公开(公告)号：US11477078B2

公开(公告)日：2022-10-18

申请号：US17215709

申请日：2021-03-29

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Wei Wei ,  Chen Huang

IPC: H04L67/101 ,  H04L41/0816 ,  H04L41/0826 ,  H04L67/1031 ,  H04L67/1097

Abstract: A data transmission method includes receiving a service request sent by a terminal device. The method also includes determining, based on the service request, a target transmission path of service data corresponding to the service request and a change policy corresponding to the target transmission path. The method further includes changing, according to the change policy, a current content delivery network (CDN) to a target CDN. The target CDN includes a plurality of serving nodes corresponding to the target transmission path. The method additionally includes sending an indication message to a first serving node in the target CDN to indicate the first serving node to send the service data to the terminal device through the target transmission path; or sending an indication message to a second serving node in the target CDN to indicate the second serving node to send the service data to the terminal device.

公开(公告)号：US12271819B2

公开(公告)日：2025-04-08

申请号：US17371590

申请日：2021-07-09

Applicant: Huawei Technologies Co., Ltd.

Inventor： Jiafeng Zhu ,  Wei Wei ,  Jianle Chen ,  Wei Wang ,  Jie Shen

IPC: G06N3/082 ,  G06N3/04 ,  H03M7/30 ,  G06F11/34

Abstract: A first aspect relates to a computer-implemented method for performing model compression. The method includes compressing a machine learning (ML) network model comprising a multiple layer structure to produce a compressed ML network model. The compressed ML network model maintains the multiple layer structure of the ML network model. The method generates a model file for the compressed ML network model. The model file includes the compressed ML network model and decoding information for enabling the ML network model to be decompressed and executed layer-by-layer.

公开(公告)号：US20170060752A1

公开(公告)日：2017-03-02

申请号：US15347776

申请日：2016-11-09

Applicant: HUAWEI TECHNOLOGIES CO.,LTD.

Inventor： Wei Wei ,  Lixin Zhang ,  Jin Xiong ,  Dejun Jiang

IPC: G06F12/0868 ,  G06F3/06 ,  G06F12/128 ,  G06F12/122

CPC classification number: G06F12/0868 ,  G06F3/061 ,  G06F3/0619 ,  G06F3/0631 ,  G06F3/065 ,  G06F3/068 ,  G06F3/0685 ,  G06F12/08 ,  G06F12/122 ,  G06F12/128 ,  G06F2212/604

Abstract: A data caching method and a computer system are provided. In the method, when a miss of an access request occurs and a cache needs to determine a to-be-replaced cache line, not only a historical access frequency of the cache line but also a type of a memory corresponding to the cache line needs to be considered. A cache line corresponding to a DRAM type may be preferably replaced, which reduces a caching amount in the cache for data stored in a DRAM and relatively increase a caching amount for data stored in an NVM., For an access request for accessing the data stored in the NVM, corresponding data can be found in the cache whenever possible, thereby reducing cases of reading data from the NVM. Thus, a delay in reading data from the NVM is reduced, and access efficiency is effectively improved.

Abstract translation: 提供了数据缓存方法和计算机系统。 在该方法中，当出现访问请求的错过并且高速缓存需要确定待替换的高速缓存行时，不仅高速缓存行的历史存取频率，而且与高速缓存行对应的存储器的类型也需要 被考虑。 可以优选地替换对应于DRAM类型的高速缓存行，这降低了存储在DRAM中的数据的高速缓存中的缓存量，并且相对地增加了存储在NVM中的数据的高速缓存量。对于访问所存储的数据的访问请求 在NVM中，尽可能在缓存中找到相应的数据，从而减少从NVM读取数据的情况。 因此，降低了从NVM读取数据的延迟，并且有效地提高了访问效率。

公开(公告)号：US20250037573A1

公开(公告)日：2025-01-30

申请号：US18915208

申请日：2024-10-14

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Wei Chen ,  Liangbing Lu ,  Peng Zhang ,  Jingjing Guo ,  Wei Wei

IPC: G08G1/005 ,  G01C21/36 ,  G06V20/50 ,  G09B21/00

Abstract: An image processing method. The method is applied to an electronic device. The electronic device is configured execute instructions to obtain a plurality of frames of images of a front angle of view of a user and generally to perform a method that includes: determining a first route if a crosswalk is detected in the plurality of frames of images, where the first route is a safe walking route on the crosswalk; obtaining, based on the plurality of frames of images, a second route that the user actually walks; and outputting first indication information to the user if an included angle between the first route and the second route in a same reference system is greater than or equal to a preset angle, where the first indication information indicates that the second route is abnormal.

公开(公告)号：US10241919B2

公开(公告)日：2019-03-26

申请号：US15347776

申请日：2016-11-09

Applicant: HUAWEI TECHNOLOGIES CO.,LTD.

Inventor： Wei Wei ,  Lixin Zhang ,  Jin Xiong ,  Dejun Jiang

IPC: G06F12/0868 ,  G06F3/06 ,  G06F12/122 ,  G06F12/128 ,  G06F12/08

Abstract: A data caching method and a computer system are provided. In the method, when a miss of an access request occurs and a cache needs to determine a to-be-replaced cache line, not only a historical access frequency of the cache line but also a type of a memory corresponding to the cache line needs to be considered. A cache line corresponding to a DRAM type may be preferably replaced, which reduces a caching amount in the cache for data stored in a DRAM and relatively increase a caching amount for data stored in an NVM. For an access request for accessing the data stored in the NVM, corresponding data can be found in the cache whenever possible, thereby reducing cases of reading data from the NVM. Thus, a delay in reading data from the NVM is reduced, and access efficiency is effectively improved.

公开(公告)号：US20150317246A1

公开(公告)日：2015-11-05

申请号：US14798094

申请日：2015-07-13

Applicant: Huawei Technologies Co., Ltd.

Inventor： Fei Xia ,  Dejun Jiang ,  Wei Wei ,  Jin Xiong

IPC: G06F12/02

CPC classification number: G06F12/0253 ,  G06F12/0238 ,  G06F12/0246 ,  G06F2212/202 ,  G06F2212/702 ,  G06F2212/7205 ,  G06F2212/7211

Abstract: A memory reclamation method and apparatus are disclosed. The memory reclamation method is performed by a processor for reclaiming memory pages of a non-volatile memory (NVM) in a terminal device. In the method, the processor receives a memory reclamation request message including a reclamation identifier which is used to indicate a quantity of memory pages requested to be reclaimed. Then, the processor reclaims inactive memory pages of the NVM according to the reclamation identifier in ascending order of the write times of inactive memory pages of the NVM. According to the memory reclamation method, a page having a relatively small quantity of inactive memory page write times of the NVM is reclaimed first, and a page having a relatively large quantity of inactive memory page write times of the NVM is reclaimed later.

Abstract translation: 公开了一种存储器回收方法和装置。 存储器回收方法由用于回收终端设备中的非易失性存储器（NVM）的存储器页面的处理器执行。 在该方法中，处理器接收包括回收标识符的存储器回收请求消息，该回收标识符用于指示请求回收的存储器页面的数量。 然后，处理器根据回收标识符以NVM的非活动存储器页面的写入时间的升序顺序回收NVM的非活动存储器页面。 根据存储器回收方法，首先回收具有相对少量NVM的非活动存储器页面写入时间的页面，并且稍后再现具有NVM的相对大量非活动存储器页面写入时间的页面。

公开(公告)号：US20210342694A1

公开(公告)日：2021-11-04

申请号：US17371590

申请日：2021-07-09

Applicant: Huawei Technologies Co., Ltd.

Inventor： Jiafeng Zhu ,  Wei Wei ,  Jianle Chen ,  Wei Wang ,  Jie Shen

IPC: G06N3/08 ,  G06N3/04

Abstract: A first aspect relates to a computer-implemented method for performing model compression. The method includes compressing a machine learning (ML) network model comprising a multiple layer structure to produce a compressed ML network model. The compressed ML network model maintains the multiple layer structure of the ML network model. The method generates a model file for the compressed ML network model. The model file includes the compressed ML network model and decoding information for enabling the ML network model to be decompressed and executed layer-by-layer.

公开(公告)号：US12099392B2

公开(公告)日：2024-09-24

申请号：US17014371

申请日：2020-09-08

Applicant: Huawei Technologies Co., Ltd.

Inventor： Wei Wei ,  Zhou Zhou ,  Yuwei Pu

IPC: G06F1/08 ,  G06F1/28 ,  G06F119/06

CPC classification number: G06F1/28 ,  G06F1/08 ,  G06F2119/06

Abstract: A dynamic power consumption estimation method includes recording a statistical value of each of n target signals of a to-be-measured target in a unit time, where the target signal is a clock signal obtained after clock gating and the statistical value is a quantity of times toggling, or the target signal is a clock gating signal and the statistical value is a quantity of enable cycles, and calculating the dynamic power consumption of the to-be-measured target in the unit time based on the statistical value of each of the n target signals in the unit time.

公开(公告)号：US09528907B2

公开(公告)日：2016-12-27

申请号：US14334320

申请日：2014-07-17

Applicant: Huawei Technologies Co., Ltd.

Inventor： Yong Cheng ,  Wei Wei ,  Yong Wang

IPC: G01M11/00 ,  H04J3/06 ,  H04J3/14 ,  H04B10/077

CPC classification number: G01M11/30 ,  H04B10/0773 ,  H04J3/0673 ,  H04J3/14

Abstract: A method for detecting symmetry of optical fibers and a device thereof are provided. The method includes: receiving, by a boundary clock, a first timestamp message carrying a first timestamp via a slave port, and a second timestamp message carrying a second timestamp via a passive port; and determining, by the boundary clock, symmetry of optical fibers according to at least the first timestamp and the second timestamp. The method and device for detecting symmetry of optical fibers provided by the present disclosure reduce the cost of detecting symmetry of optical fibers in the conventional technology, and improve the efficiency of detecting symmetry of optical fibers in the conventional technology.

Abstract translation: 提供了一种用于检测光纤对称性的方法及其装置。 该方法包括：通过边界时钟接收经由从端口携带第一时间戳的第一时间戳消息和经由被动端口携带第二时间戳的第二时间戳消息; 以及根据所述边界时钟，至少根据所述第一时间戳和所述第二时间戳来确定光纤的对称性。 本发明提供的用于检测光纤的对称性的方法和装置降低了传统技术中检测光纤对称性的成本，提高了传统技术中光纤对称性检测的效率。

公开(公告)号：US09326261B2

公开(公告)日：2016-04-26

申请号：US14154749

申请日：2014-01-14

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Wei Wei

IPC: H04J3/06 ,  H04W56/00 ,  H04W4/18 ,  H04L29/08

CPC classification number: H04W56/001 ,  H04L67/2842 ,  H04L67/288 ,  H04W4/18

Abstract: Embodiments of the present invention include a method for synchronizing a popularity value of distributed cache data, a method, an apparatus, and a system for cooperative and distributed caching. The synchronizing method includes: receiving, by a first base station, a request for accessing a packet domain service sent by a user equipment; if the first base station determines that the packet domain service is locally cached, sending the packet domain service to the user equipment and incrementing a current popularity value of the packet domain service in a list of locally cached content of the first base station according to a preset function; and sending, when the current popularity value of the packet domain service is accumulated to a preset threshold, the current popularity value to a popularity information synchronizing node for synchronization, and clearing the current popularity value locally cached. The present invention saves a transmission bandwidth of an IUB interface.

Abstract translation: 本发明的实施例包括用于同步分布式高速缓存数据的普及值的方法，方法，装置和用于协作和分布式缓存的系统。 所述同步方法包括：由第一基站接收对用户设备发送的分组域服务的访问请求; 如果第一基站确定分组域服务是本地缓存的，则将分组域服务发送到用户设备，并且根据第一基站本地缓存内容的列表增加分组域服务的当前流行度值 预置功能; 并且当将分组域服务的当前流行度值累积到预设阈值时，将当前流行度值发送到流行信息同步节点进行同步，并且清除当前流行度值本地缓存。 本发明节省了IUB接口的传输带宽。