-
公开(公告)号:US11640528B2
公开(公告)日:2023-05-02
申请号:US16660259
申请日:2019-10-22
申请人: Baidu USA LLC
发明人: Zhiyu Cheng , Baopu Li , Yingze Bao
摘要: A method for information processing for accelerating neural network training. The method includes: acquiring a neural network corresponding to a deep learning task; and performing iterations of iterative training on the neural network based on a training data set. The training data set includes task data corresponding to the deep learning task. The iterative training includes: processing the task data in the training data set using a current neural network, and determining, based on a processing result of the neural network on the task data in a current iterative training, prediction loss of the current iterative training; determining a learning rate and a momentum in the current iterative training; and updating weight parameters of the current neural network by gradient descent based on a preset weight decay, and the learning rate, the momentum, and the prediction loss in the current iterative training. This method achieves efficient and low-cost deep learning-based neural network training.
-
公开(公告)号:US11636332B2
公开(公告)日:2023-04-25
申请号:US16506519
申请日:2019-07-09
申请人: Baidu USA, LLC
发明人: Haichao Zhang , Jianyu Wang
摘要: Described herein are embodiments for a feature-scattering-based adversarial training approach for improving model robustness against adversarial attacks. Conventional adversarial training approaches leverage a supervised scheme, either targeted or non-targeted in generating attacks for training, which typically suffer from issues such as label leaking as noted in recent works. Embodiments of the disclosed approach generate adversarial images for training through feature scattering in the latent space, which is unsupervised in nature and avoids label leaking. More importantly, the presented approaches generate perturbed images in a collaborative fashion, taking the inter-sample relationships into consideration. Extensive experiments on different datasets compared with state-of-the-art approaches demonstrate the effectiveness of the presented embodiments.
-
3.
公开(公告)号:US11630953B2
公开(公告)日:2023-04-18
申请号:US16960014
申请日:2019-07-25
发明人: Hongliang Fei , Xu Li , Dingcheng Li , Ping Li
摘要: Described herein are embodiments for end-to-end reinforcement learning based coreference resolution models to directly optimize coreference evaluation metrics. Embodiments of a reinforced policy gradient model are disclosed to incorporate reward associated with a sequence of coreference linking actions. Furthermore, maximum entropy regularization may be used for adequate exploration to prevent a model embodiment from prematurely converging to a bad local optimum. Experiments on datasets compared with state-of-the-art methods verified the effectiveness of embodiments.
-
公开(公告)号:US20230115186A1
公开(公告)日:2023-04-13
申请号:US17484800
申请日:2021-09-24
申请人: BAIDU USA LLC
发明人: Tianyi GAO
IPC分类号: H05K7/20
摘要: In one embodiment, an electronic device cooling package includes an enclosure to house an electronics package, the enclosure including at least one panel with an integrated conducting plate exposed to both an inner surface of the at least one panel and an outer surface of the at least one panel. The electronic device cooling package further includes a mounting mechanism attached to the enclosure to detachably couple a cooling device to the outer surface of the at least one panel of the enclosure. Different electronics packaged within the cooling package includes different co-design of the cooling package and cooling devices.
-
公开(公告)号:US20230100305A1
公开(公告)日:2023-03-30
申请号:US17485952
申请日:2021-09-27
申请人: Baidu USA LLC
发明人: ZHIHONG PAN , DAMING LU , XI CHEN
摘要: According to one embodiment, a method is provided for video repainting performed by at least one processor of a computer device. The method includes receiving a video sequence having one or more image frames; detecting presences of a target object within the one or more image frames and determining pose condition and style shift of the detected objects; generating content representing a replacement object for the one or more image frames by applying the corresponding pose condition and style shift to the replacement object; and repainting the detected target object in the one or more image frames with the generated content.
-
公开(公告)号:US11616651B2
公开(公告)日:2023-03-28
申请号:US16315973
申请日:2019-01-04
申请人: Baidu USA LLC , Baidu.com Times Technology (Beijing) Co., Ltd. , KUNLUNXIN TECHNOLOGY (BEIJING) COMPANY LIMITED
发明人: Yong Liu , Yueqiang Cheng , Jian Ouyang , Tao Wei
摘要: According to one embodiment, a system receives, at a host channel manager (HCM) of a host system, a request from an application to establish a secure channel with a data processing (DP) accelerator, where the DP accelerator is coupled to the host system over a bus. In response to the request, the system generates a first session key for the secure channel based on a first private key of a first key pair associated with the HCM and a second public key of a second key pair associated with the DP accelerator. In response to a first data associated with the application to be sent to the DP accelerator, the system encrypts the first data using the first session key. The system then transmits the encrypted first data to the DP accelerator via the secure channel over the bus.
-
公开(公告)号:US20230089909A1
公开(公告)日:2023-03-23
申请号:US17483704
申请日:2021-09-23
申请人: Baidu USA LLC
发明人: Tianyi Gao
IPC分类号: H05K7/20
摘要: A fluid distribution apparatus for servers mounted onto a rack. The apparatus includes one or more pairs of distribution units that are rotationally symmetric and are attached to each other to form the pair. Each of the distribution units includes a main supply port and main return port that are connected to the rack's fluid manifolds. Each of the distribution units also includes plurality of supply and return connectors that connect to the servers. A pump in each of the distribution units delivers fluid from the rack's fluid manifolds to the servers for cooling. Each server is connected to fluid ports on both distribution units, so as to provide full redundancy. A controller receives signals form the servers and sends operation signals to the pumps to control the fluid delivery for cooling the servers.
-
公开(公告)号:US11612081B2
公开(公告)日:2023-03-21
申请号:US17409657
申请日:2021-08-23
申请人: Baidu USA LLC
发明人: Tianyi Gao
IPC分类号: H05K7/20
摘要: A two-phase immersion cooling system for cooling electronics. The electronics are immersed in immersion tank filled with phase change liquid. As liquid evaporates due to heat generated by the electronics, it enters a vapor passageway that leads the vapor to a condenser situated remotely from the immersion tank. Upon condensing at the condenser, the condensed liquid is directed to a resupply tank, wherein the condensed liquid cools. When the level of the two phase liquid in the immersion tank drops below a set threshold, a pump is activated to deliver the condensed liquid from the resupply tank to the immersion tank. The immersion tank, vapor passageway and condenser are position inside a containment passageway. The containment passageway captures any vapor not entering the vapor passageway and direct such vapor to the condenser. An air mover generates pressure differential within the containment passageway to direct the vapor towards the condenser.
-
公开(公告)号:US11609342B2
公开(公告)日:2023-03-21
申请号:US16649559
申请日:2020-03-05
发明人: Yan Cui , Quanwei Liu , Yao Zhuo
摘要: In one embodiment, a method for calculating a location of an autonomous driving vehicle includes receiving new global navigation satellite system (GNSS) data. The method further includes identifying a first previously estimated location from a plurality of previously estimated locations with a timestamp that is closest to the timestamp of the new GNSS data and identifying a second previously estimated location from the plurality of previously estimated locations with a most recent timestamp. The method further includes calculating a difference between the first previously estimated location and the second previously estimated location, adjusting the new GNSS data based on the difference; and calculating a current estimated location of the ADV based on the adjusted GNSS data.
-
公开(公告)号:US20230084203A1
公开(公告)日:2023-03-16
申请号:US17846555
申请日:2022-06-22
申请人: Baidu USA, LLC
发明人: Baopu LI , Qiuling SUO , Yuchen BIAN
摘要: Model pruning is used to trim large neural networks, like convolutional neural networks (CNNs), to reduce computation overheads. Existing model pruning methods mainly rely on heuristics rules or local relationships of CNN layers. A novel hypernetwork based on graph neural network is disclosed for generating and evaluating pruned networks. A graph is first constructed according to information flow of channels and layers in a CNN network, with channels and layers represented as nodes and information flows represented as edges. A graph neural network is applied to aggregate both local and global dependencies across all channels and layers of the CNN network, resulting in informative node embeddings. With such embeddings, pruned CNN networks including their architectures and weights may be effectively generated and evaluated.
-
-
-
-
-
-
-
-
-