21 results about "Adapter (computing)" patented technology

This invention discloses a non-embedded UAV information acquisition device and method. The non-embedded UAV information acquisition device includes a shell assembly, a three-mode adapter structure, a standardized physical interface, a main control module, a communication module, a positioning and sensing module, and a power management module. The internal cavity wall of the shell assembly is fitted with an anti-magnetic shielding layer made of permalloy. The three-mode adapter structure is integrated into the bottom of the shell assembly and includes a magnetic suction component, a suspension component, and a screw fixing interface. The standardized physical interface is located on the surface of the shell assembly and uses a gold-plated gold finger contact structure, which is electrically connected to the main control module. This invention enables deployment without disassembly of the UAV through a non-intrusive design while maintaining the UAV's airworthiness. It utilizes edge computing to alleviate cloud scheduling pressure and reduce task latency, and with independent power supply, achieves long endurance and high reliability, making it widely adaptable to various UAV models and harsh operating environments.

The application belongs to the technical field of collaborative learning, and particularly relates to a parameter efficient collaborative learning method and system suitable for intelligent sensing devices. The application is directed to a collaborative application scene of multiple distributed terminal devices with sensing, computing and communication capabilities. Under the premise of freezing backbone parameters, a shared adapter and a personalized adapter are introduced. Personalized parameters are only updated and saved locally to maintain single-device adaptation capability. Shared parameters realize knowledge sharing through interaction and weighted aggregation of neighboring devices. An adaptive weight matrix is used to fuse shared prediction and personalized prediction, so that global collaboration and personalized adaptation are considered in the distributed sensing scene with significant data distribution difference. Meanwhile, only the lightweight trainable part is updated and exchanged, which significantly reduces the end-side computing burden and communication overhead, and maintains the efficiency and convergence stability of collaborative learning under resource-limited conditions.

This application discloses an adapter structure, a board assembly, a quantum computing measurement and control system, and a quantum computer. The adapter structure is used to electrically connect a first signal connector of a functional board to a second signal connector of a measurement and control circuit. The adapter structure includes: a mounting base for detachable installation with the functional board; and a signal adapter connector installed within the mounting base. One end of the signal adapter connector is used to insert and cooperate with the first signal connector under the support of the mounting base to form a first connection interface, and the other end is used to insert and cooperate with the second signal connector to form a second connection interface. The first connection interface provides low insertion and extraction force to extend the insertion and extraction life of the first signal connector, while the second connection interface provides high insertion and extraction force to enhance the connection reliability of the second signal connector. This adapter structure significantly improves the insertion and extraction life of the functional board and reduces maintenance costs while ensuring connection reliability.

The application discloses a kind of based on flow perception intelligent network interface card optimization system and method, it belongs to computer network technical field.The optimization system of the application includes flow perception module, flow prediction module, resource scheduling module, priority queue module and load balancing module.The application introduces resource monitoring, flow perception, resource prediction and dynamic scheduling module, real-time monitoring network flow and predicting its change trend, according to flow mode dynamically adjusts resource allocation.The application effectively solves the problem of resource contention, through priority ordering and load balancing, ensure the stable operation of high priority task, improve the throughput and stability of system, reduce resource waste.The application is suitable for resource management of intelligent network interface card in data center and cloud computing environment, improves network processing efficiency and resource utilization, guarantees service quality, reduces resource waste, and enhances the adaptive capacity of system to dynamic load and complex flow mode.

This invention discloses a cluster component update method, apparatus, electronic device, and program product, relating to the field of cloud computing technology or other related technical fields. The method includes: executing an update task; issuing an update instruction to a node agent module deployed on a specified component node; the update task includes at least: a specified component node and a target version; downloading the binary file of the target version through the node agent module; and, if the file integrity verification passes, calling the component hot-loading adapter corresponding to the specified component node; starting a new process through the component hot-loading adapter; using the new process to update the original version to the target version, and notifying the old process to exit; and confirming the update task is complete after verifying that the component functions of the specified component node are normal. This invention solves the technical problem in related technologies where large-scale cluster component updates rely on complete node reconstruction, which is time-consuming and reduces the response capability to sudden security threats.

The present application relates to a kind of training phase perception type SHARP in-network collection communication operator dynamic scheduling system and method, system includes: training framework domain: perception training phase and generate standardization phase descriptor;Host channel adapter domain: descriptor is forwarded to switch SMA by out-band channel, and listens to GPU progress token to realize submission gate;Switch CFU dynamic scheduling domain: according to strategy table, reconfigure parameter in CFU pipeline idle gap, realize no-interruption dynamic scheduling.The present application establishes a light training phase perception collaborative link between training framework in host side and CFU in switch side, realizes the real-time dynamic scheduling of CFU execution strategy with training phase, so that the in-network computing acceleration benefit of SHARP is extended from the current limited All-Reduce scene to the complete communication life cycle of AI large model training.

This invention discloses a multi-condition testing and inspection device for liquid cooling circuits of AI servers, comprising a housing, an adapter mechanism inside the housing, and a fixed frame. The outer surface of the fixed frame is fixedly connected to the interior of the housing, and a fixed wheel is fixedly connected to the outer surface of the fixed frame. The outer surface of the fixed wheel has two symmetrical annular grooves, and a sealing ring is rotatably connected inside each of the two annular grooves. A connecting groove is formed at an equal angle on the outer surface of the sealing ring, and the interior of the connecting groove is connected to the interior of the annular groove. A spring tube is connected inside the connecting groove, and one end of the spring tube passes through the body of the sealing ring and extends to the outside of the sealing ring. This invention relates to the technical field of AI server testing equipment and solves the problem that existing AI server testing equipment cannot adapt to the interfaces of liquid cooling circuits of AI servers from different manufacturers and with different computing power levels, resulting in poor compatibility and insufficient scalability of existing testing equipment.

A method for adjusting fan speed based on a cable connection includes reading an identifier (“ID”) of a network interface card (“NIC”) installed in a computing device, the identifier indicating a NIC type, determining one or more cable types connectable to the NIC from the specification information of the NIC, and measuring power consumption of the NIC. The method includes determining from the power consumption the one or more cable types connectable to the NIC a current cable type connected to the NIC or a lack of a cable connected to the NIC, determining power requirements for the NIC based on the lack of a cable connected to the NIC or the current cable type connected to the NIC, and adjusting a fan speed of one or more fans of the computing device in response to determining the power requirements for the NIC.

Systems and techniques for listening for input / output (I / O) events in a computing system are described. An example technique includes obtaining a configuration including a plurality of listening space profiles, each listening space profile indicating a respective range of memory addresses mapped to a respective completion queue. The technique also includes monitoring input / output (I / O) traffic exchanged on a communication interface between an I / O adapter and a processor in the computing system based on the configuration. The technique further includes performing one or more actions to assist in processing of the I / O traffic based in part on the monitoring.

The application discloses an AI controllable intelligent agent task scheduling and management system based on a task control ticket. The task control ticket is taken as a carrier to establish a balanced boundary of three separated rights, i.e., scheduling and guiding right (A right), management and control right (B right) and model / tool execution right (C right). Through a sandbox three radius (semantics, data and action) and a risk budget parameter, a management decision module is driven to perform ticket and data verification, and to realize release, interception, degradation or ticket splitting. A model / tool adapter adjusts tasks between a CPU deterministic channel and a GPU inference channel according to a budget strategy. At a submission gate, a governance strategy, a tool adapter and a knowledge base snapshot are executed to version pinning, and a playback anchor point is generated. The application realizes minimum permission control through a ticket pedigree, and ensures audit certainty through version pinning and a playback mechanism, so that the whole process is auditable and traceable, and the offline and low-computing-power deployment environment with strict requirements for security and controllability is adapted.

A system for communicating over a physical port includes a Remote Direct Memory Access (RDMA) capable network adapter, a host computing device communicatively coupled to the RDMA capable network adapter, a physical port associated with the host computing device, and a plurality of interfaces associated with the physical port, the plurality of interfaces having different RDMA protocols.

This invention discloses a method, device, medium, and product for generating operator code for chip adaptation operators. The method includes: extracting adaptation requirement information for generating target operator code in response to an operator code generation request from a target operator; obtaining multiple update large model prompts corresponding to each preset exploration strategy based on the adaptation requirement information and multiple preset exploration strategies; inputting each update large model prompt into multiple working nodes equipped with the same target large model, and then triggering each working node to cyclically execute the operator code generation task in parallel; monitoring and assisting in the task execution in real time until the target operator code that has been verified and meets the performance requirements is successfully obtained in multiple working nodes, and then using the target operator code as the response result of the operator code generation request. This technical solution can improve the reliability and efficiency of generating operator code adapted to different hardware platforms and improve the utilization of computing resources.

A system and method of propagating link aggregation control protocol status to single root input / output virtualization virtual function status. The method includes obtaining, by a processing device and using a relay agent executing on the processing device, state information from a Link Aggregation Control Protocol (LACP) speaker of a host machine. The method includes detecting, based on the state information, a communication failure between a physical function (PF) of a network adapter of a host machine and a first network switch of a plurality of network switches. The method includes identifying one or more computing environments that are communicatively coupled to the first network switch through a first virtual function (VF) associated with the PF. The method includes notifying the one or more computing environments about the communication failure by modifying a link state of the first VF.

The application provides a SoP packaging structure and a preparation method thereof. Based on the settings of a TGV metal column, a TSV composite adapter plate and a rewiring layer, a high-density integrated composite adapter plate can be prepared. After the first chip and the second chip are combined, a high-density SoP packaging structure can be formed. After a cutting process, multiple computing power modules can be accommodated in one SoP packaging monomer structure, meeting the needs of small size, high integration and high economic efficiency. Furthermore, based on the groove and the TGV groove, a metal heat dissipation layer covering the TSV composite adapter plate can be formed while the TGV metal column is prepared, which can reduce the process flow, reduce the cost, achieve good heat dissipation and improve the product quality.

A computer network system and a task scheduling system and method are provided. The computer network system includes a plurality of computers without shunt jumper components, and a computer provided with a shunt jumper component is arranged between every two adjacent computers without the shunt jumper components. Each of the computers is provided with two central processing units, and each of the central processing units is connected to a high-speed network interface card. The task scheduling method is employed for scheduling, a single-node computing task is allocated to a shunted computer provided with the shunt jumper component, and remaining computers are configured to execute single-node and / or multi-node computing tasks. Upon completion of task execution, shunt jumper wires are disconnected through a shunt controller, such that adjacent nodes are communicated through the high-speed network interface card. All the computers equipped with the shunt jumper components execute single-node computing tasks.

This invention discloses a dual-mode EEG interaction system and method supporting zero-calibration and online incremental learning, belonging to the fields of human-computer interaction systems, embedded artificial intelligence, and edge computing. The system includes a wearable acquisition front-end and an embedded computing core. The computing core has a read-only storage area and a read-write storage area, which respectively pre-store frozen general backbone network parameters and variable lightweight adapter parameters. The system uses a manifold-aware signal conditioning module to project the EEG covariance matrix onto the Riemann tangent space to eliminate individual differences; a parameter-separated decoding engine extracts common features using the frozen backbone and performs personalized corrections using the adapter; and a dual-mode state machine dynamically switches between zero-calibration statistical adaptation mode and online incremental learning mode based on trigger signals. This invention achieves plug-and-play and online adaptive evolution of the EEG system, reduces the computing power and storage overhead required for online learning at the edge, and improves user experience and system robustness.

Systems and methods for supporting a single logical IP subnet across multiple independent layer 2 subnets in a high performance computing environment. A method can provide, at a computer including one or more microprocessors, a logical device, the logical device being addressed by a layer 3 address, wherein the logical device comprises a plurality of network adapters, each of the network adapters comprising a physical port, and a plurality of switches. The method can arrange the plurality of switches into a plurality of discrete layer 2 subnets. The method can provide a mapping table at the logical device.

This invention provides a PS / 2 keyboard compatibility handling method based on the BIOS of a computing platform. A bridging adapter module detects the physical characteristics of the PS / 2 keyboard and marks its status. Domestic computing motherboards read the status marks during the BIOS stage, call pre-stored multi-version PS / 2 driver adapter libraries, and dynamically match the keyboard model through a protocol parsing module. The system has a built-in hot-swappable adapter module, using an I2C bus to realize communication between data transmission and reception and the processing unit, simulating bidirectional initialization signal interaction between the host and the keyboard, supporting re-adaptation under hot-swappable conditions. This invention can reduce device detection time by more than 150ms, is compatible with mainstream PS / 2 keyboard models, and is suitable for terminal devices such as servers and industrial control systems that have high requirements for input device stability.

The application relates to the technical field of computer architecture, and discloses a memory-computing integrated large language model DRAM fault-tolerant method, which comprises the following steps: constructing a space-time related multi-granularity DRAM fault model, defining a memory bank level space constraint, and a single-bit or multi-bit flip fault mode, and establishing a DRAM fault probability calculation method based on the Arrhenius formula; performing multi-precision and module level fault sensitivity quantitative analysis on the large language model; based on the fault sensitivity quantitative result, performing precision-aware bit-level differentiated fault injection, and using a fine-tuning framework of a low-rank self-adaptive adapter based on module heterogeneity perception to perform fault-tolerant enhancement on the target large language model; the application has small parameter redundancy, does not add a hardware error correction circuit, and does not increase the hardware critical path delay, thereby improving the tolerance and robustness of the large language model to single-bit and multi-bit flip faults under the DRAM memory-computing integrated architecture.

The application discloses an AI computing power demand-oriented data center resource integrated scheduling and transaction platform, and relates to the technical field of distributed computing and resource scheduling; comprising a resource access layer, a core platform layer, a blockchain network layer and an application interaction layer. The resource access layer contains a resource adapter; the core platform layer contains an AI job perception and analysis engine, a global intelligent scheduler, a blockchain middleware and a billing engine, wherein the AI engine adopts a domain fine-tuning large model to perform chain reasoning on natural language demand, generates a structured resource specification list, the scheduler generates a resource allocation scheme according to a multi-objective optimization algorithm, and the blockchain middleware realizes assetization evidence and trusted transaction of computing power NFT; the blockchain network layer is a permissioned chain network, used for storing power digital identity, usage proof and executing smart contracts, through the cooperation of each layer, an integrated closed-loop service process is formed, the AI computing power use threshold is significantly reduced, and the resource utilization is improved.