledge migration of the diffusion model can be completed on the premise of not accessing original training data and not training.

The invention relates to the technical field of elevator intelligent operation and maintenance and fault prediction, and provides an elevator predictive fault early diagnosis method based on data driving. Constructing a physical constraint library containing dynamic constraints, time sequence continuity constraints and multi-sensor coupling rules; a conditional generative adversarial network with physical constraints is designed, the rule base is embedded into the training process in a loss function form, and a high-fidelity fault sample conforming to the elevator operation rule is generated; constructing a balanced data set after double screening of time sequence similarity and physical rules; on this basis, a meta-learning mechanism is introduced, a small sample fault diagnosis model is trained, and model migration can be rapidly completed only by using a very small number of real samples of the target elevator through the two stages of meta-training and meta-adaptation. Accurate predictive diagnosis of two typical early faults of elevator traction machine bearing micro-wear and portal crane belt slight slip is realized.

The application provides a black box confrontation evaluation method based on strategy driving and frequency-space residual remodeling, belongs to the technical field of confrontation attack, selects neural networks with multiple different network architectures to construct a heterogeneous agent model matrix, obtains a source detection tensor and a calibrated true value and extracts a benchmark response gradient; the current evolution state tensor, the benchmark response gradient, the frequency domain amplitude spectrum feature and the iteration progress scalar are spliced to form a multi-dimensional joint state vector; a frequency domain attenuation barrier and a space domain remodeling tensor are generated through a strategy decision network, the frequency domain filtering denoising and the space domain nonlinear residual remodeling are sequentially performed on the benchmark response gradient, and the evolution state tensor is updated in the residual fusion remodeling direction. After the multi-round iteration and the composite feedback signal optimization strategy network parameter, the tensor with the optimal comprehensive threat effectiveness is selected as the confrontation evaluation carrier output. The application can effectively strip high-frequency overfitting noise, solve the cross-model response offset problem, and improve the cross-model migration and evaluation stability of the black box confrontation evaluation carrier.

The invention discloses a project full-life-cycle intelligent management method and device, equipment and a medium. The method comprises the following steps: acquiring multi-source heterogeneous data in a full life cycle of a project to obtain original multi-source data of the project; preprocessing the original multi-source data of the project, and generating comprehensive project state data through cross-system fusion processing; and based on the comprehensive project state data, performing calculation processing through a reinforcement learning decision framework to obtain a management strategy of each stage, wherein the reinforcement learning decision framework comprises the steps of predicting a time sequence change trend by adopting a GRU model of an embedded model migration sub-module, performing risk assessment by adopting a Bayesian network, and performing calculation by adopting a PPO algorithm after an objective function is adjusted. And actual data and a prediction result can be obtained for comparative analysis, and a management strategy is iteratively optimized. According to the method, the problems of fragmentation and stage splitting of traditional management data are solved, accurate decision support is provided, intelligent management is promoted, and management efficiency and decision scientificity are improved.

The application provides a medical image segmentation method based on a polar fine-tuning large model migration method, and solves the technical problem that when a large model is migrated from a natural image to a more complex medical image task, a traditional fine-tuning strategy has discrete fitting effects and cannot fully capture potential complex features in data when facing high-dimensional input and complex images. The method comprises the following steps: S1, rotating an image of a skin lesion with a center point to form a new picture; S2, using a polar fine-tuning strategy to inject self-attention in a decoder and feed-forward MLP in parallel with a low-rank branch; S3, inputting decoupled features and prompt flow features into the decoder; S4, inputting the decoupled features and the prompt flow features into the decoder to output a predicted mask image; and S5, building a large model migration learning deep learning network PoSAM based on the polar fine-tuning strategy. The application performs decoupled correction on features, and quickly aligns the offset without damaging the original representation structure.

The application provides a PDAF model migration method based on deep learning, comprising: training a network under a target module dataset to obtain a target model; calibrating physical differences between an existing module of the same type and the target module to obtain a correction module; and migrating the target model to the existing module and correcting according to the correction module. The application can migrate the target model trained based on the target module to the existing module of the same type, reduces the repeated data acquisition and model training process, and greatly improves the model reusability and algorithm reference efficiency.

The application relates to the technical field of network security, and discloses an intrusion detection method and system based on data enhancement and model migration, which comprises the following steps: obtaining local network data of each edge client node and constructing a local training set; based on the local training set, completing adversarial verification enhancement of small sample attack data and local model training to obtain local model update parameters; performing dynamic trust evaluation on the local model update parameters to obtain comprehensive trust scores of the edge client nodes; based on the comprehensive trust scores, adopting a meta-gradient aggregation strategy to aggregate the local model update parameters to generate a global intrusion detection model; and deploying the global intrusion detection model to each edge client node, executing intrusion detection through a local detector optimized by multi-target cooperative reinforcement learning to obtain an intrusion detection result.

Embodiments of the present application provide a model migration method, device and equipment and a computer storage medium. The method comprises: performing model migration on a base model to generate a first function; constructing a second function according to the first function, a global regularization term and a local regularization term; updating the second function to the first function, returning to construct the second function according to the first function, the global regularization term and the local regularization term until the number of updates reaches a first threshold to obtain a third function; solving a hyperplane set corresponding to the minimization of the third function; generating a first model based on the hyperplane set; and training the first model based on a sample data set until a training stop condition is met to obtain a target model. The model migration method, device and equipment and the computer storage medium provided by the present application can enable the obtained target model to identify one or more newly added class labels without the need to re-establish a model, thereby saving time, manpower and computing resources.

The application discloses a kind of comprehensive traffic low altitude global intelligent perception methods based on autonomous evolution, belong to the cross field of intelligent transportation and computer vision.Method includes: collection multi-source perception data and extract structured causal variable;Dynamic causal diagram is constructed, and attribution analysis of perception error is realized by intervention learning;Cross-scene model migration and few-sample self-adaptation are realized based on knowledge graph;Adaptive inference is carried out using dynamic graph neural network of environment perception;Autonomous evolution closed loop driven by causality is realized based on multi-source feedback.The application realizes the leap of perception system from "perception-optimization" to "understanding-evolution" in low-altitude complex traffic environment, significantly improves perception accuracy, scene adaptability and system explainability, and reduces artificial operation cost.

The invention relates to a model dynamic migration decision-making method based on parallel discrete event scheduling. The method comprises the following steps: acquiring multi-dimensional load data of each model in the distributed system; constructing a hybrid prediction model, inputting each piece of multi-dimensional load data into the trained hybrid prediction model, establishing a mapping model of task load and operation time for each simulation event scheduler according to an output load prediction result, dynamically updating parameters of the mapping model by adopting an incremental learning strategy, and obtaining predicted operation time of the simulation event scheduler; constructing a dynamic migration model by taking a scheduler operation time variance in the engine operation process as a target and taking a model unique mounting constraint, a self-migration prohibition constraint and a migration logic coherence constraint as constraint conditions; and solving the dynamic migration model by taking the predicted running time of each simulation event scheduler as a decision basis, and outputting an optimal model migration scheme. By adopting the method, the operation efficiency and the simulation precision of the simulation system can be improved.

The invention discloses a deep model resource fair scheduling method for multi-task service quality assurance in a cloud-edge collaborative environment. The method comprises the following steps: S1, a task request sensing module; s2, an online model evaluation module; s3, a model fair selection module; s4, an isomorphic model migration module; s5, a heterogeneous model collaborative scheduling module; and S6, a service quality fairness guarantee module. The method has high engineering practical value, and can be applied to various real-time reasoning scenes such as intelligent medical treatment, intelligent security and protection, industrial internet and the like.

The invention provides a model training and reasoning method, a control device and an automatic driving vehicle, and relates to the field of artificial intelligence, in particular to the field of automatic driving. The model training method comprises the following steps: extracting feature information of first sensor data collected under first sensor configuration; determining spatio-temporal information of the first sensor data; training a task model in the automatic driving scene by taking the feature information and the spatio-temporal information of the first sensor data as training data, so that the task model learns a first association relationship between the feature information and the spatio-temporal information of the sensor data and a task prediction result, the first association relationship is used for enabling the task model to automatically adjust the task prediction result when the spatio-temporal information changes due to the change of the sensor configuration. The model migration cost is reduced, and the compatibility of the model to sensor configuration is improved.

The application provides a cross-platform model migration method and device, and belongs to the field of artificial intelligence. The method provided by the application comprises: managing chip information configuration in a chip configuration subpage; adjusting a target model migration task in a model migration management subpage, determining a target tab based on a model corresponding to the target model migration task, executing the target model migration task and displaying the corresponding task progress in the target tab; determining pre-migration chip information corresponding to a pre-migration model completing the model migration task; generating an evaluation task of the pre-migration model based on the pre-migration chip information, and modifying the number of model migration task display bars in the tab corresponding to the pre-migration model in the model migration management subpage based on the result of the evaluation task. The cross-platform model migration method and device provided by the application can realize efficient, reliable and visual cross-platform migration of artificial intelligence models from Nvidia chips to domestic chips.

The invention provides a model training method and device and a storage medium, and the method comprises the steps: determining a target frame of a target model and the related information of a target robot according to a configuration operation, and obtaining a pointing relation between modules and the attribute information of each module; generating intermediate representation information of the target model; according to the related information of the target robot, standard modal information is constructed and obtained; according to the standard modal information, performing modal architecture adaptive enhancement processing on the intermediate representation information to obtain enhanced representation information; generating a training script of the target model and configuration information corresponding to the training script according to the enhanced representation information, the target framework and pre-registered instance information; and training to obtain a target model. According to the method and the device, the compatible training script of the target robot can be generated through the configuration operation in the graphical user interface, and the target model is obtained through training, so that automatic training of the model is realized, and the model migration and reuse cost is effectively reduced.

The invention provides a model training method and electronic equipment, and the method comprises the steps: determining a target frame of a target model according to a configuration operation of a user in a graphical user interface, and obtaining at least one module, a pointing relation between the modules, and attribute information of each module; generating intermediate representation information of the target model according to the modules, the pointing relationship between the modules and the attribute information of the modules; according to the target framework and the intermediate representation information, generating a training script of the target model and configuration information corresponding to the training script; and based on the training script and the configuration information, training to obtain a target model. According to the method, the compatible training script can be generated through the configuration operation in the graphical user interface, and the target model is obtained through training, so that automatic training of the model can be realized, and the model migration and reuse cost is effectively reduced. Meanwhile, the development period can be remarkably shortened, and errors caused by manual coding are reduced.

The application discloses an indoor personnel state feature and a variety of energy demand estimation method based on information mining, and belongs to the field of building energy consumption accurate prediction.The application applies data mining technology, applies data expression to the influence of "people", and adds the simulation platform based on the traditional physical model, so that the problem that the traditional physical model cannot depict human behaviors is overcome.Meanwhile, the method can better perform model migration on different buildings, so that the data-driven model can better perform generalization.

The invention provides a method for measuring an optimal quantization bit width in a neural network model. The method comprises the following steps: acquiring a fitting relationship between overall data distribution similarity and accuracy loss before and after model migration; according to the accuracy loss threshold value of the model and the fitting relation, determining the threshold value of the overall data distribution similarity of the model; according to the weight corresponding to each layer in the model, the threshold value of the overall data distribution similarity is allocated to each layer, and the threshold value of the single-layer data distribution similarity of each layer is obtained; and determining the optimal quantization bit width of each layer according to the threshold value of the single-layer data distribution similarity. Correspondingly, the invention further provides a system for measuring the optimal quantization bit width in the neural network model, a storage medium and electronic equipment. Therefore, the optimal quantization bit width can be measured under the condition that hardware deployment and quantization bit width traversal are not carried out, the deployment cost is reduced, and the time overhead and the calculation frequency are reduced.

The invention belongs to the technical field of software communication, and relates to an SCA model automatic migration method and terminal.The complete structure information of a to-be-migrated old version model is automatically loaded and analyzed through a model analyzer, then elements are automatically replaced by a mapping engine according to a preset mapping rule, a dependency relationship is rebuilt, and external reference is recorded; the method comprises the following steps of: firstly, distributing unique identifiers for new and old version models and recording change information through a version controller to realize visual tracking of change content, time and range, and finally, sequentially executing new version SCA grammar rule verification and UML grammar semantic verification through a verifier after the models are generated, so as to realize visual tracking of the change content, time and range. According to the method, the compatibility problem is found and fed back in time, a large amount of subsequent debugging work caused by lack of systematic verification in a traditional method is avoided, automation, process and standardization of SCA model migration are integrally achieved, migration efficiency and reliability are remarkably improved, and meanwhile version management experience is optimized.