Abstract: A fluid distribution system is incorporated into a server chassis. Connector plates have fluid connectors designed to connect with fluid cooling receptacles on server boards. A linkage system can be manually manipulated to translate the connector plates so as to engage the electronics boards and connect the fluid connectors to the receptacle. A pair of main connectors connect to the rack's fluid manifold to circulate cooling fluid to cool devices mounted onto the server boards. A locking mechanism can be included to mechanically lock the chassis to the rack to prevent accidental dismount of the chassis from the rack. Also, an anchor mechanism may be provided to apply counter force when engaging the connector plates with the server boards so as to prevent accidental dismount of the server board by the moving connector plate.

Abstract: A perception and planning system includes a system frame and one or more cooling fans mounted on a first side of the system frame to generate an airflow flowing outwardly to remove at least a portion of the heat generated by the processors. The system further includes an air filtering module mounted on a second side of the system frame to filter the air flow flowing inwardly to remove durst or contaminated material from the airflow. The air filtering module includes an air filter, a mounting plate, and a cover plate. The mounting plate can be mounted on the second side of the system frame to receive the air filter. The mounting plate includes one or more magnets. The cover plate can be attached to the mounting plate by a magnetic force generated from the magnets to keep the air filter between the mounting plate and the cover plate.

Abstract: Embodiments are disclosed of a hybrid heat sink. The hybrid heat sink includes a dry base and a wet base spaced apart from the dry base. The wet base including a fluid inlet and a fluid outlet. A plurality of air channels is formed between the dry base and the wet base; and a plurality of liquid channels are formed between the dry base and the wet base. The plurality of liquid channels are coupled to the fluid inlet and the fluid outlet, and the plurality of liquid channels are thermally coupled to the plurality of air channels. Other embodiments are disclosed and claimed.

Abstract: Existing research on cross-lingual retrieval cannot take good advantage of large-scale pretrained language models, such as multilingual BERT and XLM. The absence of cross-lingual passage-level relevance data for finetuning and the lack of query-document style pretraining are some of the key factors of this issue. Accordingly, embodiments of two novel retrieval-oriented pretraining tasks are presented herein to further pretrain cross-lingual language models for downstream retrieval tasks, such as cross-lingual ad-hoc retrieval (CUR) and cross-lingual question answering (CLQA). In one or more embodiments, distant supervision data was constructed from multilingual texts using section alignment to support retrieval-oriented language model pretraining. In one or more embodiments, directly finetuning language models on part of an evaluation collection was performed by making Transformers capable of accepting longer sequences.

Abstract: A thermal management system includes a cooling unit, a condenser, and a processor. The processor is located within a server, and the system also includes a phase change cooling device in thermal communication with the processor, and in fluid communication with the condenser. The system also includes a single phase cooling device in thermal communication with the phase change cooling device, and in fluid communication with the liquid cooling unit. The system also includes a temperature sensor in thermal communication with the single phase cooling device, and a fluid pump to move fluid between the liquid cooling unit and the single phase cooling device. A TEC device may also be implemented between the phase change cooling device and the single phase cooling device.

Abstract: A printed circuit board includes first and second surfaces, first and second layers, and first and second vias. The first via extends from a first layer to the second surface and includes a first portion that is on a conductive path between the first layer and the second layer and a second portion that is not on the conductive path. A length of the first portion of the first via is greater than that of the second portion of the first via. The second via extends from the second surface to the second layer. The second via includes a first portion that is on the conductive path between the first layer and the second layer and a second portion that is not on the conductive path. A length of the first portion of the second via is greater than that of the second portion of the second via.

Abstract: Immersion cooling system for electronics. The electronics are immersed in immersion tank filled with immersion liquid. A universal supply utilizes a supply panel to fill cooled liquid into the tank. Localized supply utilizes a manifold to deliver cooled liquid to specific devices within the tank. Cooled liquid is delivered via connection of the manifold and the flexible hoses to cooling plates mounted onto the devices. Each of the cooling plates has a cavity for passing cooled liquid, a supply inlet connected to one of the flexible hoses, and an outlet hole releasing the liquid into the tank. The cooled liquid is supplied to the universal supply from a main cooler. The cooled liquid is supplied to the manifold from the main cooler or from a secondary cooler. The immersion tank has a single recirculation outlet to return liquid mixed from both the universal supply and the localized supply.

Abstract: A computer-implement method for operating an ADV is disclosed. A first trajectory for the ADV to drive along is planned. The ADV is to autonomously drive along the first trajectory. An obstacle in an affected region of the ADV is detected based on sensor data obtained from a plurality of sensors mounted on the ADV. An expected residence time of the obstacle in the affected region is determined. Whether to plan a second trajectory or to wait for the obstacle to leave the affected region is determined based on the expected residence time of the obstacle in the affected region. A second trajectory for the ADV to drive along is planned and the ADV is to autonomously drive along the second trajectory, or the ADV is to wait for the obstacle to leave the affected region and to autonomously drive along the first trajectory afterwards.

Abstract: A method, apparatus, and system for determining average lane travel speeds is disclosed. A plurality of vehicles traveling in a same direction as the ADV in a plurality of lanes are identified. Over a first time period, the plurality of vehicles is tracked. At least a first quantity of representative vehicles within the plurality of vehicles that are representative of vehicles traveling in the lane over the first time period are identified. For each of the plurality of lanes, an average speed over the first time period of the representative vehicles associated with the lane is determined. A trajectory is planned for the ADV, wherein the planned trajectory moves toward a lane whose representative vehicles have a fastest average speed. Thereafter, control signals are generated to control operations of the ADV based on the planned trajectory.

Abstract: In one embodiment, an immersion cooling system includes a container to contain first coolant received from a first cooling source and server chassis at least partially submerged into the first coolant. Each server chassis includes an electronic device and a cooling plate attached thereon to extract at least a portion of heat generated by the electronic device. The cooling plate includes an inlet port to receive second coolant from a second cooling source, a coolant channel to distribute the second coolant, and an outlet port to return the second coolant back to the second cooling source. The cooling system further includes a return manifold to be coupled to the second cooling source, the return manifold having one or more manifold return connectors respectively coupled with the server chassis and to receive and return the second coolant from the server chassis back to the second cooling source.

Abstract: Methods, systems, and devices for designing and implementing power and cooling fluid in a computing environment such as an electronics rack are disclosed. The disclosed methods and systems may provide for a high degree of power distribution and cooling fluid distribution reliability. To provide for a high degree of reliability, the system may include a number of protective features that may reduce the likelihood of connectors used for power and cooling fluid distribution from being damaged. The system may also provide for segregation of power distribution components from cooling fluid distribution components. The rack configurations include codesign of the server and rack to form the physical segregation. The segregation may reduce the chance of these components impacting the operation of other components.

Abstract: Developed and presented herein are embodiments of a new end-to-end approach for audio denoising, from a synthesis perspective. Instead of explicitly modelling the noise component in the input signal, embodiments directly synthesize the denoised audio from a generative model (or vocoder), as in text-to-speech systems. In one or more embodiments, to generate the phonetic contents for the autoregressive generative model, it is learned via a variational autoencoder with discrete latent representations. Furthermore, in one or more embodiments, a new matching loss is presented for the denoising purpose, which is masked on when the corresponding latent codes differ. As compared against other method on test datasets, embodiments achieve competitive performance and can be trained from scratch.

Abstract: A coolant management unit includes a server supply manifold, a server return manifold, a power distribution, and a controller. A server supply manifold is to receive cooling fluid from a cooling fluid source. The server supply manifold is to distribute the cooling fluid to server blades. The server return manifold is to receive vapor from the one or more server blades. The cooling fluid is two-phase cooling fluid to extract heat from one or more servers and to evaporate into the vapor into the server return manifold, and the vapor is transmitted to an external condenser via the rack return manifold to be condensed back to a liquid form. A power distribution bus is configured to distribute power to the one or more servers. A controller is configured to control a fluid pump coupled to the server supply manifold based on one or more signals received from different sensors.

Abstract: An obstacle state evolution of a spatial position of a moving obstacle over a period of time is determined. A lane-obstacle relation evolution of the moving obstacle with each of one or more lanes near the moving obstacle over the period of time is further determined. An intended movement of the moving obstacle is predicted based on the obstacle state evolution and the lane-obstacle evolution. Thereafter, a trajectory of the ADV is planned to control the ADV to avoid a collision with the moving obstacle based on the predicted intended movement of the moving obstacle. The above process is iteratively performed for each of the moving obstacles detected within a predetermined proximity of the ADV.

Abstract: A data format converter rearranges data of an input image for input to a systolic array of multiply and accumulate processing elements. The image has a pixel height and a pixel width in a number of channels equal to a number of colors per pixel. The data format converter rearranges the data to a second, greater number of channels and inputs the second number of channels to one side of the systolic array. The second number of channels is less than or equal to the number of MAC PEs on the one side of the systolic array, and results in greater MAC PE utilization in the systolic array.

Abstract: A server chassis for a server rack can include a connector assembly. The connector assembly can include a first pair of connectors to fluidly connect to one or more cold plates that are thermally coupled to one or more electronic server components, and a second pair of connectors, to fluidly connect to a supply line and return line of the server rack. The connector assembly may further include a mobile plate upon which the second pair of connectors are attached to. The mobile plate may be attached as part of the connector assembly and movable to and from a direction that connects or disconnects the second pair of connectors to the supply line and the return line of the server rack.

Abstract: In one embodiment, method for determining capability boundary of a safety redundancy of an autonomous driving vehicle (ADV) includes obtaining a sensor layout associated with the ADV representing a system having a plurality of sensors mounted on a plurality of locations of the ADV. A zone failure risk of one or more sensors within the predetermined zones is estimated based on statistical operational data of the one or more sensors for each of the plurality of predetermined zones. An overall failure risk of the sensors is determined based on the zone failure risks of the predetermined zones based on relative locations of the sensors across the predetermined zones. A dynamic risk adjustment is determined based on the overall failure risk of the sensors, the dynamic risk adjustment representing a reliability of a sensor system associated with the ADV for estimating a safety of autonomous driving of the ADV.

Abstract: Disclosed are designs to facilitate blind-mating of fluid connectors on servers and on racks that house the servers. The design includes two plates. A first plate has a connector channel to guide the connectors of the server through the first plate and spring structures to apply a force to push the connectors of the server toward the center of the connector channel. A second plate has positioning holes around its perimeter for mounting the connectors of the server using the force of the spring structures. The first plate is rotatable around a rotation axis to align the connectors of the server with the connectors of the rack. The second plate may have an elastic layer that is compressed to change a width of the second plate by inserting the second plate to the rack to automatically align the connectors of the server and of the rack in the horizontal direction.

Abstract: Deep neural network (DNN) models have been widely used for user-relevance content prediction. Presented herein is a new user-relevance framework, embodiments of which may be referred as Gating-Enhanced Multi-task Neural Networks (GemNN). In one or more, 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.

Abstract: The disclosure provides a system, for designing and developing immersion cooling in data centers. The system includes multiple aisles that perform different functions. One of the aisles includes space for housing immersion tanks in which components of the data center may be positioned to be cooled. A second aisle allows the immersion tanks to be placed in the other aisle and for backside access. A third aisle transfers vapor from the aisle housing the immersion tanks to a vapor return unit to condense the vapor. A distribution unit distributes the coolant to cool the components of the data center is designed at the second aisle. The aisles and other units can be modularized to allow for deployment to meet different types of data center requirements.