23results about How to "Reduced cold start time" patented technology

The invention discloses an inference system, an inference method, an inference device, a storage medium and a program product, and the inference system comprises a large language model parameter warehouse which is used for providing downloading services of shared parameters and fine tuning parameters corresponding to various large language models, the shared parameters and the fine tuning parameters are parameters in a corresponding large language model which are split into pre-training parameters shared by multiple users and parameters for fine tuning of the shared parameters according to user requirements based on a LoRA large language model fine tuning normal form; the resource pools comprise a plurality of reasoning nodes, and the reasoning nodes serve as execution nodes when being loaded with the shared parameters and the fine tuning parameters; and the control node is used for scheduling an execution node capable of executing the reasoning task corresponding to the reasoning request from the plurality of resource pools as an actual execution node, and sending the reasoning task to the actual execution node, so that the actual execution node instantiates the reasoning task into a large language model according to the input data carried in the reasoning task to execute the reasoning task.

The invention relates to an AEM water electrolysis hydrogen production system and a control method, and the system comprises an electrolytic cell module, a water tank module, a power supply module, a circulation module, a heat balance module, a purging module, a control module, a communication module and a display module. The functions of water replenishing of the water tank, direct current setting of the electrolytic bath, electrolyte circulation, electrolyte heat balance, startup purging, shutdown purging and the like are realized; the cold start time is shortened by controlling the technological process, and the hydrogen production efficiency is effectively improved; the service life of the electrolytic bath is effectively guaranteed by controlling the constant current output of the controllable direct-current power supply; safety production is effectively guaranteed through safety measures such as automatic purging of a triggering system; and the anti-risk capability is effectively improved by monitoring the running state of the equipment in real time and giving an alarm prompt.

The utility model discloses an intake preheating system of a methanol engine, the methanol engine and a vehicle, the intake preheating system comprises a methanol heater and an intake pipe, and a heated air outlet of the methanol heater is connected with an air inlet of the intake pipe through a heater throttle valve; an air outlet of the air filter is connected with an air inlet of the air inlet pipe through a throttle valve; according to the utility model, when the engine is in cold start, hot air heated by the heater is continuously input and directly enters the air inlet header pipe of the engine, so that the temperature after the hot air is mixed with fresh inlet air is increased, the air participating in ignition is hot air, and the success of low-temperature start of the methanol engine is further realized.

The invention provides a novel compressed air and hydrogen energy storage coupling system and a working method thereof. The system comprises an air and hydrogen coupled energy storage circulating system, an air and hydrogen coupled energy release circulating system and a waste heat multi-stage recycling heat storage circulating system. The energy storage cycle of the coupling system converts electric energy into hydrogen energy and pressure energy of air for storage, the energy release cycle converts the pressure energy and the hydrogen energy of the air into electric energy for release, the heat storage cycle system recycles and utilizes heat in the system operation process in a multi-stage mode through heat conduction oil and cooling water, and tight coupling of compressed air energy storage and hydrogen energy storage is achieved. According to the coupling system, cooling water and heat conduction oil are used as heat exchange media, heat energy generated in the energy storage process of the system is recycled in a multi-stage mode, the stored heat energy is used for heating air and hydrogen in the energy release process, and therefore the power capability of high-pressure air and hydrogen is improved. The system mixes the exhaust gas of the air expansion machine with oxygen, and the mixed gas is depressurized by the pressure stabilizing valve and then supplied to the hydrogen fuel cell, so that the oxygen concentration in the mixed gas is increased, and the power generation performance of the hydrogen fuel cell can be effectively improved.

The invention provides a system and a method for optimizing the cold start time of a. NET Serverless function. The system comprises a dependency analysis module, an AOT compiling optimization module, a dynamic code optimization module, a program set loading optimization module, a type system optimization module, a memory management optimization module, a state storage and recovery module and an adaptive optimization system. Dependence analysis is carried out through static analysis and dynamic tracking, partial AOT compiling and self-adaptive JIT compiling are achieved, program set loading and a type system are optimized, a special memory allocation strategy and lightweight state serialization are achieved, and the strategy is continuously monitored and dynamically adjusted and optimized. The method further comprises the steps of asynchronous operation optimization and security assurance. The method has the advantages that the cold start time of the. NET Serverless function is remarkably shortened, the function execution efficiency is improved, meanwhile, cross-platform compatibility and safety are guaranteed, and a more efficient solution is provided for Serverless calculation.

The embodiment of the invention discloses an image display method and device based on auxiliary rendering, computer equipment and a medium, and the method comprises the steps: carrying out the construction of a display function region, so as to obtain a video memory rendering buffer region and a frame buffer region, constructing a display algorithm operation space according to the video memory rendering buffer area and the frame buffer area, and binding the video memory rendering buffer area and the frame buffer area according to the rendering base point; generating a first display image in the video memory rendering buffer area according to a pre-stored first image rendering strategy; mapping the first display image to a frame buffer area; performing auxiliary rendering on the first display image according to a pre-stored second image rendering strategy to obtain a corresponding second display image; and mapping the second display image to an image real-time display screen of the image processing display system based on the image memory address of the second display image. According to the method, through an auxiliary rendering strategy, rapid image generation and mapping are realized, a blocking phenomenon in a traditional process is avoided, and the image display speed is further accelerated.

The invention provides a Java Serverless function cold start optimization system and a Java Serverless function cold start optimization method based on predictive class loading and code caching. The system comprises an intelligent class analysis and prediction module, a hierarchical class loading system, a fine-grained code caching module, an asynchronous preheating mechanism, a state storage and recovery module and a self-adaptive optimization module. The system predicts a required class set through static analysis and dynamic tracking, optimizes a class loading process by utilizing a hierarchical loading strategy, improves code execution efficiency by adopting fine-grained cache, and reduces starting delay through asynchronous preheating and rapid state recovery. Meanwhile, the system continuously monitors performance indexes and dynamically adjusts strategies to achieve self-adaptive optimization. According to the method, the function cold start time is remarkably shortened, the function response speed and the resource utilization efficiency are improved, meanwhile, the dynamic characteristics and safety of Java are kept, and an efficient performance optimization solution is provided for Java application in the Serverless computing environment.

The present application relates to a kind of preparation methods of essence.The problems that traditional process parameter adjustment relies on artificial experience and is difficult to realize cross batch self-adaption, multi-modal feature extraction, parameter-function-stage three-dimensional label matrix generation, graph neural network behavior atlas construction and expert strategy distribution constraint dual-channel strategy learning method are proposed.By process parameter and derived feature multidimensional fusion, time series causal strength analysis and action instruction simulation verification, optimization process parameter combination with explainable function synergistic relationship can be output.The technical scheme effectively realizes the automatic optimization of process control decision and the rapid self-adaption of model under new raw material batch, improves product consistency, quality stability and control energy efficiency.

The invention provides an optimization method and device for low-temperature cold start of a fuel cell stack and electronic equipment. The method comprises the following steps: determining a low-temperature cold start model based on internal parameters of a fuel cell stack; determining the initial membrane water content and the segmented step current as optimization variables; determining the cold start time of the low-temperature cold start model based on the optimization variable and the low-temperature cold start model, and taking the cold start time as an optimization target; based on the optimization variable and the optimization target, generating a particle swarm by adopting a particle swarm algorithm, and performing iterative global optimization in the particle swarm; and acquiring a local optimal solution of any iteration, and performing next iteration based on the local optimal solution and the low-temperature cold start model until the iteration is finished to obtain a global optimal solution meeting the optimization constraint condition. Therefore, according to the scheme, the cold start time is taken as the optimization target, and the optimization constraint condition of the cold start of the model is established, so that the cold start time can be shortened on the premise of ensuring that the low-temperature cold start of the electric pile is successful.

The invention discloses a container mirror image preloading method and device based on demand prediction, equipment, a medium and a product, and the method comprises the steps: collecting a current node state, a task execution state and mirror image information in real time, and carrying out the data preprocessing; inputting the preprocessed data into a preset hybrid prediction model, and outputting the demand probability of each node for each container mirror image in a future time window; the hybrid prediction model comprises a time sequence feature extraction network used for learning a time sequence rule and an attention coding network used for learning global relation features; determining a to-be-loaded container mirror image and a target node according to the demand probability and the resource constraint of the node; and preloading the to-be-loaded container mirror image to the target node. According to the method, the demand probability of the container mirror image on the node is predicted, only the necessary container mirror image is preloaded, the cold start time of the container is shortened, the resource utilization rate of the node is optimized, the consumption of the storage space and the network bandwidth is reduced, and the operation cost is effectively reduced.

The invention relates to the technical field of cluster management, in particular to a container group creation method and device, computer equipment and a readable storage medium. The method comprises the steps of selecting a target sandbox container environment from existing container sandbox environments of resource nodes in a cluster according to a Pod creation request under the condition that the Pod creation request of a container group is obtained; the existing container sandbox environment of each resource node is pre-established according to a sandbox environment establishment request sent by the scheduler to the resource node; and in the target sandbox container environment, creating a target Pod corresponding to the Pod creation request.

The invention relates to a vehicle-mounted carbon-free fuel mixing proportional valve with a preheating function, and belongs to the technical field of new energy automobile fuel cells. According to the technical key points, the front valve component comprises a front valve body and an ammonia water valve; an upper fluid cavity, a lower fluid cavity and a confluence cavity are formed in the front valve body; the upper fluid cavity and the lower fluid cavity are arranged in parallel up and down; the rear valve component comprises a rear valve body and a main valve; a protruding part is arranged at the left end of the rear valve body, and external threads are arranged on the outer surface of the protruding part. The protruding part of the rear valve body is in threaded connection with the interior of the inserting opening of the front valve body, and the confluence cavity is correspondingly communicated with the flowing hole channel. By adopting the scheme provided by the invention, the hydrogen fuel cell can be started more quickly in a low-temperature environment, and the cold start time is shortened.

The invention provides a system and a method for optimizing the cold start time of a Node.js server function. The system comprises a dependency analysis module, a module loading optimizer, a code optimization engine, a V8 engine optimizer, an asynchronous operation scheduler, a state manager and an adaptive optimization system. The system performs dependency analysis through static analysis and dynamic tracking, realizes module preloading and lazy loading, pre-compiles key code paths, optimizes V8 engine parameters and heap memory usage, improves asynchronous operation processing, realizes lightweight state management, and continuously monitors performance to dynamically adjust an optimization strategy. According to the method, the cold start time of the Node.js serverless function can be remarkably shortened, the execution efficiency is improved, and the resource consumption is reduced, so that the Node.js application performance and the user experience in the serverless environment are improved.

The invention provides a panoramic image display method and device, a vehicle and a storage medium. The method comprises the steps that a state machine thread and an image processing thread of a panoramic image process are initialized; if the state of the panoramic image application is an activated state, initializing a panoramic image display window and creating a display carrier through the state machine thread, and initializing a shader through the image processing thread; and processing the display carrier through the image processing thread to create a rendering surface, processing image data, rendering the image data to the rendering surface, and displaying a panoramic image. In the embodiment of the invention, different tasks in the panoramic image display process are distributed to independent threads, and the threads process the tasks in a parallel manner, so that the operation efficiency and response speed of the panoramic image application are improved, the cold start time of the panoramic image application is shortened, and the user experience is improved.

The utility model provides a water electrolysis hydrogen production system, and relates to the technical field of electrolysis hydrogen production. The system comprises a renewable energy source, an electrolytic cell, a solar heat collector, a heat exchanger and an electrolyte tank, the renewable energy source is electrically connected with the electrolytic cell, the solar heat collector is connected with the heat exchanger, the heat exchanger is connected with the electrolyte tank, and the electrolyte tank is connected with the electrolytic cell. According to the water electrolysis hydrogen production system disclosed by the utility model, the solar heat collector is additionally arranged, and the temperature of the electrolyte in the electrolyte tank is controlled through the heat exchanger, so that the temperature drop of the electrolyte in the electrolyte tank and the long-time temperature rise operation caused by the temperature drop when the electrolytic cell is restarted are avoided, and the cold start time of the electrolytic cell is further shortened; the electrolysis efficiency is improved.

The utility model relates to a megawatt anion exchange membrane water electrolysis hydrogen production integrated system which comprises an electrolytic bath subsystem comprising a plurality of large standard AEM electrolytic baths and is used for water electrolysis hydrogen production. The electrolyte circulation subsystem is used for providing electrolyte and maintaining circulation; the oxygen treatment subsystem is used for drying and dehydrogenating oxygen; the drying and purifying subsystem is used for drying, deoxidizing and purifying the hydrogen; the purging subsystem is used for purging the residual gas; the power supply and distribution subsystem is used for supplying power; the heat dissipation subsystem is used for cooling and heat dissipation; and the centralized control subsystem is used for controlling the whole integrated system. According to the utility model, the centralized control subsystem is adopted to realize remote control and real-time monitoring on the whole integrated system, so that the whole integrated system is in an optimal operation state, and the efficiency and the dynamic response rate are effectively improved; according to the utility model, the purging system is adopted to purge residual gas, so that the safety of the system is improved; and the low-concentration alkaline liquid is adopted, and hydrogen is removed from oxygen, so that the environmental protection property is effectively improved.

This invention discloses a container image preloading method, apparatus, device, medium, and product based on demand prediction. The method includes: real-time acquisition of current node status, task execution status, and image information, and performing data preprocessing; inputting the preprocessed data into a preset hybrid prediction model to output the demand probability of each node for each container image within a future time window; the hybrid prediction model includes a time-series feature extraction network for learning time-series patterns and an attention encoding network for learning global relationship features; determining the container image to be loaded and the target node based on the demand probability and the node's resource constraints; and preloading the container image to be loaded onto the target node. This invention, by predicting the demand probability of container images on a node, preloads only the necessary container images, shortening container cold start time while optimizing node resource utilization, reducing storage space and network bandwidth consumption, thereby effectively reducing operating costs.

This invention discloses a starting control method, system, equipment, aircraft, and medium for a propeller-driven aircraft piston engine, relating to the field of aero-engine control technology. The method includes: performing a power-on self-test before starting, checking the minimum propeller blade angle, throttle opening, and fuel pressure; after the self-test, the flight control computer automatically and adaptively controls the starting throttle opening and target starting speed based on altitude and atmospheric static temperature; during starting, status monitoring is performed, towing the engine to a state capable of ignition and fuel injection control; after successful starting, the lubricating oil temperature is determined to automatically set the corresponding throttle opening and warm-up speed; finally, a dual ignition system check and pre-flight condition checks are performed. This invention achieves fully closed-loop adaptive control, effectively overcoming starting difficulties in complex environments such as high altitudes and frigid conditions, eliminating the hidden dangers of starting with defects, and significantly improving the timeliness of the aircraft's maneuverability and operational safety.

The invention provides a lightweight function operation method and system oriented to a server perception-free computing scene. The method comprises the following steps: providing a communication primitive for rapid cold start and high-performance intermediate data transmission of an isolated environment; for a workflow scene without perceptual function calculation of a server, providing a novel software stack architecture during operation; a function and a bottom layer on which the function depends during operation are all deployed in the same CPU privilege level and address space for cold start operation; a library operating system of workflow granularity is used for providing bottom layer runtime service for functions, and meanwhile, a memory security language is used for providing memory isolation and sharing; and realizing an on-demand loading mechanism for the library operating system. The technical problems of slow cold start in an isolation environment, poor intermediate data transmission performance, large cold start range, high cost and poor memory isolation and sharing flexibility are solved.

This invention proposes a causally explainable anomaly analysis and root cause tracing method, apparatus, and readable storage medium, aiming to solve the problems of poor anomaly identification stability, root cause localization bias, and insufficient diagnostic interpretability caused by non-stationary data, small sample sizes, and dynamic evolution of process logic in rapid batch production changeover scenarios. This invention collects and preprocesses multi-source heterogeneous data, monitors production indicators based on adaptive thresholds and extracts anomaly window feature data, migrates the initial causal structure from historical batches based on multi-dimensional similarity of product model, process route, and anomaly pattern, and performs dynamic attribution updates combined with online data. Based on a causal graph, it applies counterfactual perturbations to candidate factors to calculate causal contribution and searches backward along the causal chain to pinpoint the root cause, mapping the inference path to physical semantics to generate a diagnostic report. This invention is applicable to anomaly diagnosis in multi-product, rapid-change flexible production lines.

The invention provides a code space mounting method and a code space mounting system for a Serverless framework, which are used for solving the problem of function cold start delay in the traditional Serverless framework. The method is characterized in that a code space prepared in advance is maintained for each user or application, and the code space is stored in a high-performance and low-delay distributed storage system. And when the function needs to be executed, the system directly mounts the corresponding code space on the function instance, so that the function instance can directly read and execute the code from the mounted code space. The invention further comprises a safe multi-tenant isolation, code space version management and updating system and a performance optimization and consistency guarantee mechanism. The method and the system have the beneficial effects that the cold start time of the function is remarkably shortened, the resource utilization efficiency is improved, and the safety and the flexibility of the system are enhanced, so that the overall performance and the user experience of the Serverless architecture are greatly improved.

The utility model relates to the technical field of vehicle chassis, and provides an exhaust catalyst and a vehicle. The exhaust catalytic converter is arranged in an exhaust system connected with an engine and comprises a catalytic converter body with a catalytic unit and an energy storage assembly arranged on the catalytic converter body. Wherein a phase change material and a heat exchange pipeline are arranged in the energy storage assembly, the heat exchange pipeline is connected with a gas outlet pipeline of the exhaust catalyst, meanwhile, when the engine works, the heat exchange pipeline can introduce gas in the gas outlet pipeline to heat the phase change material, and when the engine stops, the phase change material can release heat to heat the catalytic unit. By means of the heat storage and heat release functions of the energy storage assembly, the catalytic unit can be maintained at a certain temperature when the engine stops, the catalytic unit can reach the ignition temperature as soon as possible when the engine is in cold start, the cold start time of the engine can be shortened, emission exceeding during cold start can be reduced or even avoided, and the service life of the engine is prolonged. And the use quality of the vehicle can be improved.

The invention relates to the technical field of station reactive power response, in particular to an Internet of Things edge device scheduling system and method based on AI reasoning hardware resources, and the system comprises an edge system module, a cloud storage module, an ML model data module and a reasoning data module. A computer cluster, an artificial intelligence reasoning service unit and a local storage unit are integrated in the edge system module; the edge system module, the cloud storage module, the ML model data module and the reasoning data module communicate with one another through the network service unit; wherein the edge system module is used for receiving Internet of Things equipment data at an edge side and executing an AI reasoning task, and providing a reasoning result for an upstream application with relatively low delay; and the cloud storage module is used for storing model backup, historical reasoning data and logs and providing cloud reasoning rollback capability. According to the invention, through intelligent sensing and scheduling of heterogeneous edge hardware, efficient deployment of the ML model in the resource-constrained IoT environment is realized.