Abstract: Embodiments of a generative framework comprise cooperative learning of two generative flow models, in which the two models are iteratively updated based on the jointly synthesized examples. In one or more embodiments, the first flow model is a normalizing flow that transforms an initial simple density into a target density by applying a sequence of invertible transformations, and the second flow model is a Langevin flow that runs finite steps of gradient-based MCMC toward an energy-based model. In learning iterations, synthesized examples are generated by using a normalizing flow initialization followed by a short-run Langevin flow revision toward the current energy-based model. Then, the synthesized examples may be treated as fair samples from the energy-based model and the model parameters are updated, while the normalizing flow directly learns from the synthesized examples by maximizing the tractable likelihood. Also provided are both theoretical and empirical justifications for the embodiments.

Abstract: In one embodiment, a light detection and range (LIDAR) device includes an array of light transmitting and receiving (TX/RX) units arranged to sense a physical range associated with a target. Each of the light TX/RX units includes a mounting board having a light pass-through opening and a light emitter mounted adjacent to the light pass-through opening. The light emitter is configured to emit a light beam towards the target according to a transmitting path. The LIDAR device further includes a light detector positioned behind the mounting board to receive at least a portion of the light beam reflected from the target through the light pass-through opening according to a light receiving path. The light transmitting path and the light receiving path are substantially in parallel and close to each other.

Abstract: A fluid deployment unit includes an expandable container containing mixed fluids in a gaseous region and a liquid region, where the expandable container includes a gas-out port, a liquid-out port, a gas-in port, and a liquid-in port. The fluid deployment unit includes a first three-way valve having a first port coupled to the liquid-out port, a second port coupled to the gas-out port, and a third port matable to an inlet of an electronic rack. The fluid deployment unit includes a second three-way valve having a first port matable to an input port of a liquid-to-liquid exchange unit of a testing assistant unit, a second port coupled to the gas-in port, and a third port matable to an outlet of the electronic rack, where the liquid-in port of the expandable container is matable to an output port of the liquid-to-liquid exchange unit.

Abstract: Presented herein are systems and methods for an end-to-end solution for object removal from a video using adaptive learning. In one or more embodiments, using a pre-trained image inpainting model, in-scene training data may be generated using optical-flow guided sampling. In one or more embodiments, the sampled patches are used to generate a training dataset, which is used to further train the image inpainting model until reaching a stop condition. The adaptively trained inpainting model may be used to generate a modified video in which the desired object or objects have been removed and the corresponding removed portion(s) have been filled (or inpainted) to preserve the integrity of the image.

Abstract: An electronic rack includes a rack manifold to be coupled to an external cooling fluid source, including a supply rack manifold and a return rack manifold, wherein the rack manifold includes a plurality of pairs of rack connectors disposed thereon. The electronic rack further includes a server chassis including a connector holder having a pair of a supply server connector and a return server connector to be connected with a corresponding pair of rack connectors of the rack manifold. The electronic rack further includes a controller, in response to detecting a leakage of the cooling fluid, configured to cause the supply server connector to disconnect from the supply rack manifold, while maintaining the return server connector connected with the return rack manifold, and to increase a flowrate of the cooling fluid on the return rack manifold to remove the cooling fluid residing within the server chassis.

Abstract: A battery system can provide backup power for information technology (IT) equipment. In response to a lithium ion based battery being inactive (not charging or discharging), a temperature of the battery can be maintained at or below an optimal storage temperature of the battery, using a primary cooling system. If the primary cooling system is insufficient, the temperature can be maintained at or below the optimal storage temperature with a secondary cooling system that runs in addition to the primary system. The optimal storage temperature of the battery is determined based on an effort to cool the battery and a degradation of the battery.

Abstract: According to one embodiment, a thermal management system includes a thermoelectric cooling (TEC) element mounted on an information technology (IT) component, wherein the element draws a first power to cool the component; a load sensor to sense a second power that is drawn by the component to perform a first set of operations; a temperature sensor to sense a temperature of the component; and a controller to determine whether the second power drawn has changed based on the load sensor, wherein the component performs a second set of operations while drawing the second power, and in response to the changed second power, adjust the power supplied to the element and component based on the temperature and the power supplied to the component, such that the component continues to perform the second set of operations while drawing the adjusted power that is less than the power drawn before the adjustment.

Abstract: 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.

Abstract: Described herein are embodiments of a reinforcement learning based large-scale multi-objective ranking system. Embodiments of the system may be used for optimizing short-video recommendation on a video sharing platform. Multiple competing ranking objective and implicit selection bias in user feedback are the main challenges in real-world platform. In order to address those challenges, multi-gate mixture of experts (MMoE) and soft actor critic (SAC) are integrated together into a MMoE_SAC system. Experiment results demonstrate that embodiments of the MMoE_SAC system may greatly reduce a loss function compared to systems only based on single strategies.

Abstract: A modular cooling system for data center. An airflow section forms a duct for air flow and a plurality of core units are serially attached to each other and to the airflow section. A blower unit is attached to each of the core units. A plurality of motorized dampers are provided: between each of the core units and the airflow unit, in between each two core units, and between each core unit and its corresponding blower unit. A plurality of fluid ports are attached to each of the core units. At least one of the core units is loaded with one or more equipment selected from: air filter, humidifier, dehumidifier, heat exchanger, evaporator, condenser, chiller, computer room air conditioner (CRAC), dry cooler, a cooling tower or other types of cooling equipment. A combination operation of the components on the compartment and the cooling units enables fast deployment and operation.

Abstract: An electronic cooling system is disclosed. The system includes a plurality of cooling plates to extract heat from their respective heat sources. The system further includes one or more vapor separators for extracting vapor from the liquid, with each vapor separator to receive mixed phase liquid and separate the mixed phase liquid into vapor and cooling liquid. The system further includes a return unit to receive the vapors from the vapor separators through one or more vapor loops, and dissipate the received vapors to an external cooling loop. The cooling plates include a first cooling plate that receives liquid phase cooling liquid to extract heat from a first heat source and produces first mixed phase liquid. The cooling plates further include a second cooling plate that uses cooling liquid from a vapor separator to extract heat from a second heat source, produces second mixed phase liquid, and supplies the second mixed phase liquid to the return unit.

Abstract: Systems and methods for managing airflow in a backup battery unit (BBU) rack are described in the disclosure. In one embodiment, a system includes a BBU rack with a number of BBU modules, the BBU rack configured to power a server rack in a data center. The system further includes one or more crossflow fans, each crossflow fan configured to dynamically adjust its air blowing direction in real time; and a rack management controller that are connected to the server rack, the BBU rack and the one or more crossflow fans. The crossflow fan is equipped with a rotatable frame for airflow variations. The one or more crossflow fans to diffuse cooling air into one or more of the BBU modules in response to a power supply incident after receiving control signals from either a rack management controller and/or the BBU rack controller.

Abstract: In one embodiment, a set of dynamic vehicle parameters of an ADV associated with a road location are determined based on outputs of an IMU of the ADV measured when the ADV is traveling through the road location. Whether the set of dynamic vehicle parameters satisfy one of a first set of criteria and a second set of criteria is determined. In response to the dynamic vehicle parameters satisfying the first set of criteria for a first predetermined quantity of times or satisfying the second set of criteria for a second predetermined quantity of times, the speed limit associated with the road location is adjusted within a limited range spanning from a minimum speed limit to a maximum speed limit. Operations of the ADV when the ADV subsequently travels through the road location are controlled based at least in part on the adjusted speed limit.

Abstract: Approximate nearest neighbor (ANN) searching is a fundamental problem in computer science with numerous applications in area such as machine learning and data mining. For typical graph-based ANN methods, the searching method is executed iteratively, and the execution dependency prohibits graphics processor unit (GPU)/GPU-type processor adaptations. Presented herein are embodiments of a novel framework that decouples the searching on graph methodology into stages, in order to parallel the performance-crucial distance computation. Furthermore, in one or more embodiments, to obtain better parallelism on GPU-type components, also disclosed are novel ANN-specific optimization methods that eliminate dynamic memory allocations and trade computations for less memory consumption. Embodiments were empirically compared against other methods, and the results confirm the effectiveness.

Abstract: Embodiments disclose a method and a system to operate an autonomous driving vehicle (ADV). In one embodiment, a system determines a source lane speed limit (SLSL) for a source lane (SL) and a target lane speed limit (TLSL) for a target lane (TL), where the SLSL is greater than the TLSL. The system generates a first trajectory for the ADV to follow the SL, the first trajectory having a modified speed corresponding to the TLSL instead of the SLSL. The system generates a second trajectory for the ADV to change from the SL to the TL, the second trajectory having a speed corresponding to the TLSL. The system determines a first cost and a second cost for the first trajectory and the second trajectory respectively based on a cost function for comparison. The system controls the ADV according to a trajectory with a lowest cost.

Abstract: A source module includes a server fluid distribution unit and a busbar unit, as well as connectors. In an embodiment, a server fluid distribution unit to be coupled to a rack fluid distribution unit and the one or more server blades for deploying one or more servers. In an embodiment, a busbar unit to be coupled with an alternating current (AC) power distribution unit and the one or more server blades. In an embodiment, the source module is to be coupled to the rack fluid distribution unit to distribute cooling fluid received from a cooling fluid source to the one or more server blades of corresponding server chassis and to extract heat from the one or more servers.

Abstract: A cooling plate includes a fluid inlet port, a cooling enclosure, and an inlet channel coupled between the fluid inlet port and the cooling enclosure to channel a two-phase fluid entering from the inlet port to the cooling enclosure. The cooling enclosure includes a number of heat spreading structures coupled to an inner surface of the cooling enclosure to form spacings, where the two-phase fluid in contact with the heat spreading structures causes portions of the two-phase fluid to change into a vapor phase. The cooling plate includes an extended vapor channel coupled to the cooling enclosure to collect the two-phase fluid in vapor phase. The cooling plate includes a vapor outlet port coupled to the extended vapor channel for the two-phase fluid in vapor phase to exit the vapor outlet port, where the cooling plate is submersible in an immersion tank containing a single-phase immersion fluid.

Abstract: Sound signals that are not audible are generated by one or more speakers disposed in the vehicle. Reflected sound signals from a driver or a passenger of the vehicle that are not audible are received by one or more microphones disposed in the vehicle. A behavior-induced acoustic pattern is detected based on the reflected ultrasound signals. The behavior-induced acoustic pattern is analyzed to recognize a behavior of the driver or the passenger of the vehicle. A response or an alert is generated according to the recognized behavior of the driver or the passenger in the vehicle.

Abstract: A self-acclimating electronics package includes an electronic chip, an electrically resistive material with a negative temperature coefficient of resistivity, the electrically resistive material being thermally coupled to the electronic chip, and a thermoelectric cooler thermally coupled to the electronic chip. The thermoelectric cooler is electrically connected in series with the electrically resistive material and a power supply to cool the electronic chip, where if a temperature of the electronic chip increases, a resistance of the electrically resistive material decreases to cause the a voltage supplied to the thermoelectric cooler to increase, and if a temperature of the electronic chip decreases, a resistance of the electrically resistive material increases to cause the a voltage/current applied to the thermoelectric cooler to decrease.

Abstract: A modularized architecture for designing and implementing air side economization system is introduced. An airside economizer system operable in plurality of modes according to outside air temperature and quality. The system can be configured as a mixing module, an evaporative cooling module, and air supply module, attached to each other and having air passages there-between. The operational modes include a direct mode wherein the heat exchanger is deactivated and the plurality of blowers and plurality of louvers are activated to deliver ambient air to the load; an indirect mode wherein the heat exchanger is activated and the plurality of blowers and plurality of louvers are activated to deliver ambient air to the heat exchanger; and a mixing mode, wherein the heat exchanger is deactivated and the plurality of blowers and plurality of louvers are activated to mix ambient air with recirculated air.