354results about "Program saving/restoring" patented technology

An online system trains a transformer architecture by an initialization method which allows the transformer architecture to be trained without normalization layers of learning rate warmup, resulting in significant improvements in computational efficiency for transformer architectures. Specifically, an attention block included in an encoder or a decoder of the transformer architecture generates the set of attention representations by applying a key matrix to the input key, a query matrix to the input query, a value matrix to the input value to generate an output, and applying an output matrix to the output to generate the set of attention representations. The initialization method may be performed by scaling the parameters of the value matrix and the output matrix with a factor that is inverse to a number of the set of encoders or a number of the set of decoders.

The embodiment of the invention provides a process freezing method and device, relates to the field of terminals, and can reduce resource consumption of background processes and improve performance and endurance of electronic equipment. The method is applied to the electronic equipment and comprises the steps of creating a target thread in response to that the electronic equipment is in a first scene; determining a to-be-frozen process based on the target thread; and the process to be frozen is set to be in an uninterruptible sleep state based on the target thread, and the process to be frozen is set to be in an uninterruptible sleep state based on the target thread.

The invention discloses a graphics processor thread mid-term preemption. Techniques are provided that enable thread mid-term (instruction level) preemption in a graphics processor without requiring software intervention by a graphics driver associated with the graphics processor. Execution of a kernel program by a graphics processor is triggered using threads of the graphics processor to dispatch hardware to facilitate hardware-based thread mid-term preemption, the graphics processor holding thread states of preempted processing resources and facilitating subsequent recovery of the thread states to the same or different sets of processing resources.

This disclosure describes a large language model (LLM)-based agent operating system comprising an application layer, a kernel layer, and a hardware layer. The kernel layer comprises an AIOS kernel comprising: LLM core(s), an agent scheduler, a context manager, a memory manager, a storage manager, and a system call interface configured to manage interactions among the agent scheduler, the context manager, the memory manager, and the storage manager.

The invention is suitable for the technical field of terminals, and particularly relates to a memory management method, electronic equipment and a computer readable storage medium. According to the method, when it is detected that the first application is switched to the background state, the electronic device can determine that the first application does not need to display an image and does not need to carry out graphic rendering, that is, it can be determined that the first application does not need to use a GPU physical memory currently. At the moment, the electronic equipment can store the first content stored in the first GPU physical memory corresponding to the first application to a disk, and can release the first GPU physical memory, so that the occupation of the GPU physical memory is reduced, the remaining available physical memory of the electronic equipment can be increased, the response speed of the electronic equipment to user operation is increased, the fluency of the system is improved, and the user experience is improved. And user experience is improved.

The invention provides a virtual machine crashing event processing method and device and a storage medium. According to the method, when a crashing event occurs in a second virtual machine facing a hardware layer of the electronic equipment, a first virtual machine which does not generate the crashing event and bears an operating system temporarily takes over resources occupied by the second virtual machine, the second virtual machine is restarted independently, and after the second virtual machine is restarted, the first virtual machine does not generate the crashing event and bears the operating system. And returning the resources of the second virtual machine taken over by the first virtual machine to the second virtual machine. Therefore, only the second virtual machine with the crashing event is self-recovered, and the electronic equipment is not triggered to restart the whole machine, so that the user perception is reduced, and the user experience is improved.

The embodiment of the invention discloses a thread priority adjustment method and related equipment which are applied to electronic equipment, and the electronic equipment runs a first application. A first process of the first application comprises a plurality of first threads; identifying an associated thread of the first thread from the threads of the second process; the second process is a process which is operated by the electronic equipment, except the first process and the resident process and provides service for the first application; when it is detected that the running state of the first application is changed, the priority of the first thread is adjusted, the priority of the associated thread is adjusted, and the priority adjustment trend of the associated thread is consistent with the priority adjustment trend of the first thread. By adopting the embodiment of the invention, the response speed and performance of the system can be improved, and the user experience is improved.

An embodiment of an integrated circuit may comprise a processor with one or more cores and circuitry coupled to the one or more cores, the circuitry to control one or more interrupts based on an interrupt expansion data structure, and report information derived from the interrupt expansion data structure to a software interrupt handler. Other embodiments are disclosed and claimed.

A task scheduling method is provided, including: acquiring task data of all to-be-started tasks, with the task data including procedure numbers, processing durations, processing positions, and processing-position stop durations corresponding to all procedures corresponding to the to-be-started tasks; and inputting the task data of each to-be-started task to a task scheduling model to obtain a start moment of each to-be-started task which is output by the task scheduling model. The task scheduling model takes minimizing total time required for performing all the to-be-started tasks as an objective function, and constraint conditions of the task scheduling model are associated with the task data. The method includes performing task scheduling on the to-be-started tasks according to the start moment of each to-be-started task.

The embodiment of the invention discloses a multi-thread scheduling method, a server and a computer system, which are used for avoiding multi-thread cross-node far-end access and reducing the overhead for running a first thread and a second thread. The method comprises the steps that genetic relationship information between a first thread and a second thread is determined according to first memory access information of the first thread in a first operation stage and second memory access information of the second thread in a second operation stage; determining a first node according to the first memory access information, the second memory access information and the genetic relationship information; determining that the resource metric value of the first node is smaller than a resource threshold value; a processor unit for running the first thread and the second thread is selected from the first node.

Computer devices and management methods are provided. In an implementation, a computer device includes a subrack common component and a plurality of service nodes, wherein each of the plurality of service nodes comprises a baseboard management controller (BMC connected to the subrack common component and configured to manage the subrack common component when being in a primary status.

The invention discloses an agent context management system and device based on importance sampling, and belongs to the technical field of artificial intelligence and agents, the agent context management system and device based on importance sampling comprise a global context manager used for maintaining and managing eight types of context information; the task stack manager is used for receiving a new request and performing task splitting; and the recursive execution engine is used for executing the task in a recursive mode and controlling the recursive depth. Eight types of context information are maintained through a global context manager, splitting and storage of complex tasks are achieved through a task stack manager, a recursive execution engine provides an efficient task execution mechanism and controls the recursive depth, an importance sampler selects related contexts based on a joint probability model, and the task execution efficiency is improved. And the context compressor combines the sampling fragments into a final context, and the task scheduler coordinates the working process of each component, so that a complete agent collaborative context management system is constructed.

Disclosed in the embodiments of the present application are a multi-thread scheduling method, a server and a computer system, used for avoiding multi-threading cross-node remote access and reducing overhead of running a first thread and a second thread. The method in the embodiments of the present application comprises: according to first memory access information of a first thread in a first running stage and second memory access information of a second thread in a second running stage, determining information of affinity between the first thread and the second thread; according to the first memory access information, the second memory access information and the information of affinity, determining a first node; determining that a resource metric value of the first node is less than a resource threshold value; and selecting from the first node a processor unit used for running the first thread and the second thread.

The invention provides a frozen screen processing method and electronic equipment. The first component can be displayed on a screen of the electronic device. The first component has the ability to respond to a first operation. When a first operation on the first component is detected, the electronic device may detect whether the first component responds to the first operation. If the first component does not respond to the first operation, the electronic device can redraw the first component on the screen. According to the method, whether the frozen screen occurs in the electronic equipment or not can be detected according to whether the component displayed on the screen responds to the user operation or not, and when the frozen screen occurs in the electronic equipment, the component required to be used by the user is quickly recovered and responded by locally redrawing the component, so that the frozen screen problem is quickly solved, and the waiting time of the user is shortened.

Certain aspects of the present disclosure provide techniques for hardware-based saving and restoring of architecture state information for processing elements (PEs). According to certain aspects, one or more circuit elements trigger saving of architecture state information of at least one PE to at least one memory prior to the at least one PE transitioning from a first state to a second state and triggering restoration of the architecture state information from the at least one memory to the at least one PE prior to the at least one PE transitioning from the second state to the first state.

In accordance with an embodiment, described herein are systems and methods for auto-completion of ICS flow using artificial intelligence / machine learning. Next actions prediction is a service that assists users in modeling the flows quickly by predicting and suggesting the next set of actions a user might be thinking of adding. The service also assists the user to follow some of the best practices while creating an integration flow.

The invention provides a target object control method and device, electronic equipment, a medium and a product, and relates to the technical field of intelligent robots, a target object comprises dual-core heterogeneous processors, instruction sets and / or operating systems corresponding to the dual-core heterogeneous processors are different, and the dual-core heterogeneous processors share hardware resources; the method comprises the steps that in the starting process of any subsystem in an operating system of a target object, the running state of the dual-core heterogeneous processor is detected; in the embodiment of the invention, under the condition that the running state is determined to be abnormal, the information of the currently running task of the dual-core heterogeneous processor is stored, and the operating system of the target object is restarted based on the task information, so that the task is quickly recovered when the operating system has a problem, the task can be continuously executed from an interruption point after restarting, and the downtime is reduced.

Techniques are provided for backing up virtual machines from a computing environment to a storage environment. A virtual machine agent is utilized to generate a snapshot of the virtual machine. Metadata comprising a snapshot identifier of the snapshot and virtual disk information of virtual disks captured by snapshot is generated at the computing environment. The metadata is retrieved and used to create a metafile that is transferred to the storage environment within which snapshots of the virtual machine are to be stored. The snapshot is retrieved from the computing environment and is packaged into a snapshot package having a protocol format used by the storage environment. The snapshot package is transferred to the storage environment.

The invention provides a multi-process shared information storage method and device, equipment, a medium and a vehicle. The method comprises the following steps: acquiring a process identifier corresponding to each process in a plurality of processes of a target application; for any first process in the multiple processes, under the condition that the target application is started, the process identifier corresponding to the first process is matched, shared information corresponding to the first process is obtained, the shared information corresponding to the first process comprises a first field and a second field, and the first fields corresponding to different processes are the same; acquiring a return address of the creation function under the condition of detecting that the first process completes creation of any first thread through the creation function; obtaining a preset in-stack function in the return address; and executing the stack-in function to replace the context environment in the first thread with shared information corresponding to the first process. According to the embodiment of the invention, the execution efficiency of the multi-process application can be improved.

Techniques are provided to enable mid-thread (instruction level) preemption in a graphics processor without requiring software intervention by a graphics driver associated with the graphics processor. Hardware based mid-thread preemption is facilitated using the thread dispatch hardware of the graphics processor to trigger execution of a kernel program by the graphics processor that saves the thread state of preempted processing resources and facilitates the subsequent restoration of that thread state to the same or a different set of processing resources.

Techniques are provided for backing up virtual machines from a computing environment to a storage environment. A virtual machine agent is utilized to generate a snapshot of the virtual machine. Metadata comprising a snapshot identifier of the snapshot and virtual disk information of virtual disks captured by snapshot is generated at the computing environment. The metadata is retrieved and used to create a metafile that is transferred to the storage environment within which snapshots of the virtual machine are to be stored. The snapshot is retrieved from the computing environment and is packaged into a snapshot package having a protocol format used by the storage environment. The snapshot package is transferred to the storage environment.

One embodiment of the present invention sets forth a technique for encapsulating compute task state that enables out-of-order scheduling and execution of the compute tasks. The scheduling circuitry organizes the compute tasks into groups based on priority levels. The compute tasks may then be selected for execution using different scheduling schemes. Each group is maintained as a linked list of pointers to compute tasks that are encoded as task metadata (TMD) stored in memory. A TMD encapsulates the state and parameters needed to initialize, schedule, and execute a compute task.

The disclosure herein describes platform-level checkpointing for deep learning (DL) jobs. The checkpointing is performed through capturing two kinds of state data: (i) GPU state (device state), and (ii) CPU state (host state). The GPU state includes GPU data (e.g., model parameters, optimizer state, etc.) that is located in the GPU and GPU context (e.g., the default stream in GPU, various handles created by the libraries such as DNN, Blas, etc.). Only a fraction of the GPU memory is copied because the checkpointing is done in a domain-aware manner. The “active” memory contains useful data like model parameters. To be able to capture the useful data, memory management is controlled to identify which parts of the memory are active. Also, to restore the destination GPU to the same context / state, a mechanism is used to capture such state-changing events on an original GPU and replayed on a destination GPU.

Techniques are disclosed for capturing information surrounding a user's interactions with an application for enriching application context to be used by a digital assistant for supporting the user's interactions. In one aspect, a method includes detecting an event from a client device, where the event is associated with a user's interaction with an application on the client device. In response to detecting the event, application context is obtained for the application. The obtaining includes accessing the application context from a data store based on an identifier. A first generative artificial intelligence model can then be used to generate a list having an executable action based on the event and the application context. An execution plan is then created and executed, which includes executing the executable action using an asset to obtain an output. The output or a communication derived from the output is sent to the client device.

In a microprocessor having an instruction execution unit and first and second sets of process execution resources, context information for a process next-to-be-executed by the instruction execution unit is loaded into a register file, translation lookaside buffer and first-level data cache of the first set of process execution resources during a first interval. During the first interval and concurrently with the loading of context information for the process next-to-be executed, the instruction execution unit executes a current process, including accessing context information for the current process within the register file, translation lookaside buffer and first-level data cache of the second set of process execution resources.

Certain aspects of the present disclosure provide techniques for hardware-based saving and restoring of architecture state information for processing elements (PEs). According to certain aspects, techniques involve triggering, via one or more circuit elements, saving of architecture state information of at least one processing element (PE) to at least one memory prior to the at least one PE transitioning from a first state to a second state; and triggering, via the one or more circuit elements, restoration of the architecture state information from the at least one memory to the at least one PE prior to the at least one PE transitioning from the second state to the first state.

Described herein is a technique to enable hardware driven checkpointing within an accelerator device without requiring explicit host software intervention to generate the checkpoint. One embodiment provides an accelerator device comprising a memory interconnect, a plurality of accelerator cores, and a scheduler coupled with the plurality of accelerator cores. The scheduler is configured to receive an checkpoint creation job to cause generation of a compressed checkpoint for a training operation executed via the plurality of accelerator cores, atomically create a compressed checkpoint for at least a portion of the training operation, and store the compressed checkpoint to checkpoint storage associated with the accelerator device.

The present disclosure describes a system and method for preempting a long-running process with a higher priority process in a machine learning system, such as a hardware accelerator. The machine learning hardware accelerator can be a multi-chip system including semiconductor chips that can be application-specific integrated circuits (ASIC) designed to perform machine learning operations. An ASIC is an integrated circuit (IC) that is customized for a particular use.

The invention provides a learning machine voice instruction interaction method and system, is applied to the technical field of learning machines, and realizes fine management of help seeking of a user by starting an independent'temporary session state record 'for each knowledge point. By generating'multi-level guide prompts' and'multi-level question exploration ', the system can gradually guide the user to think instead of directly giving answers, so that not only is the independent thinking ability of the user cultivated, but also the system can more accurately judge the real mastering degree of the user on the knowledge point. When the user answers correctly, the system confirms the mastering of the user; and when the answer is incorrect, the system adjusts an auxiliary strategy, such as recommending a teaching video, so that more targeted help is provided. Therefore, according to the method provided by the invention, false triggering of dependence intervention caused by multi-user sharing can be avoided, the accuracy and individuation of an auxiliary strategy of the learning machine are ensured, and the user experience and learning efficiency are remarkably improved.

An apparatus comprises exception return state register storage, and processing circuitry. In response to a guarded control stack (GCS) exception return state push instruction, the processing circuitry obtains exception return state information from the exception return state register storage and push the state information to a GCS data structure. In response to a GCS exception return state pop instruction, the processing circuitry obtains GCS-protected exception return state information from the GCS data structure. In at least one operating state, the processing circuitry detects, in response to an attempt to modify the exception return state information stored in the exception return state register storage, whether an exception return state lock parameter is in a locked state or an unlocked state, and signals a fault when it is in the locked state.