381 results about "Memory footprint" patented technology

The invention provides a target recognition model reasoning optimization method and device, and the method comprises the steps: firstly carrying out the structural analysis and sensitivity evaluation of a pre-training model, extracting the structural features of each network layer, activating the distribution features, carrying out the quantitative sensitivity scoring, and constructing a data set reflecting the hierarchical features and fault-tolerant capability; and querying a quantitative configuration knowledge base based on the data set to generate a heterogeneous quantitative strategy. Layered low-bit quantization is executed according to the strategy, and a layered weighted loss function is introduced to carry out quantization perception training, so that precision loss caused by bit width compression is effectively compensated. According to the method, through hierarchical heterogeneous quantification, the model recognition precision is preserved to the maximum extent while high compression ratio and reasoning acceleration are achieved, and particularly, the performance of a high-sensitivity layer is protected. The generated heterogeneous quantitative model remarkably reduces memory occupation and power consumption, is suitable for an edge hardware platform with limited resources, forms a set of complete automatic process from analysis and configuration to training compensation, and has good universality and engineering practical value.

The embodiment of the invention provides a dynamic resource scheduling method and system based on reinforcement learning, and the system comprises a state sensing module which is used for collecting and preprocessing the resource state data of each node in a cluster, and the resource state data at least comprises a CPU utilization rate, a memory occupancy rate, a network bandwidth utilization rate, a task queue length and a node load; the action decision module is used for outputting a scheduling action according to the current state representation vector, and the reward calculation module is used for calculating a reward value according to an actual operation result of the system; the model training module is used for training the strategy model by using a reinforcement learning algorithm, and the deployment optimization module is used for deploying the trained strategy model to a production environment. Complicated and changeable workloads and resource states in the cloud environment can be automatically dealt with, the manual intervention cost is remarkably reduced, and the scheduling efficiency and accuracy are improved. And a comprehensive and unified environment perception capability can be constructed.

A method of serving a generative transformer model includes determining a batch size to use in processing inference requests and allocating at least one prompt pipeline and at least on token pipeline to the generative transformer model to process the batch of inference requests. The number of prompt pipelines and the number of token pipelines, and the depths of the pipelines are determined based on the batch size, an average prompt length, a cache requirement per stage, and a memory footprint of model weights for the generative model per stage using a resource allocator component of the model serving system. Cache streaming is used to stream prompt cache from prompt pipelines to token pipelines to generate tokens. Cache streaming involves gather-copy operations which may be performed using compute kernels.

The invention belongs to the technical field of integrated circuit simulation verification, particularly relates to an RISC-V simulation resource dynamic generation method and system, and solves the problems of instruction consistency and variable-length instruction truncation through a metadata double-table structure and a cross-boundary instruction splicing mechanism. Virtualized two-stage translation support and abnormal injection are realized through a recursive multiple hit detection and dynamic attribute bit modification mechanism. According to the method, complete storage is replaced with lightweight metadata, memory occupation is remarkably reduced, the problem that memory occupation is linearly increased along with time is solved, simulation efficiency and consistency are improved, and the method is suitable for full-system verification of a high-performance RISC-V processor.

The invention relates to a lightweight neural network target detection model optimization method for an embedded device, and belongs to the technical field of target detection, and the method comprises the steps: determining the hardware constraint and detection task boundary of the embedded device; adjusting a network infrastructure based on hardware and task characteristics; network parameter redundancy is eliminated; designing a scene adaptive dynamic feature selection mechanism; compressing parameter storage precision; optimizing calculation-intensive operation; adjusting a memory access mode; an embedded deployment framework is adapted; and constructing a closed-loop iterative optimization process. The method is beneficial for realizing lightweight and efficient operation of the model, improving detection precision and stability, reducing memory occupation and power consumption, and meeting hardware constraints of embedded equipment.

The invention discloses an interactive document editing canvas method, which comprises the following steps of: establishing a data-view mapping relation directly driven by a Vue2 responsive system between a canvas container and an element array, any addition, deletion and modification operation aiming at the node description data set can be instantly and accurately reflected to the style and attribute of the corresponding DOM child node without manual intervention, so that element positioning, rendering and follow-up editing are completed in the same technical action, the delay and inconsistency risk caused by frequent and direct operation of a real DOM in a traditional scheme is eliminated, and the operation efficiency is improved. The interaction fluency, the layout intuition and the system maintainability are obviously improved; by means of a virtual DOM mechanism of Vue2, batch diff is carried out and then submitted to a real DOM in a unified mode, the number of rearrangement and redrawing times of a browser is reduced, CPU and memory occupation is further reduced, and the canvas still keeps smooth response even in the scene with dense elements and frequent updating.

A data processing apparatus, a processor, a board card, and a data processing method. The data processing apparatus may be comprised by a combined processing apparatus (20), the combined processing apparatus (20) comprising a computing apparatus (201), an interface apparatus (202), a processing apparatus (203) and a storage apparatus (204). The computing apparatus (201) is configured to execute an operation specified by a user, so as to execute deep learning or machine learning calculation. The computing apparatus (201) may interact with the processing apparatus (203) by means of the interface apparatus (202), so as to jointly complete the operation specified by the user. The interface apparatus (202) is used to transmit data and a control instruction between the computing apparatus (201) and the processing apparatus (203). The storage apparatus (204) is used to store data of the computing apparatus (201) and the processing apparatus (203). The data processing apparatus provides a vector computation scheme integrating different types of fine-grained quantization, so that processing can be simplified, and the advantages of a low bit width operation and small memory occupation of a fine-grained quantization format are fully utilized.

The invention provides a long text abstract generation method and device, and belongs to the technical field of natural text processing, the method comprises the following steps: using Euclidean norm to carry out importance sorting on Tokens and carrying out compression storage according to a sparse rate, so that global key information is completely reserved and memory occupation is obviously reduced; in the decoding stage, local attention scores and global attention scores are calculated in parallel, entropy differences are mapped into fusion weights through Sigmoid by combining temperature adjusting parameters, and dynamic balance of local details and long-distance dependence is achieved. The local key value pairs and the global key value pairs are subjected to weighted integration based on the fusion weight, a continuous semantic spectrum is formed in a single decoding layer, splicing breakage caused by traditional partitioning is eliminated, the problems of input limitation and semantic splitting are effectively relieved, the context length capable of being processed by a model is expanded under the condition that the calculation amount is not remarkably increased, and the method has the advantages of being simple in structure and convenient to operate. And local and global context information is adaptively fused, so that the accuracy and continuity of the abstract are effectively improved.

The invention relates to the technical field of large model reasoning, in particular to a reasoning performance optimization method and system based on a large model, and the method comprises the following steps: initializing a reasoning performance optimization agent; collecting hardware environment indexes in real time, wherein the hardware environment indexes comprise a video memory utilization rate, a CPU (Central Processing Unit) exchange number, residual video card resources, storage IOPS (Input / Output Per Second) and network throughput; the method has the beneficial effects that related statistical indexes, including model types, model weight file total volume, model average sequence length, the number of tokens per second output by the model, first token time of the model, a display card list occupied by the model, the size of a KVcache block, the size of a KVcache sliding window and the like, of each model in a service system are comprehensively collected and analyzed; the system performance is comprehensively evaluated, and the defects that in the prior art, performance evaluation is not comprehensive, and a real-time monitoring mechanism for key indexes such as memory occupation and network bandwidth is lacked are overcome.

The invention belongs to the technical field of edge computing, and discloses a lightweight AI model adaptive deployment system oriented to edge computing. A dependency intensity matrix is constructed by accurately quantifying a data dependency relationship between model layers, network stability is monitored in real time, an abnormal time period is identified, a communication penalty factor and a bandwidth attenuation factor are calculated based on time delay peak distribution characteristics, and then a communication overhead prediction model for cooperative calculation between edge devices is constructed. The system innovatively generates a multi-granularity model segmentation candidate scheme set, identifies a segmentation boundary triggering resource competition by monitoring and calculating load fluctuation and memory occupancy change, and dynamically adjusts a model partition granularity and a mapping strategy according to a gradient transmission quantity and an activation value size at the segmentation boundary. According to the invention, the overhead of cross-device communication is reduced, the utilization balance of computing resources is improved, and the real-time response capability in a time delay sensitive scene is ensured.

The invention relates to the technical field of heat supply network metering supervision, and discloses a heat supply network metering data supervision system based on edge calculation. The system comprises an edge data acquisition module, a metering task management module, an edge resource scheduling module and a heat supply network supervision execution module. The edge data acquisition module acquires original data from a heat supply network metering sensor in real time, and sends the original data to the metering task management module after verification and format conversion; the latter receives the data, generates and stores a metering task record according to the type and the supervision rule, and updates the state to be to-be-processed; the edge resource scheduling module queries a to-be-processed task, monitors the CPU utilization rate, the memory occupancy rate and the network delay of an edge node, determines an execution sequence in combination with the number of tasks and generates a resource allocation scheme; and the heat supply network supervision execution module activates a corresponding node, obtains a task record, executes data analysis and supervision logic processing, and stores a result. The system optimizes data processing and resource configuration by means of edge computing, and adapts to heat supply network metering data supervision requirements.

The invention discloses a file pre-reading method and device, a medium and equipment, and the method comprises the steps: firstly obtaining a file type of a target file, and executing pre-reading processing through a pre-reading window only when the target file is a continuous file; and dynamically adjusting the size of the window in combination with the hit condition of the pre-reading content on the basis: increasing the window in hit to improve the sequential reading efficiency, and reducing the window in miss to reduce the invalid pre-reading proportion. Through the above self-adaptive adjustment mechanism based on hit feedback, the size of the pre-reading window can be flexibly optimized for different access modes, and excessive pre-reading caused by a fixed pre-reading window in a random access or access mode change scene is avoided, so that unnecessary memory occupation and input and output resource waste are reduced, and the stability of reading performance is kept.

The invention discloses a large model fine tuning optimization method based on multi-strategy fusion, which comprises the following steps: designing a dynamic parameter selection mechanism, adaptively determining a parameter subset needing fine tuning according to a task demand and a model structure, and reducing unnecessary parameter updating calculation; constructing a dynamic low-rank decomposition framework, dynamically adjusting the rank of a low-rank matrix according to a model training state and data characteristics, and keeping key information while compressing a parameter scale; a self-adaptive task sensing mechanism is introduced, a fine adjustment strategy is automatically adjusted according to different task characteristics, and the adaptability of the model to various tasks is improved; and a mixed precision training method is adopted, so that the calculation complexity and the memory occupation are reduced on the premise of ensuring the model precision. According to the method, a parameter efficient fine tuning technology and a dynamic low-rank decomposition strategy are innovatively combined, and an adaptive task perception mechanism and a mixed precision training technology are introduced, so that the operand and resource requirements of model training are effectively reduced, and the fine tuning efficiency and the model performance are improved.

The invention discloses an Internet of Things device anti-quantum password migration method based on a national cryptographic algorithm, and relates to the field of data security, the method comprises the following steps: configuring a Hash agility module taking a national cryptographic SM3 as a default algorithm, and dynamically switching to an anti-quantum or compatibility alternative algorithm based on monitoring; carrying out identity authentication and temporary key negotiation by adopting SM2, and generating a key pair only used for a single session; deriving a session key based on the key pair, and adopting a differential double-layer signature verification mechanism according to the data security level; for firmware updating, a differential package is generated through function level differential analysis, and atomic updating and rollback are achieved through A / B system partition. According to the method, quantum computing threats are effectively resisted, national security compliance and forward security are ensured, the volume of an update package is remarkably reduced through a resource optimization technology, memory occupation is reduced, verification efficiency is improved, and the method is particularly suitable for resource-limited Internet of Things equipment.

The invention relates to the technical field of computer vision, and discloses an industrial product visual defect detection method and system based on multi-task reverse knowledge distillation, by introducing a multi-scale projection layer and a multi-task learning framework, the expression ability of a model to normal sample features and the inhibition ability to abnormal signals are significantly improved, and the detection accuracy is improved. Therefore, industry-leading defect detection and positioning precision is realized on industrial reference data sets such as MVTec, extremely low reasoning delay and memory occupancy are kept, the reasoning speed is far higher than that of mainstream methods such as PatchCore, an efficient and accurate technical scheme is provided for industrial large-scale real-time defect detection, low delay and low memory consumption are kept, and the defect detection accuracy is greatly improved. The abnormal detection and positioning precision of industrial product visual defect detection is remarkably improved, and the defect that performance and efficiency are difficult to consider in the prior art is overcome.

The embodiment of the invention provides a flash memory processing failure processing method and system, computer equipment and a storage medium, and relates to the technical field of flash memory process.The method comprises the steps that the write operation return state of first flash memory data written into a first physical block is obtained; when the write operation return state is a failure state, obtaining first flash memory data in the internal cache of the flash memory, storing the first flash memory data in the main control cache, and erasing the first flash memory data in the internal cache of the flash memory; and writing the first flash memory data stored in the main control cache into a second physical block, wherein the second physical block is a healthy physical block. According to the embodiment of the invention, the write failure of the flash memory can be processed, the data security of the flash memory is ensured, and a master control buffer does not need to be additionally occupied. And when writing failure occurs, reading buffer data cached in the flash memory, and re-writing the buffer data into the healthy flash memory physical block. According to the method, the flash memory writing failure can be effectively processed, the data security is ensured, the occupation of a main control memory buffer of the flash memory can be reduced, and the cost is reduced.

The invention belongs to the technical field of artificial intelligence and edge computing, particularly relates to an industrial edge-oriented lightweight AI model dynamic loading system, and aims to solve the problems of rigid model deployment, high reasoning delay and update lag caused by resource limitation and variable working conditions of industrial edge equipment. According to the system, an AI model is decoupled into function fragments with attribute tags through a model fragment manager, a configuration strategy is generated by combining a context sensing adaptation module with an equipment state and a task requirement, a scheduling engine determines an optimal fragment combination according to the configuration strategy, an incremental loading executor realizes millisecond-level hot plugging through symbol analysis and a double-buffering mechanism, and the configuration strategy and the scheduling engine are combined. The localized fine tuning unit optimizes key parameters online by using a small amount of data. According to the scheme, loading on demand, dynamic adaptation and non-inductive upgrading are supported, memory occupation and energy consumption are remarkably reduced, the response speed and the model generalization ability are improved, and efficient autonomy of edge intelligent services is achieved.

The invention relates to an LBMS underlying software architecture based on AUTOSAR, and belongs to the technical field of automobile controllers, and the software architecture comprises a management module which comprises a target Os, a target EcuM, a target CanTp, a target Com, a target CanNm, a target NvM, a target WdgM and a target E2e which are redesigned based on LBMS requirements; the target E2e is also used for realizing the verification function of the Crc. According to the software architecture provided by the invention, the standard AUTOSAR underlying software architecture is simplified, nine MCAL modules and three complex driving modules are inherited, Dcm and Dem are statically reserved and dynamically optimized, eight management modules are redesigned according to LBSM requirements, seven non-essential modules are directly deleted, and the modules are further subjected to function integration, so that the number of the modules is reduced; therefore, the FLASH memory occupation of the MCU is obviously reduced, and the system performance is optimized.

The invention relates to the technical field of simulation training application integration frameworks based on micro-service architecture, and discloses a simulation training application integration framework method and system based on micro-service architecture. According to the simulation training application integration framework method and system based on the micro-service architecture provided by the invention, a simulation training process is divided into a preparation stage, an execution stage and an evaluation stage, and independent data storage spaces and communication rules are respectively configured, so that the isolation and controllability of task processing of each stage are remarkably improved; and inter-stage data interference is avoided. A staged resource use monitoring mechanism is introduced, a dynamic capacity expansion and contraction strategy based on service request changes, the CPU utilization rate and the memory occupancy rate is achieved, the system resource utilization rate is effectively improved, and the operation stability is guaranteed. By setting copy health detection and copy switching logic, fault copy elimination and automatic taking-over of healthy copies are realized, and the fault-tolerant capability and service continuity of the system are enhanced.

A system method for performing distributed training of models. The method includes co-optimizing memory optimizations with parallelism to increase model training throughput under a memory constraint by orchestrating a plurality of optimizations to utilize system resources in consideration of computation, communication and memory footprint. The system can include an overlap-centric schedule template that determines granularity and order of how techniques utilized by the plurality of optimizations are applied to a model. The overlap-centric schedule template mitigates tuning complexity by applying heuristics to orchestrate optimizations in an overlapped manner.

The invention provides an applet on-demand dynamic loading and subpackaging method based on a unified subpackaging mode, relates to the technical field of mobile application development, and comprises the step of realizing on-demand loading and page rendering by establishing standardized configuration information, a subpackaging resource list, a subpackaging life cycle state machine and a subpackaging dependency graph and combining a subpackaging runtime manager. According to the method, the initialization loading time can be shortened, the memory occupation can be optimized, the small program starting performance can be improved, and meanwhile, the development complexity can be reduced and the cross-platform compatibility can be improved through a standardized subpackage mechanism.

The present disclosure relates to techniques for improving inference efficiency and memory utilization in transformer-based neural networks. A modified architectural design is introduced that decouples key and value matrix generation from inter-layer dependencies, enabling statically computed or parallelizable projections across layers. The disclosed approach may eliminate the need for layer-wise prefilling, support linear-time inference, and substantially reduce the memory footprint associated with key-value (KV) caching. The disclosed architecture may use shared or layer-specific projections, with a single KV-cache serving all or subsets of layers. In some embodiments, a non-linear transformation (e.g., implemented via a feed-forward network), may preprocess input embeddings prior to query, key, and value generation. A lookup table of transformed embeddings may be precomputed to further accelerate inference. The disclosed system can enhance scalability and may allow deployment of large models on resource-constrained hardware, offering practical benefits for latency-sensitive applications and long-context processing in transformer-based models.

The embodiment of the invention provides a memory leak detection method and device, equipment and a storage medium, and relates to the technical field of operating systems and performance tuning. The method comprises the following steps: monitoring memory allocation and release operation of a process by adopting a probe program; collecting memory use information of a stack of the process on memory allocation and release operation; through a collaborative caching mechanism of a kernel mode and a user mode, sharing memory use information to the user mode, and analyzing the memory use information in the user mode to obtain a comprehensive hotspot score representing a memory occupation growth trend of a stack; and determining whether the stack has memory leak according to the comprehensive hotspot score. Based on the method, whether the memory leaks or not can be accurately detected under the condition of low performance overhead.

The invention relates to the technical field of artificial intelligence video processing and hardware acceleration, and discloses an NPU (Network Processing Unit) high-concurrency reasoning method and system based on single-pipeline multi-channel video stream batch processing, which can avoid frequent memory copying and CPU (Central Processing Unit) participation by carrying out dynamic batch merging processing on multi-channel video streams at an NPU end. The data transmission and model loading times, memory occupation and loading delay are reduced; according to the technical scheme, 32 paths of video input can be supported through a single reasoning pipeline, and compared with a multi-pipeline scheme, the model does not need to be loaded repeatedly, so that the model loading times and the memory occupation are greatly reduced, the concurrency capability and the throughput rate are remarkably improved, cutoff fault tolerance and dynamic batch size adjustment are supported, and the stability is improved. According to the method, the processing efficiency and the resource utilization rate of the system can be improved while low delay is kept, and the method is suitable for multi-channel video AI analysis scenes such as video monitoring, intelligent transportation and security and protection.

The invention discloses a re-calculation opportunity selection method and device in large model training, and belongs to a large model distributed training optimization technology in the cross field of artificial intelligence and high-performance calculation. The method comprises the following steps: dynamically predicting the influence of pipeline bubbles and memory usage by monitoring calculation, communication tasks and memory usage states in training iteration in real time and utilizing a pre-trained machine learning model, and judging the opportunity of inserting a re-calculation operation. According to the method, the optimal recalculation triggering opportunity is intelligently output by taking large model training state data as input and taking iteration delay minimization as a target. The device comprises a state monitoring module, an intelligent decision-making module and a feature engineering processing and executing module which are used for realizing the method. According to the method, a static rule is upgraded into an intelligent decision, calculation and communication gaps can be utilized more accurately, calculation, communication and memory resources are balanced in a self-adaptive mode, and therefore the throughput and efficiency of large model training are remarkably improved.

A system, method, and computer-program product includes receiving a plurality of vector embeddings having an initial dimensionality and projecting the plurality of vector embeddings into lower-dimensional spaces using at least two different dimension reduction algorithms to generate corresponding sets of projected vector embeddings. Each set of projected embeddings may be quantized and nearest neighbors for the original embeddings and for each quantized set of projected embeddings may be calculated. Additionally, a neighbor preservation metric may be evaluated for each quantized set by comparing its nearest neighbors to those of the original embeddings. Based on the neighbor preservation metrics and a predefined error tolerance, an optimal compression configuration may be selected.

The application provides a webpage application data processing method and device, equipment and readable storage medium; the method comprises the following steps: sending the path of a dynamic link library to a browser process in a browser embedded framework through a host process, so as to trigger the browser process to create a webpage application; creating a rendering process through the browser process, wherein the rendering process is used for rendering a data item in the webpage application; calling an extension plug-in in the browser embedded framework through the rendering process to perform the following processing: loading the dynamic link library according to the path, and starting a calculation thread; sending the data item to the calculation thread through the browser process, so as to trigger the calculation thread to obtain the calculation result of the data item through a calculation function in the dynamic link library; sending the calculation result to the rendering process through the browser process, so as to trigger the rendering process to display the calculation result in the webpage application. Through the application, the data processing efficiency of the webpage application can be improved and the memory occupation can be reduced.

This application relates to the field of agricultural Internet of Things (IoT), and particularly to a method, apparatus, device, and program product for calculating vegetation indices. The method includes: receiving a user-configured vegetation index calculation task, wherein the calculation task includes at least one vegetation index to be calculated; based on a preset dependency relationship between vegetation indices and spectral channels, and combined with the vegetation indices included in the calculation task, parsing to obtain the necessary set of spectral channels required for the calculation task; acquiring the necessary spectral data corresponding to the necessary set of spectral channels, and caching the image data of the current processing row and adjacent rows in the necessary spectral data; calculating the vegetation index in the calculation task row by row based on the cached image data; and when the calculation of a single row of vegetation indices is completed, outputting the single row of vegetation indices and releasing the cached data related to the calculation of the single row of vegetation indices. This method can significantly reduce the memory usage of vegetation index calculation and improve memory utilization efficiency.

This invention relates to the field of information processing technology, and in particular to a gridded interpolation method, apparatus, electronic device, and storage medium that integrates KNN search. It employs a KD-tree with contiguous memory storage and aggregated leaf node storage, which helps reduce memory usage and tree depth. Combined with an improved fast selection algorithm and iterative construction, it can increase tree construction speed and mitigate the risk of recursive stack overflow. By integrating KNN search with heap sort, the nearest neighbor is directly obtained during traversal. Compared to the two-stage method of querying first and then sorting, this reduces intermediate data storage and secondary sorting overhead. This invention improves interpolation speed and reduces memory usage when processing gridded interpolation of millions of scattered points. In typical browser scenarios, it can achieve millisecond-level response times, making it suitable for real-time visualization and rapid analysis in fields such as meteorology, oceanography, and the environment.

The invention relates to the technical field of data testing and the field of smart medical treatment and financial science and technology, and provides a load balancing testing method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: generating a load balancing testing request based on a testing module of a client; forwarding the load balancing test request to a load balancing proxy node based on a switch; determining a target test script from the load balancing agent node according to the load balancing test request; performing data forwarding processing on the client and the communication test instrument based on the load balancing agent node according to the target test script, and detecting on the load balancing agent node to obtain a CPU utilization rate, a memory occupancy rate, connection pool state information and request processing delay information; and determining a load balancing test result according to the CPU utilization rate, the memory occupancy rate, the connection pool state information and the request processing delay information. Through the technical scheme, comprehensive and reliable test processing can be performed on the load balancing server simply, conveniently and quickly.