MULTIPLE-MODEL HETEROGENEOUS COMPUTING
    1.
    发明公开

    公开(公告)号:US20240211724A1

    公开(公告)日:2024-06-27

    申请号:US18556619

    申请日:2021-08-11

    IPC分类号: G06N3/04

    CPC分类号: G06N3/04

    摘要: Modern deep neural network (DNN) models have many layers with a single layer potentially involving large matrix multiplications. Such heavy calculation brings challenges to deploy such DNN models on a single edge device, which has relatively limited computation resources. Therefore, multiple and even heterogeneous edge devices may be required for applications with stringent latency requirements. Disclosed in the present patent documents are embodiments of a model scheduling framework that schedules multiple models on a heterogeneous platform. Multiple-model heterogeneous computing is partitioned into a neural computation optimizer (NCO) part and a neural computation accelerator (NCA) part. The migration, transition, or transformation of DNN models from cloud to edge is handled by the NCO, while the deployment of the transformed DNN models on the heterogeneous platform is handled by the NCA. Such a separation of implementation simplifies task execution and improves the flexibility for the overall framework.

    HUMAN-MACHINE INTERFACE (HMI) ARCHITECTURE
    2.
    发明申请

    公开(公告)号:US20190004510A1

    公开(公告)日:2019-01-03

    申请号:US15640867

    申请日:2017-07-03

    申请人: Baidu USA LLC

    摘要: An autonomous driving vehicle (ADV) is operated using a human-machine interface (HMI). The web server provides the HMI to a computing device in response to an input received from the computing device. An ADV command is entered into the HMI and passed to an interface of the web server. In response to receiving the ADV command, the web server calls a remote procedure call to a proxy server in a backend server for processing by a perception and control module of the ADV. Results of the ADV command are received by the web server interface and stored in results memory with a unique identifier. The web server interface opens a socket that accesses the results memory. If the results memory changes, the socket reads the results memory and provides the ADV command results to the HMI. Multiple HMIs can simultaneously communicate with the web server interface and socket.