Abstract: Apparatuses, systems, and techniques for caching of compiled shader programs in a cloud computing environment. An initial request for a compiled shader program for an application executed by a first client device is received. The initial request includes a first shader key generated based on first state data. If the compiled shader program is determined, based on the first shader key, to not be stored using a shader cache, the compiled shader program is received from the first client device and stored with the first shader key using the shader cache. A subsequent request for the compiled shader program is received for the application running at a second client device. The subsequent request includes a second shader key generated based on second state data. If the second shader key is determined to match the first shader key, the compiled shader program is transmitted to the second client device.

Abstract: Disclosed are apparatuses, systems, and techniques that may use machine learning for implementing speaker recognition, verification, and/or diarization. The techniques include processing audio data channels (ADCs) using a voice detection model to determine voice activity likelihoods (VALs) that individual ADCs include speech, obtaining, using VALs, a second set of ADC(s), and processing, using an audio processing a neural network (NN) model, the second set of ADCs to obtain association of the speech to the one or more speakers. The techniques also include generating a plurality of embeddings associated with the ADCs, processing the plurality of embeddings to obtain aggregated embedding(s) that represent audio data of multiple ADCs, and processing the aggregated embedding(s), using the audio processing NN model, to obtain association of the speech to the one or more speakers.

Abstract: One or more processors include circuitry to, in response to a call to an application programming interface (API), check for information about one or more memory transactions after a timeout event indicated by one or more input parameters of the API. A system includes one or more processors to, in response to a call to an application programming interface (API), check for information about one or more memory transactions after a timeout event indicated by one or more input parameters of the API. A method includes receiving an application programming interface (API) call and, in response to receiving the API call, checking for information about one or more memory transactions after a timeout event indicated by one or more input parameters of the API.

Abstract: Apparatuses, systems, and techniques to control a real-world and/or virtual device (e.g., a robot). In at least one embodiment, the device is controlled based, at least in part on, for example, one or more neural networks. Parameter values for the neural network(s) may be obtained by training the neural network(s) to control movement of a first agent with respect to at least one first target while avoiding collision with at least one stationary first holder of the at least one first target, and updating the parameter values by training the neural network(s) to control movement of a second agent with respect to at least one second target while avoiding collision with at least one non-stationary second holder of the at least one second target.

Abstract: Apparatuses, systems, and techniques to perform an application programming interface (API) to indicate one or more graph node functions of one or more graph nodes to be added to a software graph based, at least in part, on a dependency type indicated by the API. In at least one embodiment, one or more graph nodes are added to a software graph based on a node type and a dependency type.

Abstract: Systems and methods of optimizing runtime of a neural network (NN) by at least one processor may include: receiving a space of untrained NN architectures, capable of performing a predefined NN function on a given computing device, after being trained; receiving a reference accuracy metric value; evaluating a latency value for each NN architecture of the architecture space; dividing the architecture space to a plurality of groups based on said evaluated latency; performing a search among the plurality of groups, to determine a group G* that corresponds to a minimal evaluated latency, and yet comprises at least one NN architecture that maintains an accuracy metric value that is at least equal to the reference accuracy metric value; and training at least one NN architecture of the determined group G* to perform the NN function.

Abstract: In various examples, systems and methods are disclosed relating to generating animatable characters or avatars. The system can assign a plurality of first elements of a three-dimensional (3D) model of a subject to a plurality of locations on a surface of the subject in an initial pose. Further, the system can assign a plurality of second elements to the plurality of first elements, each second element of the plurality of second elements having an opacity corresponding to a distance between the second element and the surface of the subject. Further, the system can update the plurality of second elements based at least on a target pose for the subject and one or more attributes of the subject to determine a plurality of updated second elements. Further, the system can render a representation of the subject based at least on the plurality of updated second elements.

Abstract: In various examples, one or more processors can include one or more circuits to obtain an image generated by executing path tracing operations. The one or more circuits can apply an effect to the image in two-dimensional screen space and can update the image to represent a difference between the original image and the image, with the effect applied to improve downstream denoising of the image. The updated image can then be provided to a denoiser that implements one or more neural networks that are trained/updated to output a final image that is denoised and displayed using a display device.

Abstract: In various examples, systems and methods are disclosed relating to generating an output 3D latent representation by encoding, using a text encoder, a text prompt and encoding, using a 2D-3D encoder, a 2D image of an object or a 3D representation of the object. A 3D output is generated by applying the output 3D latent representation to a decoder. A reconstruction loss and a SDS loss are determined for the 3D output. At least one of the text encoder, the 2D-3D encoder, and the decoder is updated using the reconstruction loss and the SDS loss.

Abstract: In various examples, one or more processors can include one or more circuits to obtain image data for a current image being generated, in accordance with one or more ray tracing operations. The image data can correspond to a current image prior to or during reconstruction of that image. The image can be grainy, include artifacts, etc. To reduce these affects, the one or more circuits can execute one or more attention operations to determine an update to the current image based at least on a plurality of patches of the current image and windowed attention regions corresponding to each patch of the plurality of patches. In some examples, the windowed attention regions can extend at least in part beyond boundaries of each patch, allowing for information from neighboring patches to be used to update the image when reconstructing the image.

Abstract: In various examples, a VPU and associated components may be optimized to improve VPU performance and throughput. For example, the VPU may include a min/max collector, automatic store predication functionality, a SIMD data path organization that allows for inter-lane sharing, a transposed load/store with stride parameter functionality, a load with permute and zero insertion functionality, hardware, logic, and memory layout functionality to allow for two point and two by two point lookups, and per memory bank load caching capabilities. In addition, decoupled accelerators may be used to offload VPU processing tasks to increase throughput and performance, and a hardware sequencer may be included in a DMA system to reduce programming complexity of the VPU and the DMA system. The DMA and VPU may execute a VPU configuration mode that allows the VPU and DMA to operate without a processing controller for performing dynamic region based data movement operations.

Abstract: In various examples, techniques for detecting occluded objects within an environment are described. For instance, systems and methods may receive training data representing images and ground truth data indicating whether the images are associated with occluded objects or whether the images are not associated with occluded objects. The systems and methods may then train a neural network to detect occluded objects using the training data and the ground truth data. After training, the systems and methods may use the neural network to detect occluded objects within an environment. For instance, while a vehicle is navigating, the vehicle may process sensor data using the neural network. The neural network may then output data indicating whether an object is located within the environment and occluded from view of the vehicle. In some examples, the neural network may further output additional information associated with the occluded object.

Abstract: Technology related to calibrating synchronization of a data signal to a forwarded clock signal is described. A circuit includes a receiver circuit and a clock-to-data synchronization circuit. The receiver circuit receives a data signal and includes two inverters and two resistors. The clock-to-data synchronization circuit receives the data signal and the clock signal and determines an indication of synchronization between the data and clock signals. The clock-to-data synchronization circuit adjusts, using the indication, a parameter of the receiver circuit to synchronize the data signal and the clock signal.

Abstract: In various examples, a technique for performing three-dimensional (3D) scene completion includes determining an initial representation of a first 3D scene. The technique also includes executing a machine learning model to generate a first update to the initial representation at a previous time step and a second update to the initial representation at a current time step, wherein the second update is generated based at least on a threshold applied to a set of predictions corresponding to the first update. The technique also includes generating a 3D model of the 3D scene based at least on the second update to the initial representation.

Abstract: An instruction to disable a first cryptographic key of a computing device is received by a processor of the device. The first cryptographic key is embedded in the device during manufacturing to facilitate a boot sequence of the device. The instruction is signed using a cryptographic signature. The first cryptographic key is associated with a first priority indicator. The cryptographic signature is verified using a second cryptographic key embedded in the device during manufacturing. The second cryptographic key is associated with a second priority indicator. The first priority indicator is compared with the second priority indicator. Responsive to determining that the second priority indicator supersedes the first priority indicator, the first cryptographic key is disabled.

Abstract: One embodiment of a computer-implemented method for processing a neural network comprises receiving a first quantized matrix that corresponds to a portion of a multi-dimensional input tensor and has been quantized based on a first scale factor; and performing one or more computational operations using the first quantized matrix and the first scale factor to generate one or more data values that correspond to a first portion of a multi-dimensional output tensor.

Abstract: In various examples, a technique for audio-driven facial animation with adaptive speech includes determining that a rate of speech associated with an audio segment exceeds a threshold. The technique also includes based at least on the rate of speech exceeding the threshold, upsampling a first set of features associated with the audio segment into a second set of features that is different in size than the first set of features. The technique further includes generating, using one or more machine learning models and based at least on at least a subset of the second set of features, a facial animation output corresponding to the audio segment.

Abstract: In examples, the number of rays used to sample lighting conditions of a light source in a virtual environment with respect to particular locations in the virtual environment may be adapted to scene conditions. An additional ray(s) may be used for locations that tend to be associated with visual artifacts in rendered images. A determination may be made on whether to cast an additional ray(s) to a light source for a location and/or a quantity of rays to cast. To make the determination variables such as visibilities and/or hit distances of ray-traced samples of the light source may be analyzed for related locations in the virtual environment, such as those in a region around the location (e.g., within an N-by-N kernel centered at the location). Factors may include variability in visibilities and/or hit distances, differences between visibilities and/or hit distances relative to the location, and magnitudes of hit distances.

Abstract: Various examples, systems, and methods are disclosed relating to orthogonal triangle strips. A first computing system can execute one or more operations to obtain an approximation of a three-dimensional (3D) curve including a plurality of linear segments. The first computing system can execute one or more operations to determine two quadrilateral primitives (quads) for each of the plurality of linear segments using a plurality of triangle primitives, wherein each of the two quads includes two of the plurality of triangle primitives, wherein the two quads corresponding to each linear segment of the plurality of linear segments are angled relative to one another and extend along the corresponding linear segment. The first computing system can execute one or more operations to render the two quads for each of the plurality of linear segments.

Abstract: Apparatuses, systems, and techniques to perform registration among images. In at least one embodiment, one or more neural networks are trained to indicate registration of features in common among at least two images by generating a first correspondence by simulating a registration process of registering an image and applying the at least two images and the first correspondence to a neural network to derive a second correspondence of the features in common among the at least two images.