Patents Assigned to BAIDU USA LLC
  • Patent number: 12014398
    Abstract: Deep neural network (DNN) models have been widely used for user-relevance content prediction. Presented herein are embodiments of a new user-relevance framework, which may be referred as Gating-Enhanced Multi-task Neural Networks (GemNN) embodiments. Neural network-based multi-task learning model embodiments herein predict user engagement with content in a coarse-to-fine manner, which gradually reduces content candidates and allows parameter sharing from upstream tasks to downstream tasks to improve the training efficiency. Also, in one or more embodiments, a gating mechanism was introduced between embedding layers and multi-layer perceptions to learn feature interactions and control the information flow fed to MLP layers. Tested embodiments demonstrated considerable improvements over prior approaches.
    Type: Grant
    Filed: July 7, 2021
    Date of Patent: June 18, 2024
    Assignees: Baidu USA LLC, Baidu.com Times Technology (Beijing) Co., Ltd.
    Inventors: Hongliang Fei, Jingyuan Zhang, Xingxuan Zhou, Junhao Zhao, Banghu Yin, Ping Li
  • Patent number: 12013193
    Abstract: A liquid manifold can be assembled to an information technology (IT) rack to deliver and distribute fluid to IT equipment. The manifold can include a plurality of sections, each of the plurality of sections having one or more shut-off valves. One or more leak detection sensors can be arranged to detect leaks in any of the sections and in any of the IT equipment. A controller can control a shut-off valve to a closed position based on a detected leak. The design enables the manifold to better manage and control the fluid for mission critical IT equipment.
    Type: Grant
    Filed: June 7, 2021
    Date of Patent: June 18, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 12008467
    Abstract: Presented herein are embodiments of an improved asymmetric quantization, which may generally be referred to as improved asymmetric quantization (IAQ) embodiments. IAQ embodiments combine the benefits of conventional asymmetric quantization and symmetric quantization but also provide additional computation efficiencies. Embodiments of IAQ adopt an asymmetric range of the weights of a neural network layer, so they circumvent the limitation of symmetric range of symmetric quantization. Moreover, the inference process of a neural network quantized by an IAQ embodiment is much faster than that of the neural network quantized by conventional asymmetric quantization by quantizing an offset value of each layer.
    Type: Grant
    Filed: May 19, 2020
    Date of Patent: June 11, 2024
    Assignee: Baidu USA LLC
    Inventors: Yingzhen Yang, Zhibiao Zhao, Baoxin Zhao, Jun Huan, Jian Ouyang, Yong Wang, Jiaxin Shi
  • Patent number: 12010815
    Abstract: A cooling module includes a first cooling plate having a first internal channel and a second cooling plate having a second internal channel. The cooling module includes an interconnect frame coupled in between the first and second cooling plates, the interconnect frame includes a third internal channel that connects the first internal channel to the second internal channel. The cooling module includes a first injection plate attached to a bottom portion of the first cooling plate and a second injection plate attached to a bottom portion of the second cooling plate, the first and second injection plates manage a distribution of coolant fluid to dedicated areas of electronic chips adjacent to the first and second injection plates. The cooling module includes a first pump frame coupled to an inlet port at the first cooling plate, a first pump disposed at the first pump frame to directly intake a coolant fluid.
    Type: Grant
    Filed: March 18, 2022
    Date of Patent: June 11, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Publication number: 20240185386
    Abstract: Image super-resolution (SR) refers to the process of recovering high-resolution (HR) images from low-resolution (LR) inputs. Blind image SR is a more challenging task which involves unknown blurring kernels and characterizes the degradation process from HR to LR. In the present disclosure, embodiments of a variational autoencoder (VAE) are leveraged to train a kernel autoencoder for more accurate degradation representation and more efficient kernel estimation. In one or more embodiments, a kernel-agnostic loss is used to learn more robust kernel features in the latent space from LR inputs without using ground-truth kernel references. In addition, attention-based adaptive pooling is introduced to improve kernel estimation accuracy, and spatially non-uniform kernel features are passed into SR restoration resulting in additional kernel estimation error tolerance.
    Type: Application
    Filed: September 30, 2021
    Publication date: June 6, 2024
    Applicants: Baidu USA LLC, Baidu.com Times Technology (Beijing) Co., Ltd.
    Inventors: Zhihong PAN, Baopu LI, Dongliang HE, Wenhao WU, Tianwei LIN
  • Publication number: 20240185587
    Abstract: Modem 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. Two different approaches, model first scheduling (MFS) and hardware first scheduling (HFS), are presented to allocate a group of models for a service into corresponding heterogeneous edge devices, including CPU, VPU and GPU. Experimental results prove the effectiveness of the MFS and HFS methods for improving the inference speed of single and multiple AI-based services.
    Type: Application
    Filed: August 16, 2021
    Publication date: June 6, 2024
    Applicants: Baidu.com Times Technology (Beijing) Co., Ltd., Baidu USA LLC
    Inventors: Haofeng KOU, Xing LI, Huimeng ZHENG, Lei WANG, Zhen CHEN
  • Patent number: 11999064
    Abstract: Embodiments of a learning-based excavation planning method are disclosed for excavating rigid objects in clutter, which is challenging due to high variance of geometric and physical properties of objects, and large resistive force during the excavation. A convolutional neural network is utilized to predict a probability of excavation success. Embodiments of a sampling-based optimization method are disclosed for planning high-quality excavation trajectories by leveraging the learned prediction model. To reduce simulation-to-real gap for excavation learning, voxel-based representations of an excavation scene are used. Excavation experiments were performed in both simulation and real world to evaluate the learning-based excavation planners. Experimental results show that embodiments of the disclosed method may plan high-quality excavations for rigid objects in clutter and outperform baseline methods by large margins.
    Type: Grant
    Filed: July 20, 2021
    Date of Patent: June 4, 2024
    Assignee: Baidu USA LLC
    Inventors: Qingkai Lu, Liangjun Zhang
  • Patent number: 11997831
    Abstract: A connector assembly may include a plate to be attached to a server chassis of an electronic rack; a connection module; and a moveable alignment module having a guiding structure and configured to be coupled to the plate via the connection module. The moveable alignment module may house one or more fluid blind mate connectors that fluidly connect to one or more cold plates that are thermally coupled to one or more electronic server components of the server chassis. The connection module may enable the moveable alignment module to be moveable when the guiding structure engages a distribution manifold of the electronic rack to align and couple the one or more fluid blind mate connectors to either at least one of a supply line connector or a return line connector of the distribution manifold, thereby enabling fluid communication between the server chassis and the distribution manifold.
    Type: Grant
    Filed: December 21, 2021
    Date of Patent: May 28, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11996708
    Abstract: A first battery of a battery system can include a thermoelectric component (TEC) that produces electric energy from thermal energy that the first battery generates. The TEC is used to charge a second battery of the battery system, while maintaining proper thermal conditions for the first battery. The battery system can be used to support information technology (IT) equipment by acting as backup power during a power outage, and/or to provide supplemental power under peak power conditions.
    Type: Grant
    Filed: March 31, 2021
    Date of Patent: May 28, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11990757
    Abstract: The power supply system comprises an inter-system bus, one or more renewable energy systems and one or more second switches, where each second switch is coupled to one renewable energy system and to connect to the corresponding renewable energy system to the inter-system bus. The power supply system further comprises one or more current detection circuits, where each current detection circuit is coupled to one renewable energy system to detect an output current of the corresponding renewable energy system. The power supply system further comprises a central controller coupled to the inter-system bus and configured to, in response to the output current of the corresponding renewable energy system is higher than a predetermined threshold current, activate a corresponding second switch to connect the corresponding renewable energy system to the inter-system bus to provide a renewable power to at least one of the one or more server clusters, or storage systems.
    Type: Grant
    Filed: January 25, 2021
    Date of Patent: May 21, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11989233
    Abstract: Presented herein are embodiments of a fast search on graph methodology for Maximum Inner Product Search (MIPS). This optimization problem is challenging since traditional Approximate Nearest Neighbor (ANN) search methods may not perform efficiently in the nonmetric similarity measure. Embodiments herein are based on the property that a Möbius/Möbius-like transformation introduces an isomorphism between a subgraph of 2-Delaunay graph and Delaunay graph for inner product. Under this observation, embodiments of a novel graph indexing and searching methodology are presented to find the optimal solution with the largest inner product with the query. Experiments show significant improvements compared to existing methods.
    Type: Grant
    Filed: September 27, 2020
    Date of Patent: May 21, 2024
    Assignee: Baidu USA LLC
    Inventors: Shulong Tan, Zhixin Zhou, Zhaozhuo Xu, Ping Li
  • Patent number: 11980008
    Abstract: A server chassis can include a first pair of connectors on a rear side of the server chassis arranged to face in a rear direction to engage with connectors of a server rack. Upon engaging of the connectors, the server chassis fluidly connects to a supply line and a return line of the server rack. The server chassis includes a second pair of connectors on a front side of the server chassis, to fluidly connect to a cooling system of a server node housed within the server chassis. The second pair of connectors are held to the server chassis with one or more movable attachments that enable coupling and decoupling of the second pair of connectors to the cooling system of the server node. Fluid channels fluidly connect the first pair of connectors to the second pair of fluid connectors.
    Type: Grant
    Filed: March 21, 2022
    Date of Patent: May 7, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11980007
    Abstract: A connector assembly may include a distribution frame having an adjusting channel; and one or more distribution modules disposed and slidably moveable along the adjusting channel. Each of the one or more distribution modules may include an inlet connector to receive cooling liquid from a cooling liquid loop, a fluid blind mate connector to-connect the cooling liquid loop to one or more cold plates of a server chassis, and a guiding element having a guiding structure and configured to house the fluid blind mate connector. When the guiding element contacts with the server chassis, the guiding element causes the corresponding distribution module to move slightly to align the fluid blind mate connector with a server connector of the server chassis, such that the fluid blind mate connector connects with the server connector to distribute the cooling liquid to the server chassis.
    Type: Grant
    Filed: December 21, 2021
    Date of Patent: May 7, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11980006
    Abstract: According to one embodiment, a system for a rack includes an equipment module for coupling to a rack module. The rack module includes a pair of main connectors for coupling to a coolant source and a pair of manifold connectors, and includes a vertical leak detection channel that has a leak detection cable. Positioned between the main connectors and the manifold connectors is a structure that at least partially contains the main connectors and has a liquid path to the channel. The equipment module includes a chassis that has a pair of connectors and a conducting path. The chassis also has an opening for receiving the manifold connectors for coupling to the connectors, and another opening, the conducting path extends from an interior of the chassis to the other opening and is arranged to guide liquid out of the opening and into the channel while both modules are coupled together.
    Type: Grant
    Filed: September 1, 2021
    Date of Patent: May 7, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11978913
    Abstract: According to one embodiment, a battery management system includes a BBU shelf with bus connectors; a battery module with battery cell packages, each package including battery cells; and a converter module having 1) a set of converters, 2) a first busbar to which each converter is connected, the first busbar includes several switches, each switch is disposed between adjacently connected converters, 3) a second busbar to which each converter is connected via a first switch and to which each package is connected via a second switch, the second busbar comprises several switches, each switch is disposed between adjacently connected converters, packages, or a combination thereof, and 4) a third busbar to which each converter is connected and to which each bus connector is connected via a third switch, the third busbar includes several switches, each switch is disposed between adjacently connected converters, bus connectors, or a combination thereof.
    Type: Grant
    Filed: April 30, 2020
    Date of Patent: May 7, 2024
    Assignee: BAIDU USA LLC
    Inventors: Huawei Yang, Tianyi Gao
  • Publication number: 20240131705
    Abstract: Presented herein are embodiments of a two-stage methodology that integrates data-driven imitation learning and model-based trajectory optimization to generate optimal trajectories for autonomous excavators. In one or more embodiments, a deep neural network using demonstration data to mimic the operation patterns of human experts under various terrain states, including their geometry shape and material type. A stochastic trajectory optimization methodology is used to improve the trajectory generated by the neural network to ensure kinematics feasibility, improve smoothness, satisfy hard constraints, and achieve desired excavation volumes. Embodiments were tested on a Franka robot arm equipped with a bucket end-effector. Embodiments were also evaluated on different material types, such as sand and rigid blocks. Experimental results showed that embodiments of the two-stage methodology that comprises combining expert knowledge and model optimization increased the excavation weights by up to 24.
    Type: Application
    Filed: October 10, 2022
    Publication date: April 25, 2024
    Applicant: Baidu USA LLC
    Inventors: Zhixian YE, Qiangqiang GUO, Liyang WANG, Liangjun ZHANG
  • Patent number: 11968803
    Abstract: A cooling plate includes an inlet port to receive two-phase fluid, a base plate to be attached to server electronics for cooling, a number of heat spreading structures disposed and spaced apart on the base plate to extend a heat exchange area from the base plate, and a number of fluid channels disposed on the base plate, where each of the fluid channels is situated between a pair of adjacent heat spreading structures. Each of the fluid channels include a channel inlet to receive at least a portion of the two-phase fluid from the inlet port, a channel outlet to allow the received portion of the two-phase fluid to flow through and exit the corresponding fluid channel, and one or more intermediate outlets disposed between the channel inlet and the channel outlet to allow at least a portion of the received two-phase fluid to exit in vapor or liquid form.
    Type: Grant
    Filed: December 22, 2021
    Date of Patent: April 23, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11963340
    Abstract: A frame for a liquid-cooled chassis in an IT rack, can include a fluid supply line contained within one or more walls of the frame and a fluid return line contained within the one or more walls of the frame. Fluid distribution function and hardware are integrated to the frame. One or more contact pads can be located on an external surface of the frame, for transferring thermal energy with IT equipment.
    Type: Grant
    Filed: June 18, 2021
    Date of Patent: April 16, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11963332
    Abstract: Methods, systems, and devices for managing coolant in data centers are disclosed. Coolant may be used in data centers to regulate the temperatures of computing devices. The disclosed methods and system may provide for containment of vapor and condensation of the vapor into coolant. The coolant may be used as part of a two phase cooling system to cool the computing devices. The vapor may be generated when the coolant removes heat from computing devices during operation while at least partially submerged in the coolant. A mobile condenser is used for condensing the vapor within the containment.
    Type: Grant
    Filed: September 8, 2021
    Date of Patent: April 16, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao
  • Patent number: 11960335
    Abstract: A leak prevention system includes a liquid cooling device for a server. The liquid cooling device includes a fluid area configured to receive a cooling fluid. The leak prevention system also includes at least one pair of detection wires disposed within the liquid cooling device and at least partially surrounding the fluid area. The leak prevention system also includes a sensor connected to the detection wires and configured to detect when a cooling fluid contacts the detection wires.
    Type: Grant
    Filed: March 26, 2021
    Date of Patent: April 16, 2024
    Assignee: BAIDU USA LLC
    Inventor: Tianyi Gao