Google Patente

This document describes technology for a runtime memory system that utilizes unpredictable multiplexing of memory bus lines. The technology dynamically reassigns data and/or address bus connections between a processor and its runtime memory during warm reset operations. The bus line assignments are made through a multiplexing mechanism-for example, data bus line from the processor may be reassigned to link with bus line at the runtime memory system interface. When a warm reset occurs, the physical mapping of memory locations changes due to this bus line reassignment. The system maintains standard memory operations while implementing this dynamic bus line configuration. Malicious software relying upon memory-resident exploit becomes inoperable after warm reset due to the changed physical memory topology created by the bus line reassignment process.

A first computing device is configured to receive an indication of subscription request to at least one data of a plurality of data fields of a media bridging protocol from a second computing device. The first computing device receives the indication of subscription request as specifying one or more types of messages that correspond to at least one of the plurality of data fields and that are to be sent via the bridging protocol to the second computing device. The first computing device is further configured to generate messages for the plurality of data fields and determine a subset of the generated messages that are associated with the subscribed data fields to be provided to the second computing device. The first computing device is further configured to provide the subset of messages to the second computing device via the bridging protocol.

Disclosed is an object storage system that can accept any number of write requests that add zero or more bytes to the end of the object. A write request may flush the data up to the current end of the object to make it durable and accessible for read operations. A write request may make the object permanently immutable.

The technology is generally for providing resilience to interconnect networks in large scale computing. Processing nodes of the network are arranged in reconfigurable pods that can contain large numbers of processing nodes. Processing nodes are grouped into building blocks of the pod, each building block may be contained within a physical computing rack. Processing nodes are interconnected via interconnect pathways that connect processing nodes to network switches. A building block's processing nodes communicate to network switches through regular communication pathways via the interconnect network. Additionally, a protection communication path is provided between a building block and the network. If a regular communication path is affected by a failure of a switch or interconnect link, the protection pathway is used to route the traffic affected by the failure without impacting the computing network. An electrical circuit switch is provided at the building block level to route data to the protection pathway.

An optical communication system includes a pair of transceivers and an optical circuit switch with multi-core fibers at each port that provides a communication bridge between the pair of transceivers. The transceivers each transmit a forward transmission signal through a first single-core fiber and receives a reverse transmission signal through a second single-core fiber. A fan-in/fan-out device can be positioned between the switch and each transceiver along an optical path. The fan-in/fan-out device is configured to couple the light or transmission signal transmitted through the single-core fiber with a core of a multi-core fiber. The multi-core fiber is further coupled to the port.

Methods, systems, and apparatus for reducing latency of configuration of image sensor and image signal processor. A computing system can include a machine learning (ML) processing engine that can process denoising diffusion ML models for execution on statically compiled ML accelerators. The system can determine that a partitioned graph representation, which includes connected subgraphs that each represent at least one layer of the neural network of the ML model, forms a directed acyclic graph. The system can insert cache nodes in the graph representation, where each cache node corresponds to a respective subgraph and is configured to cache output of the respective subgraph. The system can generate an execution dataflow graph including a plurality of iterations of the partitioned graph representation, where, at one or more iterations during model inference operations, the execution dataflow graph uses inputs from cache nodes and excludes execution of subgraphs corresponding to the cache nodes.

This document describes systems and techniques for delivering patch partitions to local function managers of a computing device. For example, a system includes a plurality of local function managers in a computing device, each of the local function managers is configured to identify an instance for execution of one or more patch partitions, each of the patch partitions including executable code to direct a local function and generate a patch request for the one or more patch partitions. A global function controller is configured to access a group of patch partitions including each of the one or more patch partitions for each of the plurality of local function managers and to respond to receipt of the patch request by delivering one or more requested patch partitions to the local function manager generating the patch request.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for performing a machine learning task on a network input to generate a network output using an attention neural network. The attention neural network includes a plurality of attention blocks arranged in a stack one after the other. One or more of the plurality of attention blocks are generalized residual attention blocks. Each generalized residual attention block includes one or more generalized residual blocks and an attention layer. The one or more generalized residual blocks includes a query generalized residual block.

A link layer architecture for D2D interconnections or communications. The architecture includes a self-contained D2D module that is capable of receiving data in multiple input bus protocols and transferring that data through a physical layer connection to another chiplet. The D2D module is a self-contained base block used to transfer data across dies. It includes a transmit (Tx) section/logic and a receive (Rx) section/logic. Each section/logic forms a building block that can be used to transfer data between dies or chiplets.

The present disclosure is generally directed to a modular busbar assembly that can be adjusted to fit varying rack power targets and connector interfaces, accommodating multiple connector geometries along one cost-effective assembly. The modular busbar assembly enables high power densities at powershelf regions by accommodating single- and double-pole connectors and connector contacts at various positions relative to the modular busbar assembly. The modular busbar assembly is customizable because the auxiliary busbar geometry may be altered independently from the primary busbar geometry.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

A slidable fiber optic adaptor slidable relative to a mounting support configured to be mounted to a fiber optic tray in a panel assembly is provided for fiber optic interconnection. The slidable adaptor may be slidable between a non-extended position and an extended position. Thus, when the fiber optic connectors having different polarities are connected to the slidable adaptors, the slidable fiber optic adaptor may be pulled out in the extended position to allow the fiber optic connectors with different polarities to be in a staggered configuration, thus, providing sufficient space therebetween for finger access for operators or technicians for ease of cable inspection and management.

Provided are computer-implemented systems and methods which improve embedding generation model performance with respect to localized quality issues. In particular, the proposed techniques can operate to improve localized embedding model performance without incurring significant re-training costs and/or re-deployment costs associated with full model re-training.

The invention relates to a computer-implemented method to facilitate verified service requests by a user computing device, comprising receiving application program code distributed by an application service system, wherein the application program code comprises identifying information; extracting the identifying information from the application program code resident on the user computing device; transmitting an access request to the application service system, wherein the access request comprises the identifying information extracted from the application program code resident on the user computing device; receiving, from the application service system, an access token comprising the identifying information extracted from the application program code resident on the user computing device; and transmitting a service request including the access token to the application service system.

Generally disclosed herein is an approach for an optimized data computation for machine learning (ML) operations using a plurality of small-sized ML accelerators integrated into a general-purpose compute system-on-chip (SoC). One or more processors of the general-purpose compute SoC may be configured to receive a workload and divide the workload into a plurality of sub-workloads. The plurality of sub-workloads may be distributed among the plurality of small-sized ML accelerators. Each of the plurality of the ML accelerators can be configured to output a partial outcome. The outputs from each ML accelerator can be combined using atomic store operations using a dedicated opcode and a specified memory address.

The present application relates to systems and methods for regulating power for one or more electronic devices. In accordance with aspects of the disclosure, one or more micro-controllers are configured to regulate both peak and average power being consumed by devices within the system. The power being consumed within the system can be determined over different durations of time and multiple independent feedback loops may be used to regulate power within the system. In addition, the system can include a feedback loop that is configured to allow the system to enter into a low-power mode.

An example computing system retrieves information from one or more applications, and receives an indication of a natural language input. The computing system applies a machine learning model to at least a portion of the information and at least a portion of the indication of the natural language input to generate instructions for generating at least one graphical component, e.g., a custom widget. The instructions may include a temporal parameter for the at least one graphical component, in which the computing system may update, based on the temporal parameter, the instructions for generating the at least one graphical component to update the at least one graphical component.

The technology is generally directed to the continued use of a link between nodes of a network in which at least one of the lanes in the link has failed. Rather than identify the link as a failure in response to a lane failure, the traffic may be redistributed amongst the remaining lanes in the link. The redistribution of traffic may result in the link operating in a degraded mode. To mitigate the reduced traffic capacity of the degraded mode, the remaining links may be overclocked, such that the bandwidth of each link is increased to return the link to a traffic capacity corresponding to the initial capacity. In some examples, the traffic may be diverted to a redundant, otherwise inactive lane, thereby allowing for the traffic to be transmitted and received via the link, without any degradation in the performance of the link.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

Provided is a system for standardizing health data for use by machine-learned models in responding to user queries. A computing system obtains health data for a user from one or more health data sources. The computing system generates standardized health data, wherein the standardized health data is generated by converting the health data into a standardized format usable by a machine-learned model. The computing system receives a user health query from the user. The computing system generates a model input for the machine-learned model comprising the user health query and the standardized health data. The computing system provides the model input to the machine-learned model. The computing system receives a model output from the machine-learned model. The computing system transmits the model output for display to the user.

Example embodiments of the present disclosure provide for an example method including obtaining data associated with media channels. The example method includes obtaining, by a computing device of a first platform, first channel data associated with a first media channel of an entity, the first media channel being on a first platform. Additionally, the method can include obtaining second channel data associated with a second media channel of the entity, the second media channel being on a second platform. Moreover, the method can include processing, using a machine-learning model, the first channel data and the second channel data to determine an allocation of resources to the first media channel and the second media channel. Furthermore, the method can include generating and presenting a recommendation based on the determination.

A method (400) includes receiving a time series anomaly detection query (20) from a user (12) and training one or more models (212) using a set of time series data values (152). For each respective time series data value in the set, the method includes determining, using the trained models, an expected data value for the respective time series data value and determining a difference between the expected data value and the respective time series data value. The method also includes determining that the difference between the expected data value and the respective time series data value satisfies a threshold (314). In response to determining that the difference between the expected data value and the respective time series data value satisfies the threshold, the method includes determining that the respective time series data value is anomalous and reporting the anomalous respective time series data value (152A) to the user.

A computer-implemented method comprising: receiving, by an application (108) executing on a user device (102) that is not an application store, a single content response from a content server (120), wherein the single content response comprises both digital content (112) referencing a second application that is not installed on the user device and application store data (116) for the second application; in response to receiving the single content response and when the application is an active application, providing, by the active application and in a user interface (110) of the active application, the digital content concurrently with an application store user interface (114) that includes (i) the application store data for the second application and (ii) an install element (118) for triggering installation of the second application on the user device from within the active application; detecting user interaction with the install element provided within the user interface of the active application; and in response to detecting the user interaction with the install element, triggering a download of the second application independent of launching an application for the application store.

A hash table system, including a plurality of hash tables, associated with respective hash functions, for storing key-value pairs; an overflow memory for storing key-value pairs moved from the hash tables due to collision; and an arbiter for arbitrating among commands including update commands, match commands, and rehash commands, wherein for each system clock cycle, the arbiter selects as a selected command one of an update command, a match command, or a rehash command, and wherein the hash table system completes execution of each selected command within a bounded number of system clock cycles.

Zusammenfassung wird geladen …

Zusammenfassung wird geladen …

A system for managed data transfer in compute network, includes a first data store and a second data store. A data transfer agent and one or more servers plan and implement a data transfer to move data stored in the first data store to the second data store via the data transfer agent. The data agents are centrally managed by a control plane and are initialized with authentication credentials that are needed to effectuate the data transfer. The source data universe and the destination data universe may have different authentication interfaces. The data transfer agent will use a first credential to access data at the source and a second credential to copy the requested data to the destination.

This system and method described herein provide for efficient monitoring of machine learning models in a data warehouse toolset integrated with a cloud platform. Computation of statistics and anomalies for training data and serving data for the machine learning models may be distributed across physical shards in the data warehouse. Computed results may be converted to a format compatible with external platforms. Model monitoring may be supported using table valued functions (TVFs) to compute descriptive statistics of data and to compute the feature distribution difference between two different pieces of data, a scalar function to validate statistics, and an internal task handler to parse monitoring statistics results and save it as a job to make it visualizable.

In accordance with aspects of the disclosure, an OLAP database is used to store a plurality of tables that are each associated with a particular client or tenant and which are each configured in accordance with a table-specific schema of metadata. A catalog service can be configured to allow for efficient read and write operations in connection with the tables stored within the OLAP database.

A wearable computing device and system are provided. The wearable computing device includes a display device, an inertial measurement unit (IMU), an optical sensor(s), and a processor(s). The processor(s) is configured to obtain optical sensor data associated with a user of the wearable computing device, obtain movement data characterizing motion of the user, process the optical sensor data and the movement data to identify a motion-induced spurious component of the optical sensor data, and modify an operation mode of the wearable computing device based on the motion-induced spurious component of the optical sensor data.

Implementations are directed to generating voice-based chatbot policy override(s) and/or utilizing voice-based chatbot policy override(s) in conjunction with existing voice-based chatbot(s). The voice-based chatbot policy override(s) can correspond to, for example, machine learning (ML) model(s) that supplement functionality of the existing voice-based chatbot(s). Notably, the voice-based chatbot policy override(s) are associated with rule(s) (e.g., by virtue of training the ML model(s) that correspond to the voice-based chatbot policy override(s)) for when the voice-based chatbot policy override(s) should be utilized in lieu of the existing voice-based chatbot(s) in responding to spoken utterance(s) of human user(s) engaged in corresponding conversation(s) with the voice-based chatbot policy override(s). Nonetheless, from a perspective of the human user(s), it appears as if they are still engaging in the corresponding conversations with the existing voice-based chatbot(s). Thus, the functionality of the existing voice-based chatbot(s) can be supplemented without having to re-train the existing voice-based chatbot(s).

A method includes detecting a session between a mobile device and an alternative playback device. In response to detecting that the session has been established, the method also includes presenting a graphical user interface (GUI) on a screen of a mobile device of a user including a voice search GUI element representing a voice search, receiving a user selection of the voice search GUI element, receiving an audio input of the user to request a search of one or more media items, the audio input specifying one or more search criteria for the search, and submitting a textual search query derived from the audio input for processing, wherein a result of the search is to be presented to the user on a screen of the alternative playback device.

A method performed by a device having, and a system comprising, a plurality of sensor controllers, wherein each sensor controller of the plurality of sensor controllers is configured to implement a respective decimation filter that is configured to generate a single sample output from multiple input samples generated by a corresponding sensor of a plurality of sensors. The method comprises, and a non-master sensor controller of the plurality of sensor controllers is configured to perform operations comprising, detecting, by a first sensor controller of the plurality of sensor controllers, an update in a memory for the first sensor controller; determining, based on the update and by the first sensor controller, whether to operate the first sensor controller as a first master sensor controller; and in response to determining to operate the first sensor controller as the first master sensor controller, configuring the first sensor controller in a master controller mode, wherein a sensor controller in the master controller mode is configured to generate a sync signal upon receiving a threshold number of input samples generated by the respective sensor for the sensor controller, and wherein the sync signal causes each sensor controller in the plurality of sensor controllers operating in a non-master controller mode to implement the respective decimation filter to generate the single sample output.

Implementations set forth herein relate to an automated assistant that is responsive to spoken utterances spoken by multiple different users simultaneously, or otherwise within a narrow window of time, in furtherance of initializing one or more actions. For instance, when a primary user is interacting with an automated assistant, a secondary user may also provide a spoken utterance. In response, the automated assistant can selectively determine whether any input from the secondary user should affect a request from the primary user. In some instances, the automated assistant can elect to either disregard the secondary input, separately respond to the secondary input, or use some amount of content of the secondary input to further the request from the primary user. Incorporation of secondary input when fulfilling a primary request can be fluidly performed to resemble human conversation and eliminate any unnecessary engagement between the automated assistant and the users.

Systems and methods for managing deployment of applications and services within a network having a plurality of edge platforms. A domain name service (DNS) request from a client device requesting for content from a third-party provider is modified to include location information indicating a client device location. A deployment manager uses the location information in the modified request to select a nearby edge platform for servicing the request. Changes to the edge platform servicing the request may be imitated based on changes in location information received in future requests. The deployment manager may also maintain a database to track resources that are already deployed in nearby edge platforms, conserving resources when selecting the edge platforms to service requests. The deployment manager also directs third-party content providers to the edge platform using a separate service so that communication with third-party content providers does not share the location information.

A method that includes receiving a plurality of workloads (102), each associated with respective security criteria (104) and scheduled for execution on a distributed computing system (140) divided into a plurality of security rings (160) each associated with a respective subset of computing devices (162) of the distributed computing system that is physically isolated from the other security rings. For each respective workload, the method includes determining, using the respective security criteria, a security level (410) of the respective workload and identifying, using the security level of the respective workload, one or more of the plurality of security rings that are eligible for executing the respective workload. The method also includes executing the respective workload on one or more computing devices selected from one of the respective subsets of computing devices associated with the identified one or more of the plurality of security rings eligible for executing the respective workload.

Aspects of the disclosure are directed to a side sideband interface (SSBI). The SSBI is a die-to-die (D2D) interface configured for various protocols to enable cross-die communication with higher robustness and flexibility and lower cost. The SSBI includes a control status register (CSR) wrapper containing the CSRs for configuring and monitoring the SSBI, a protocol handler for converting the various protocols both to and from a standard streaming interface as well as performing arbitration, and a link layer for performing serialization and deserialization of data as well as flow control management and error detection.