Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, are described for scheduling tasks of ML workloads. A system receives requests to perform the workloads and determines, based on the requests, resource requirements to perform the workloads. The system includes multiple hosts and each host includes multiple accelerators. The system determines a quantity of hosts assigned to execute tasks of the workload based on the resource requirement and the accelerators for each host. For each host in the quantity of hosts, the system generates a task specification based on a memory access topology of the host. The specification specifies the task to be executed at the host using resources of the host that include the multiple accelerators. The system provides the task specifications to the hosts and performs the workloads when each host executes assigned tasks specified in the task specifications for the host.

Abstract: A computer-generated virtual object manipulator having one or more affordances for manipulating a computer-generated virtual object is disclosed. Selection of a virtual object can cause an object manipulator to be displayed over the virtual object. The object manipulator can include a cone-shaped single-axis translation affordance for each of one or more object axes, a disc-shaped single-axis scale affordance for each of the one or more object axes, an arc-shaped rotation affordance for rotation about each of the one or more object axes, and a center of object affordance for free space movement of the virtual object. The object manipulator can also include a slice-shaped two-axis translation affordance that can be displayed after hovering over an area in a particular plane.

Abstract: In one implementation, a method of playing audio is performed at a device including one or more processors coupled to non-transitory memory. The method includes playing audio associated with an object at a frequency-dependent volume based on a distance between a user location and an object location, an object orientation with respect to the user location, and the frequency-dependent three-dimensional audio emission pattern. In various implementations, methods of visualizing the frequency-dependent three-dimensional audio emission pattern and changing properties thereof are disclosed.

Abstract: In one implementation, a method of playing audio is performed at a device including one or more processors coupled to non-transitory memory. The method includes playing audio associated with an object at a frequency-dependent volume based on a distance between a user location and an object location, an object orientation with respect to the user location, and the frequency-dependent three-dimensional audio emission pattern. In various implementations, methods of visualizing the frequency-dependent three-dimensional audio emission pattern and changing properties thereof are disclosed.

Abstract: Methods, systems, and apparatus, including computer-readable storage media for uncorrectable memory recovery. Different sources of uncorrectable memory error are handled to provide for recovery actions by a host kernel of a machine hosting one or more virtual machines. Rather than defaulting to kernel panic behavior, the host kernel can identify the source of uncorrectable error, and cause the host machine and/or the affected virtual machines to take recovery action that is less disruptive than abrupt shutdown from panic. For example, the host kernel can handle uncorrectable memory error caused by kernel accesses to guest memory of a host virtual machine, as well as uncorrectable memory error improperly raised as a result of known defects in host machine hardware. The host kernel can also be configured to detect sources of overflow in exceptions raised by a processor as a result of uncorrectable memory error.

Abstract: Embodiments of the present disclosure disclose an anomaly detector for detecting an anomaly in a sequence of poses of a human performing an activity. The anomaly detector includes an input interface configured to accept input data indicative of a distribution of the sequence of poses, a memory configured to store a discriminative one-class classifier having a pair of complementary classifiers bounding normal distribution of pose sequences in a reproducing kernel Hilbert space (RKHS), a processor configured to embed the input data into an element of the RKHS and classify the embedded data using the discriminative one-class classifier, and an output interface configured to render a classification result.

Abstract: In one embodiment, a method includes accessing first sets of tokens associated with a desired task and one or more modalities associated with a context of the desired task, determining a second set of tokens for each of the one or more modalities using a classifier network associated with the modality, generating a number of embedding vectors by mapping the first sets of tokens and the second set of tokens associated with each of the one or more modalities to an embedding space, and producing a sequence of words addressing the desired task by processing the number of embedding vectors with an encoder-decoder network.

Abstract: An optical element including an optically transparent lens which defines a curved surface having a steepness given by an R/# of from about 0.5 to about 1.0. A film is positioned on the curved surface. The film includes an index layer. A composite layer is positioned on the curved surface having a refractive index greater than the index layer. The composite layer includes HfO2 and Al2O3. The composite layer has a mole fraction X of HfO2, wherein X is from about 0.05 to about 0.95 and a mole fraction of Al2O3 in the composite layer is 1?X.

Abstract: Methods, systems, and apparatus, including computer-readable storage media for uncorrectable memory recovery. Different sources of uncorrectable memory error are handled to provide for recovery actions by a host kernel of a machine hosting one or more virtual machines. Rather than defaulting to kernel panic behavior, the host kernel can identify the source of uncorrectable error, and cause the host machine and/or the affected virtual machines to take recovery action that is less disruptive than abrupt shutdown from panic. For example, the host kernel can handle uncorrectable memory error caused by kernel accesses to guest memory of a host virtual machine, as well as uncorrectable memory error improperly raised as a result of known defects in host machine hardware. The host kernel can also be configured to detect sources of overflow in exceptions raised by a processor as a result of uncorrectable memory error.

Abstract: The present invention provides improved methods of protein purification using CEX chromatography. Such methods generally comprise the steps of: contacting a protein of interest (e.g., an antibody) with a cation exchange resin at a first pH, that is less than the pI of the most acidic isoform of the protein of interest, such that the protein of interest binds to the resin; washing the cation exchange resin at a second pH that is greater than the first pH, but less than the pI of the most acidic isoform of the protein of interest; and eluting the protein of interest from the resin at a third pH that is about equal to or less than the first pH. The methods of the invention are particularly useful for the commercial purification of recombinant therapeutic proteins (e.g., antibodies).

Abstract: The disclosure relates to a method and a system for predicting the operation time of sparse matrix vector multiplication. The method comprises constructing a convolutional neural network comprising an input layer, a feature processing layer, a data splicing layer and an output layer for outputting prediction results. The method further comprises acquiring a plurality of groups of sparse matrices with known sparse matrix vector multiplication operation time as sample data, inputting the sample data into the convolutional neural network to train the convolutional neural network, and inputting the sparse matrix to be classified into the trained convolutional neural network to realize the prediction of the operation time of sparse matrix vector multiplication.

Abstract: Various implementations relate generally to multi-sensor devices. Some implementations more particularly relate to a multi-sensor device including a ring of radially-oriented photosensors. Some implementations more particularly relate to a multi-sensor device that is orientation-independent with respect to a central axis of the ring. Some implementations of the multi-sensor devices described herein further include one or more additional sensors. For example, some implementations include an axially-directed photosensor. Some implementations also can include one or more temperature sensors configured to sense an exterior temperature, for example, an ambient temperature of an outdoors environment around the multi-sensor. Additionally or alternatively, some implementations include one or more of an infrared sensor or infrared sensors, a cellular communication circuit, and a GPS module.

Abstract: One or more frames sampled from a media item of an event that is concurrently being streamed to one or more users of a content delivery platform are received. The one or more frames are analyzed to identify one or more candidate event clocks within the one or more frames. Whether a candidate event clock of the one or more candidate event clocks satisfies one or more conditions for each of the one or more frames is determined. Responsive to determining that the candidate event clock of the one or more candidate event clocks satisfies the one or more conditions, the candidate event clock is identified as an actual event clock used to time the event and mapping data that maps the actual event clock to a timestamp associated with a respective frame of the one or more frames of the media item is generated.

Abstract: The disclosed technology provides techniques, systems, and apparatus for containing and recovering from uncorrectable memory errors in distributed computing environment. An aspect of the disclosed technology includes a hypervisor or virtual machine manager that receives signaling of an uncorrectable memory error detected by a host machine. The virtual machine manager then uses information received via the signaling to identify virtual memory addresses or memory pages associated with the corrupted memory element so as to allow for containment and recovery from the error.

Abstract: The disclosed technology provides techniques, systems, and apparatus for containing and recovering from uncorrectable memory errors in distributed computing environment through migration of virtual machines and associated memory to a target host machine. An aspect of the disclosed technology includes a hypervisor or virtual machine manager that receives signaling of an uncorrectable memory error detected by a host machine. The virtual machine manager then uses information received via the signaling to identify virtual memory addresses or memory pages associated with the corrupted memory element so as to allow for containment and recovery from the error, and for live migration of the virtual machine.

Abstract: An optical element including an optically transparent lens which defines a curved surface having a steepness given by an R/# of from about 0.5 to about 1.0. A film is positioned on the curved surface. The film includes an index layer. A composite layer is positioned on the curved surface having a refractive index greater than the index layer. The composite layer includes HfO2 and Al2O3. The composite layer has a mole fraction X of HfO2, wherein X is from about 0.05 to about 0.95 and a mole fraction of Al2O3 in the composite layer is 1?X.

Abstract: Embodiments of the present disclosure disclose an anomaly detector for detecting an anomaly in a sequence of poses of a human performing an activity. The anomaly detector includes an input interface configured to accept input data indicative of a distribution of the sequence of poses, a memory configured to store a discriminative one-class classifier having a pair of complementary classifiers bounding normal distribution of pose sequences in a reproducing kernel Hilbert space (RKHS), a processor configured to embed the input data into an element of the RKHS and classify the embedded data using the discriminative one-class classifier, and an output interface configured to render a classification result.