中科微点-全球专利检索

  1. Login
  2. Sign Up

Search Results

8 records (took 0.019 seconds)

    COMPUTATIONAL GRAPH CRITICAL SECTIONS
    2.
    发明申请
    COMPUTATIONAL GRAPH CRITICAL SECTIONS 审中-公开

    公开(公告)号：US20200167207A1

    公开(公告)日：2020-05-28

    申请号：US16695884

    申请日：2019-11-26

    Applicant: Google LLC

    Inventor： Eugene Brevdo ,  Alexandre Tachard Passos

    IPC: G06F9/52 ,  G06F9/48 ,  G06F9/50 ,  G06F16/901

    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing critical section subgraphs in a computational graph system. One of the methods includes executing a lock operation including providing, by a task server, a request to a value server to create a shared critical section object. If the task server determines that the shared critical section object was created by the value server, the task server executes one or more other operations of the critical section subgraph in serial. The task server executes an unlock operation including providing, by the task server, a request to the value server to delete the shared critical section object.

    Computational graph critical sections
    3.
    发明授权
    Computational graph critical sections 有权

    公开(公告)号：US11188395B2

    公开(公告)日：2021-11-30

    申请号：US16695884

    申请日：2019-11-26

    Applicant: Google LLC

    Inventor： Eugene Brevdo ,  Alexandre Tachard Passos

    IPC: G06F9/52 ,  G06F9/46 ,  G06F9/48 ,  G06F9/50 ,  G06F16/901

    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing critical section subgraphs in a computational graph system. One of the methods includes executing a lock operation including providing, by a task server, a request to a value server to create a shared critical section object. If the task server determines that the shared critical section object was created by the value server, the task server executes one or more other operations of the critical section subgraph in serial. The task server executes an unlock operation including providing, by the task server, a request to the value server to delete the shared critical section object.

    Systems and Methods for Reducing Idleness in a Machine-Learning Training System Using Data Echoing
    4.
    发明申请
    Systems and Methods for Reducing Idleness in a Machine-Learning Training System Using Data Echoing 审中-公开

    公开(公告)号：US20200372407A1

    公开(公告)日：2020-11-26

    申请号：US16871527

    申请日：2020-05-11

    Applicant: Google LLC

    Inventor： Dami Choi ,  Alexandre Tachard Passos ,  Christopher James Shallue ,  George Edward Dahl

    IPC: G06N20/00

    Abstract: A method for reducing idleness in a machine-learning training system can include performing operations by computing devices. A first set of training operations can access and prepare a plurality of training examples of a set of training data. A second set of training operations can train a machine-learned model based at least in part on the set of training data and can include one or more repeat iterations in which at least a portion of the second set of training operations is repeatedly performed such that the training example(s) are repeatedly used to train the machine-learned model. A rate of the repeat iteration(s) can be based at least in part on an echo factor that can be based at least in part on a comparison of a first computational time of the first set of training operations to a second computational time of the second set of training operations.

    Systems and methods for reducing idleness in a machine-learning training system using data echoing
    7.
    发明授权
    Systems and methods for reducing idleness in a machine-learning training system using data echoing 有权

    公开(公告)号：US11537949B2

    公开(公告)日：2022-12-27

    申请号：US16871527

    申请日：2020-05-11

    Applicant: Google LLC

    Inventor： Dami Choi ,  Alexandre Tachard Passos ,  Christopher James Shallue ,  George Edward Dahl

    IPC: G06N20/00 ,  G06N3/08

    Abstract: A method for reducing idleness in a machine-learning training system can include performing operations by computing devices. A first set of training operations can access and prepare a plurality of training examples of a set of training data. A second set of training operations can train a machine-learned model based at least in part on the set of training data and can include one or more repeat iterations in which at least a portion of the second set of training operations is repeatedly performed such that the training example(s) are repeatedly used to train the machine-learned model. A rate of the repeat iteration(s) can be based at least in part on an echo factor that can be based at least in part on a comparison of a first computational time of the first set of training operations to a second computational time of the second set of training operations.

    COMPUTATIONAL GRAPH CRITICAL SECTIONS
    8.
    发明申请
    COMPUTATIONAL GRAPH CRITICAL SECTIONS 有权

    公开(公告)号：US20220083400A1

    公开(公告)日：2022-03-17

    申请号：US17533223

    申请日：2021-11-23

    Applicant: Google LLC

    Inventor： Eugene Brevdo ,  Alexandre Tachard Passos

    IPC: G06F9/52 ,  G06F16/901 ,  G06F9/50 ,  G06F9/48 ,  G06F9/46

    Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for implementing critical section subgraphs in a computational graph system. One of the methods includes executing a lock operation including providing, by a task server, a request to a value server to create a shared critical section object. If the task server determines that the shared critical section object was created by the value server, the task server executes one or more other operations of the critical section subgraph in serial. The task server executes an unlock operation including providing, by the task server, a request to the value server to delete the shared critical section object.

Patent Agency Ranking