09 - Scientific and electric apparatus and instruments
42 - Scientific, technological and industrial services, research and design
Goods & Services
Downloadable software libraries being computer software
development tools; downloadable software and software
libraries for data compression and decompression;
downloadable software and software libraries for data
compression and decompression using graphics processing
units (GPUs); downloadable software development tools for
data compression and decompression using graphics processing
units (GPUs). Providing temporary use of online non-downloadable software
libraries being computer software development tools;
providing temporary use of online non-downloadable software
and software libraries for data compression and
decompression; providing temporary use of online
non-downloadable software and software libraries for data
compression and decompression using graphics processing
units (GPUs); providing temporary use of online
non-downloadable software development tools for data
compression and decompression using graphics processing
units (GPUs); design and development of computer software.
2.
DATA ASSOCIATION USING CORRELATION RESPONSE VALUES
In various examples, image areas may be extracted from a batch of one or more images and may be scaled, in batch, to one or more template sizes. Where the image areas include search regions used for localization of objects, the scaled search regions may be loaded into Graphics Processing Unit (GPU) memory and processed in parallel for localization. Similarly, where image areas are used for filter updates, the scaled image areas may be loaded into GPU memory and processed in parallel for filter updates. The image areas may be batched from any number of images and/or from any number of single-and/or multi-object trackers. Further aspects of the disclosure provide approaches for associating locations using correlation response values, for learning correlation filters in object tracking based at least on focused windowing, and for learning correlation filters in object tracking based at least on occlusion maps.
G06T 1/20 - Processor architecturesProcessor configuration, e.g. pipelining
G06T 11/20 - Drawing from basic elements, e.g. lines or circles
G06V 10/764 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
G06V 10/82 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
G06V 20/58 - Recognition of moving objects or obstacles, e.g. vehicles or pedestriansRecognition of traffic objects, e.g. traffic signs, traffic lights or roads
3.
ENCODING IMAGE REGIONS FOR MACHINE LEARNING AND AI APPLICATIONS
In various examples, properties may be determined for image regions, where the image regions are indicated by output data generated using MLMs. An encoder may use the properties to generate encoded images using encoding quality settings for the image regions. When an encoded image is decoded and applied to the MLMs, corresponding output data may indicate an image region which is likely to correspond to an encoded image region of the encoded image, and which may be applied to at least one MLM. Thus, the properties for encoding an image region to an encoded image can be adapted to control the visual quality of an image region determined from a decoded version of the encoded image. The properties may be determined based at least on performance metric values for the MLMs or based at least on a ranking of the image regions.
H04N 19/42 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
H04N 19/154 - Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
H04N 19/167 - Position within a video image, e.g. region of interest [ROI]
H04N 19/176 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
4.
COMPENSATING COUPLING TRANSFORMERS FOR TRANS-INDUCTOR VOLTAGE REGULATORS
Various embodiments disclose a trans-inductor voltage regulator comprising a first group pairing of switching circuits that includes, a first group of switching circuits, a first compensating coupling inductor coupled to the first group of switching circuits, a second group of switching circuits, and a second compensating coupling inductor coupled to the second group of switching circuits, where the first compensating coupling inductor comprises a first winding of a compensating coupling transformer, and the second compensating coupling inductor is a second winding of the compensating coupling transformer.
H02M 3/335 - Conversion of DC power input into DC power output with intermediate conversion into AC by static converters using discharge tubes with control electrode or semiconductor devices with control electrode to produce the intermediate AC using devices of a triode or a transistor type requiring continuous application of a control signal using semiconductor devices only
5.
ENHANCED OBJECT IDENTIFICATION USING ONE OR MORE NEURAL NETWORKS
Apparatuses, systems, and techniques to identify one or more objects in one or more images. In at least one embodiment, one or more objects are identified in one or more images based, at least in part, on a likelihood that one or more objects is different from other objects in one or more images.
In various examples, an obstacle detector is capable of tracking a velocity state of detected objects or obstacles using LiDAR data. For example, using LiDAR data alone, an iterative closest point (ICP) algorithm may be used to determine a current state of detected objects for a current frame and a Kalman filter may be used to maintain a tracked state of the one or more objects detected over time. The obstacle detector may be configured to estimate velocity for one or more detected objects, compare the estimated velocity to one or more previous tracked states for previously detected objects, determine that the detected objects corresponds to a certain previously detected object, and update the tracked state for the previously detected object with the estimated velocity.
B60T 7/12 - Brake-action initiating means for automatic initiationBrake-action initiating means for initiation not subject to will of driver or passenger
B60W 30/09 - Taking automatic action to avoid collision, e.g. braking and steering
Apparatuses, systems, and techniques are presented to generate one or more images. In at least one embodiment, two or more pixels from two or more images are blended based, at least in part, on a distance of the two or more pixels from a region of the two or more images, in which pixel colors are substantially similar.
A system includes a gain module, an optical module, and a feedback module. The gain module processes a first portion of an electrical signal to generate a first compensated portion of the electrical signal, the first portion of the electrical signal having a first frequency range. The optical module generates an optical signal based on a combination of the first compensated portion of the electrical signal and a second portion of the electrical signal having a second frequency range that is higher than the first frequency range. The feedback module provides, to the gain module, an electrical feedback signal based at least in part on a signal strength of the optical signal. The gain module updates the first compensated portion of the electrical signal based on the electrical feedback signal.
In various examples, the disclosed techniques include receiving, from an application executing in a virtual machine (VM), a request to perform a cryptographic operation, wherein the request specifies an ephemeral key identifier and source data. The techniques also determine, using key metadata received from a trusted execution environment, a key slot identifier associated with the ephemeral key identifier, wherein the key slot identifier identifies a key slot in which a cryptographic key is stored. The techniques further cause the cryptographic operation to be performed on the source data in the trusted execution environment using the cryptographic key, where the cryptographic key used to perform the cryptographic operation is accessed from the key slot identified by the key slot identifier. The techniques further provide, to the application, a cryptographic operation result received form the trusted execution environment.
Apparatuses, systems, and techniques to perform one or more operations using a tensor. In at least one embodiment, one or more circuits are to perform an application programming interface (API) to cause two or more tensor contractions to be performed based, at least in part, on one or more input parameters of the API.
Apparatuses, systems, and techniques to optimize processor performance. In at least one embodiment, a method optimizes linked code based, at least in part, on storing an indication of whether two portions of code have been linked.
In various embodiments, systems and methods for distributed simulation of time dynamics of quantum systems with an optimized tensor mode redistribution schedule are provided. A scheduling optimizer may create a schedule of tensor redistributions based on a hypergraph constructed from a quantum many-body operator. The scheduling optimizer iteratively distributes a quantum state tensor into state tensor slices based on a clustering process that partitions the hypergraph into clusters, and redistributes the quantum state tensor across processing units based on identified sliced or non-sliced tensor modes according to the obtained clusters. Operators that can be applied to non-sliced modes of the redistributed quantum state tensor are applied, and the iterative process repeats until each of the operators of the full quantum many-body operator have been applied. Because these operators act on non-sliced modes, inter-processing unit communication is minimized while executing these operations, and tensor redistributions across all processing units is minimized.
Embodiments of the present disclosure relate to an AI agentic system for scene understanding. Some embodiments perform such scene understanding by extracting, indexing, and iteratively refining scene data through an AI agent that autonomously generates and refines queries in a continuous loop until a predefined completeness threshold is met. This ensures that scene data is not only captured but also refined over time, producing a fully indexed and queryable representation of the scene.
G06F 16/787 - Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using geographical or spatial information, e.g. location
14.
SIMULATING DIFFERENTIABLE OBJECT ELASTICITY USING IMPLICIT FUNCTIONS
Approaches presented herein provide for the use of implicit functions to simulate differentiable object elasticity. An implicit continuous function, such as a signed distance function (SDF), can be used to approximate the surface of an object by providing scalar values from a set of vertices of a regular grid in which the object representation is to be generated. Interpolation can be applied to determine an approximate surface location and shape within each boundary cell. A trained neural network, such as a multilayer perceptron (MLP), can be used to determine appropriate quadrature points that fall within the volume of the object. A finite element analysis can integrate over these quadrature points, using both continuous and discrete settings, as a basis for performing efficient differentiable elasticity simulations including the deformable object.
In various examples, techniques for determining perception zones for object detection are described. For instance, a system may use a dynamic model associated with an ego-machine, a dynamic model associated with an object, and one or more possible interactions between the ego-machine and the object to determine a perception zone. The system may then perform one or more processes using the perception zone. For instance, if the system is validating a perception system of the ego-machine, the system may determine whether a detection error associated with the object is a safety-critical error based on whether the object is located within the perception zone. Additionally, if the system is executing within the ego-machine, the system may determine whether the object is a safety-critical object based on whether the object is located within the perception zone.
G05D 1/00 - Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
G06V 20/58 - Recognition of moving objects or obstacles, e.g. vehicles or pedestriansRecognition of traffic objects, e.g. traffic signs, traffic lights or roads
16.
CONVERTING NON-UNIQUE WIRELESS DEVICE IDENTIFIERS TO UNIQUE WIRELESS DEVICE IDENTIFIERS
Apparatuses, systems, and techniques to cause one or more non-unique wireless device identifiers to correspond to one or more unique wireless device identifiers. In at least one embodiment, one or more non-unique wireless device identifiers are mapped to one or more unique wireless device identifiers.
Systems and methods are disclosed that monitor for supply noise from a power source using a voltage noise monitor (VNM). For instance, the VNM may include voltage sense circuitry comprising a controllable resistor that is controlled using threshold information. The resistance of the controllable resistor may be changed based on closing and/or opening one or more switches associated with step resistors using the bits from the threshold information. Furthermore, the VNM may include digital circuitry that comprises a hold finite state machine and a sticky hold counter. Using the digital circuitry, the VNM may be configured to hold a noise detection event for a plurality of clock cycles. In addition, the VNM may perform a calibration process based on setting two voltages for the power source to obtain two codes, and determining a transfer function based on the two voltages and the two codes.
Apparatuses, systems, and techniques to translate instructions that use operand stack structures to instructions that use registers and/or memory for operands. In at least one embodiment, one or more instruction operand stack structures are duplicated if one or more branch instructions use one or more instruction operands stored in the one or more stack structures.
Various embodiments include techniques for encoding media frames using a multistage search without the processing overhead of pyramidal motion estimation techniques. The first stage of the multistage search generates many motion vectors, where each motion vector is based on a small range full pixel search of a pixel group in the media frame. The video encoder selects a full pixel best motion (FBM) vector for all pixel groups from the motion vectors generated during the first stage. The second and subsequent phases of the multistage search is based on the FBM vector from the prior stage as a starting point, where the search range for each subsequent phase is larger than the search range of the prior stage. The multistage search can be performed over a fixed number of stages. Alternatively, the multistage search can be terminated when the cost value for the current stage is below a threshold value.
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a matrix multiplication is performed based on an indication of whether information to be used by part of the matrix multiplication has been loaded.
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a processor performs an API to indicate whether a partial result of a matrix multiplication has been loaded.
In various examples, techniques for optimizing object detection models are described herein. Systems and methods are disclosed that process sensor data using a backbone of a machine learning model(s) in order to generate feature maps at different resolutions. The systems and methods then use the machine learning model(s) to generate a vector based at least in part on one or more of the feature maps. For example, if the backbone generates four feature maps, then the machine learning model(s) may generate the vector using two feature maps from the four feature maps. The systems and methods then process the vector using a transformer of the machine learning model(s) in order to generate data representing a class label(s) for an object(s) depicted by an image represented by the sensor data and/or a location(s) of the object(s) within the image.
G06V 10/82 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
G06V 10/77 - Processing image or video features in feature spacesArrangements for image or video recognition or understanding using pattern recognition or machine learning using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]Blind source separation
Apparatuses, systems, and techniques to generate an image. In at least one embodiment, one or more neural networks are to generate a second image based, at least in part, on a first image and information indicating zero or more differences between the first and second image.
Apparatuses, systems, and techniques to modify a set of training data used for machine learning. In at least one embodiment, a set of images used for training a machine learning system is resampled by augmenting the set of images with additional images of under represented object types extracted from portions of existing training images in the set.
G06F 18/2411 - Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
G06F 18/214 - Generating training patternsBootstrap methods, e.g. bagging or boosting
Various embodiments include a printed circuit board cover assembly comprising a cover layer that attaches to at least a portion of a printed circuit board layer, and a retention hook that extends from the cover layer and physically interfaces with at least a portion of a connection port. In various embodiments, a printed circuit board package comprises a printed circuit board layer, a cover layer that attaches to at least a portion of the printed circuit board layer, and a retention hook extending from the cover layer, the retention hook configured to physically interface with at least a portion of a connection port.
Technologies for generating a set of models for each account, where each model is a fine-grained, unsupervised behavior model trained for each user to monitor and detect anomalous patterns are described. An unsupervised training pipeline can generate user models, each being associated with one of multiple accounts and is trained to detect an anomalous pattern using feature data associated with the one account. Each account is associated with at least one of a user, a machine, or a service. An inference pipeline can detect a first anomalous pattern in first data associated with a first account using a first user model. The inference pipeline can detect a second anomalous pattern in second data associated with a second account using a second user model.
Disclosed are apparatuses, systems, and techniques that implement training and deployment of streaming multimodal language systems capable of generating live text outputs. The techniques include predicting, over a plurality of iterations, a plurality of text tokens of a streaming text output associated with a streaming audio input. An individual iteration updates audio embeddings representative of the streaming audio input, processes, using a cross-modality network, the audio embeddings and text embeddings representative of a text input associated with the streaming audio input to obtain a plurality of cross-attention states, provides, to a language model (LM), a prompt including output embeddings obtained based at least on the plurality of cross-attention states, and receives, from the LM, a text token predicted for the respective iteration. The streaming text output is generate using the predicted text tokens.
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a processor performs an API to indicate a portion of results of a matrix multiply operation that is to be partially loaded.
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a processor performs an API to indicate operands of a matrix multiply operation that are to be partially loaded.
Apparatuses, systems, and techniques to perform one or more operations using a tensor. In at least one embodiment, one or more circuits are to perform an application programming interface (API) to cause parameters indicating operands and dimensions of operands of two or more tensor contractions to be stored.
A method includes providing heat produced by a fuel cell from converting a source of gas to generate electrical power for a datacenter to an absorption chiller to produced cooled liquid. The method further includes utilizing the produced cooled liquid to cool one or more electronic components of the datacenter.
H05K 7/20 - Modifications to facilitate cooling, ventilating, or heating
G06F 1/26 - Power supply means, e.g. regulation thereof
H01M 8/04007 - Auxiliary arrangements, e.g. for control of pressure or for circulation of fluids related to heat exchange
H01M 8/04089 - Arrangements for control of reactant parameters, e.g. pressure or concentration of gaseous reactants
H01M 8/04119 - Arrangements for control of reactant parameters, e.g. pressure or concentration of gaseous reactants with simultaneous supply or evacuation of electrolyteHumidifying or dehumidifying
H05K 7/14 - Mounting supporting structure in casing or on frame or rack
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a processor performs an API to indicate whether operands to be used by a matrix multiplication have been partially loaded.
Apparatuses, systems, and techniques to perform a digital simulation of a wireless network. In at least one embodiment, a processor generates a virtual network simulation and combine user equipment of said virtual network simulation into virtual cells in order to cause at least one of two or more mobile device emulation programs to be selected based, at least in part, on two or more distinct physical resource block parameters corresponding to said two or more mobile devices.
Apparatuses, systems, and techniques are presented for generating instructional text. In at least one embodiment, an instructional video is analyzed to determine logical steps of a process or task demonstrated in that video, and instructive text is generated for those logical steps.
G06V 30/262 - Techniques for post-processing, e.g. correcting the recognition result using context analysis, e.g. lexical, syntactic or semantic context
G09B 9/00 - Simulators for teaching or training purposes
Apparatuses, systems, and techniques to help identify one or more directions to transmit a first fifth generation new radio (“5G NR”) signal. In at least one embodiment, said one or more identified directions to be used to transmit a first 5G NR signal is based, at least in part, on channel state information of one or more second 5G NR signals.
H04B 7/06 - Diversity systemsMulti-antenna systems, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station
36.
LONG-RANGE 3D OBJECT DETECTION USING 2D BOUNDING BOXES
3D object detection is a computer vision task that generally detects (e.g. classifies and localizes) objects in 3D space from the 2D images or videos that capture the objects. Current techniques used for 3D object detection rely on machine learning processes that learn to detect 3D objects from existing images annotated with high-quality 3D information including depth information generally obtained using lidar technology. However, due to lidar's limited measurable range, current machine learning solutions to 3D object detection do not support detection of 3D objects beyond the lidar range, which is needed for numerous applications, including autonomous driving applications where existing close or midrange 3D object detection does not always meet the safety-critical requirement of autonomous driving. The present disclosure provides for 3D object detection using a technique that supports long-range detection (i.e. detection beyond the lidar range).
G06T 7/70 - Determining position or orientation of objects or cameras
G06T 7/80 - Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
G06V 10/22 - Image preprocessing by selection of a specific region containing or referencing a patternLocating or processing of specific regions to guide the detection or recognition
G06V 10/82 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Apparatuses, systems, and methods to store non-graphics kernels with graphics kernels. In at least one embodiment, performance of an application programming interface (“API”) causes non-graphics kernels to be stored sequentially with graphics kernels to be performed by one or more processors.
In various examples, techniques are described for adapting a multilingual Large Language Model (LLM) into a bilingual Small Language Model (SLM) that exhibits model capacity to understand, process, and generate content in both English and a Low-Resource Language (LRL). The techniques include compressing the LLM to generate a multilingual SLM and performing continued pre-training on the multilingual SLM to generate the bilingual SLM. The techniques also include performing one or more alignment techniques on the bilingual SLM to adapt the SLM's outputs to human values and expectations regarding, e.g., profanity, privacy, politeness, bias, and/or conversational style. The techniques may generate various training corpora, each including one or more of natural English content, natural LRL content, synthetic LRL content generated via translation from English sources, and transliterated synthetic LRL content based on transliterations of natural and/or synthetic LRL content.
Apparatuses, systems, and techniques to perform one or more operations using a tensor. In at least one embodiment, one or more circuits are to perform an application programming interface (API) to cause parameters indicating operands and dimensions of operands of two or more tensor contractions to be stored. In at least one embodiment, one or more circuits are to perform an application programming interface (API) to cause two or more tensor contractions to be performed based, at least in part, on one or more input parameters of the API.
Apparatuses, systems, and methods to perform matrix multiplication. In at least one embodiment, a matrix multiplication is performed based on an indication of whether information to be used by part of the matrix multiplication has been loaded.
Systems and methods for a datacenter cooling system are disclosed. In at least one embodiment, a secondary cooling loop interfaces with a primary cooling loop and includes at least one processor that is adapted with a first proportional-integral-derivative (PID) controller to control a first flow controller, which can cause a flow rate for a secondary coolant from a coolant distribution unit (CDU) to a plurality of second flow controllers, and where the plurality of second flow controllers are associated with a plurality of second PID controllers and the flow rate is based in part of on feedback from the plurality of second PID controllers.
Aspects of the present invention are directed to computer-implemented techniques for improving the training of artificial neural networks using a reduced precision (e.g., float16) data format. Embodiments of the present invention rescale tensor values prior to performing matrix operations (such as matrix multiplication or matrix addition) to prevent overflow and underflow. To preserve accuracy throughout the performance of the matrix operations, the scale factors are defined using a novel data format to represent tensors, wherein a matrix is represented by the tuple X, where X=(a, v[.]), wherein a is a float scale factor and v[.] are scaled values stored in the float16 format. The value of any element X[i] according to this data format would be equal to a*v[i].
In various examples, systems and methods are disclosed relating to dexterous arm-hand grasping with geometric fabrics. One or more processors can update, during a first update stage, a teacher model to generate first actions for a geometric fabric associated with a simulated autonomous machine of a simulation using state information of the simulation. During a second update stage, the processors can update a student model to generate second actions for the geometric fabric using at least one rendered image of the simulation, the teacher model, and noised state information of the simulation. The student model and the geometric fabric can be used to control a physical autonomous machine with respect to a physical object based at least on an image of an environment including the physical autonomous machine and the physical object.
In various examples, systems and methods are disclosed relating to disclosed relating to dexterous arm-hand grasping with geometric fabrics. One or more processors can cause a teacher model to generate first actions for a geometric fabric associated with a simulated autonomous machine in a simulated environment using state information of the simulated environment and position information of a simulated object in the simulated environment. Using the teacher model and a depth image of the simulated environment, a student model can be updated to generate second actions for the geometric fabric associated with the simulated autonomous machine. A depth image of an environment can be provided as input to the student model to cause the student model to infer at least one action to control a physical autonomous machine with respect to a physical object using the geometric fabric.
The hybrid-head architecture model can be used to train a language model (LM). It uses a combination of attention heads and state space models (SSMs) to improve the speed and efficiency of inferencing a received input sequence. This disclosure combines the high-resolution recall capabilities of attention heads with the efficient context summarization of SSM heads. The model can be separated into a set of layers, and the input sequence can be processed layer by layer. Each layer can have its own number of attention heads and SSM heads. Fine-tuning and optimization can be applied to each layer, as well as normalization and scaling. To further optimize the performance of the hybrid-head architecture model, learnable meta tokens can be used, which act as a learned cache for attention and SSM heads, enhancing the model's focus on salient information. The attention heads and the SSMs can be processed in parallel.
Apparatuses, systems, and techniques to generate a memory allocation plan for a set of tensors. In at least one embodiment, tensor data corresponding to the set of tensors is stored into memory locations at run time based, at least in part, on the memory allocation plan generated at a compile time.
Systems and methods are disclosed for training and using a digital human foundational model (DHFM) comprising a generative adversarial network (GAN) generator. For instance, the method may include obtaining one or more inputs comprising pose information indicating a three-dimensional (3D) pose representation of a human and processing the one or more inputs using a mapping network to generate intermediate latent code. The method may further include processing the intermediate latent code using the trained generator to generate texel-aligned Gaussian maps that align Gaussian attributes to a coarse mesh template of the human and performing linear blend skinning and deformation on the texel-aligned Gaussian maps to obtain modified texel-aligned Gaussian maps. The method may also include processing the modified texel-aligned Gaussian maps using a multi-part renderer to generate a synthetic human representation of the human indicating facial and hand features of the human.
G06T 7/246 - Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
G06T 13/40 - 3D [Three Dimensional] animation of characters, e.g. humans, animals or virtual beings
G06V 10/764 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
G06V 10/774 - Generating sets of training patternsBootstrap methods, e.g. bagging or boosting
Systems and methods are disclosed for generating and curating a training dataset for training one or more machine learning-artificial intelligence (ML-AI) models. For instance, the method may include extracting 2D landmarks of a human from an obtained image that is within the training dataset and extracting 3D poses of the human from the obtained image. The method may further include using camera coordinates associated with the obtained image to project the 3D poses of the human into 2D space and fine-tuning the 3D poses of the human based on comparing the projected 3D poses in 2D space with the extracted 2D landmarks. The method may also include generating labels for the obtained image within the training dataset, augmenting the training dataset with a plurality of generated synthetic images of humans, and training the one or more ML-AI models.
G06T 7/246 - Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
G06T 13/40 - 3D [Three Dimensional] animation of characters, e.g. humans, animals or virtual beings
G06V 10/26 - Segmentation of patterns in the image fieldCutting or merging of image elements to establish the pattern region, e.g. clustering-based techniquesDetection of occlusion
G06V 10/72 - Data preparation, e.g. statistical preprocessing of image or video features
G06V 10/764 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
G06V 10/774 - Generating sets of training patternsBootstrap methods, e.g. bagging or boosting
Various embodiments include techniques for rendering an image. The techniques include allocating a plurality of voxels to represent a scene, sorting the plurality of voxels based on a plurality of associated Morton codes to obtain a rendering order, and rendering the plurality of voxels based on the rendering order to generate an image.
One embodiment of a method for training a machine learning model to control a robot includes causing a model of the robot to move within a simulation based on one or more outputs of the machine learning model, computing an error within the simulation, computing at least one of a reward or an observation based on the error, and updating one or more parameters of the machine learning model based on the at least one of a reward or an observation.
In at least one embodiment, a system includes one or more servers of a datacenter, and a datacenter cooling system. The datacenter cooling system includes a set of sensors, a liquid-to-air heat exchanger, one or more fans, and one or more processors to cause the datacenter cooling system to operate in a cooling mode of a plurality of cooling modes based at least in part on sensor data obtained from the set of sensors. The plurality of cooling modes includes a first mode that causes one or more fans to blow air in a first direction to cool a coolant circulating within the liquid-to-air heat exchanger that is enabled, and a second mode that causes the one or more fans to blow air in a second direction, different from the first direction, to cool the one or more servers, wherein, in the second mode, the liquid-to-air heat exchanger is disabled.
Apparatuses, systems, and techniques to execute CUDA programs. In at least one embodiment, an application programming interface is performed to indicate two or more blocks of threads to be scheduled in parallel.
In various examples, a technique for securely transmitting CAN (Controller Area Network) messages is disclosed that includes receiving, using a cryptographic engine, a message from an application to be transmitted over a CAN (Controller Area Network) bus, wherein the cryptographic engine executes a secure firmware and is implemented on an on-die discrete processor. The technique further includes accessing, using the secure firmware, a key from a plurality of keys associated with an authentication process from a secure memory associated with the cryptographic engine. Additionally, the technique includes computing an authentication tag using the key and the message and transmitting the message with the authentication tag over the CAN bus to a destination address.
Linear-Swept Sphere (LSS) primitives alongside hardware linear curve intersection testing and traversal logic allow traversal of curves directly within a hardware ray tracer. The introduction of the LSS primitive block, LSS primitive ranges, ray-LSS intersection testing, and LSS fetch removes the requirement that curve traversal utilize ray tracer item-ranges for traversal of curve primitives within a BVH and ray-curve intersection testing through software intersection shaders. Definition of an LSS primitive supports enablement of a single LSS endcap, query and return of the ray-LSS exit hit point, and degenerate-shell LSS intersection testing.
Disclosed are systems and techniques for improved task execution throughput by a processor. The techniques include executing, by a processing unit, a first task with associated processor configuration requirements. The techniques further include receiving a second task and a third task to be executed by the processing unit. Each of the second task and the third task has associated processor configuration requirements. The techniques further include determining a first next task to execute by the processing unit based on a comparison of a current processor configuration of the processing unit with the processor configuration requirements of the second task and the third task. The current processor configuration of the processing unit is based on the processor configuration requirements associated with the first task. The techniques further include providing the first next task to the processing unit for execution.
Apparatuses, systems, and methods to perform thread memory addressing. In at least one embodiment, a processor, comprising: one or more circuits to perform an application programming interface (API) to cause one or more instructions to be performed based, at least in part, one or more API parameters indicating a size of one or more operands to be used by the one or more instructions.
Apparatuses, systems, and techniques to generate blue noise masks for real-time image rendering and enhancement. In at least one embodiment, a vector-valued noise mask is generated and applied to one or more images to generate one or more enhanced images for image processing (e.g., real-time image rendering). In at least one embodiment, the noise mask includes vector values per pixel and is able to handle the temporal domain (e.g., add time to the spatial domain) to improve image quality when rendering images over multiple frames.
A method includes generating semantic data corresponding to appearance features of a person within a first image. One or more models and the semantic data are used to generate attribute features of the person. The one or more models, the semantic data, and the attribute features are used to generate an embedding. The one or more models and the embedding are used to identify the person within a second image.
Apparatuses, systems, and techniques to dynamically perform neural network portions. In at least one embodiment, one or more portions of one or more neural networks are dynamically performed by two or more processors based on, for example, one or more performance metrics estimated of the one or more portions of the one or more neural networks.
Apparatuses, systems, and techniques to use one or more neural networks to generate texture maps are described. In at least one embodiment, one or more neural networks are used to generate one or more texture maps of a second resolution based, at least in part, on one or more texture maps of a first resolution, less than the second resolution.
Apparatuses, systems, and methods to modify compressed videos. In at least one embodiment, a processor includes circuitry to perform one or more instructions to cause information to be stored and to store an indication of one or more decompression algorithms to decompress the information.
H04N 19/423 - Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation characterised by memory arrangements
G06F 9/50 - Allocation of resources, e.g. of the central processing unit [CPU]
In various examples, data collection vehicles or machines may be equipped with one or more LiDAR sensors (and/or other sensors), and the LiDAR sensor(s) may be used to collect frames of LiDAR data representing various real-world conditions. The LiDAR data may be processed using one or more deep neural networks (DNNs) such as a transformer neural network to generate auto-labels representing detected dynamic obstacles of any designated class. Tracking may be applied to generate object tracks (tracklines), estimate velocity, and/or handle occlusions. In some embodiments, the object tracks may be refined based on geometry and/or confidence to improve their accuracy. In some embodiments, the auto-labels are classified to generate an estimated representation of quality, and auto-labels with at least a threshold quality score may be skipped during human labeling. As such, auto-label quality scores may be used to accelerate human validation of auto-labeled scenes by skipping high quality auto-labels.
Embodiments described herein provide a hybrid data center cooling system. In at least one embodiment, a data center cooling system includes one or more air-and-liquid heat exchangers to use one or more first coolant flows to transfer heat from one or more heated air flows and one or more heated second coolant flows returned from one or more computing hardware.
Apparatuses, systems, and techniques to use one or more neural networks to cause one or more error detection and correction (EDC) algorithms to be performed for one or more radio access network (RAN) signals. In at least one embodiment, one or more EDC algorithms are to be performed on said one or more RAN signals based, at least in part, on one or more quality indicators of said one or more RAN signals.
Processors, systems and techniques to perform different inferencing tasks using a unified multi-task inferencing mode are disclosed. In at least one embodiment, multiple task-specific models are created through fine tuning of a common pre-trained model, a single unified model is derived from the multiple task-specific models, and a task-specific model is then reconstructed from the unified model according to a type of inferencing task to be performed.
Systems and methods are disclosed that use a Foundational Stereo Model to generate an output disparity map. The Foundational Stereo Model includes side-tuning adapters (STA) that utilize a vision transformer (ViT) and a convolutional neural network (CNN) to generate feature maps. Specifically, the CNN may be used to adapt the ViT-based monocular depth estimation network for the stereo setup, which synergizes the strengths of both CNN and ViT architectures. In addition, the Foundational Stereo Model includes an attentive hybrid cost filtering (AHCF) that uses two branches that also utilizes the advantages of both a transformer architecture and the CNN architecture. Furthermore, the Foundational Stereo Model may perform iterative refinement of an initial disparity map to obtain the output disparity map based on performing a convolutional gated recurrent unit (GRU) operation.
Processors, systems and techniques to efficiently select numeric encoding formats of varying precision for processing within neural networks is disclosed. In at least one embodiment, tensor data of respective layers of a neural network may be quantized according to a plurality of low precision formats, either during or subsequent to training, and a preferred encoding format selected according at least to quantization error of the tensor data.
In various examples, a technique for performing unified generative and discriminative learning includes converting, via execution of a machine learning model, a plurality of training data samples into a first plurality of latent representations. The technique also includes computing one or more losses based on the plurality of latent representations, wherein the loss(es) include a contrastive term that approximates an expected similarity between a latent representation of a training data sample and a second plurality of latent representations associated with a distribution of training data samples that includes the plurality of training data samples. The technique further includes updating one or more parameters of the machine learning model based on the one or more losses to generate a trained machine learning model.
In various examples, a technique for generating an embedding of a data sample comprises generating, via execution of an encoder included in a trained machine learning model, a latent representation of the data sample. The technique also includes converting, via execution of one or more hidden layers within a decoder included in the trained machine learning model, the latent representation into one or more sets of hidden outputs. The technique further includes generating an embedding of the data sample based on at least a portion of the one or more sets of hidden outputs and causing a task-based output to be generated based on the embedding of the data sample.
Embodiments of the present disclosure relate to applications, platforms, architecture, etc. for assessing annotations generated by annotators in the evaluation of generated outputs. In particular one or more of a first ground truth annotation or a second ground truth annotation corresponding to one or more generated outputs may be obtained. The first ground truth annotation may be based at least on a plurality of assessment annotations and the second ground truth annotation may correspond to an expert related to the one or more generated outputs. Further, one or more assessments related to one or more assessment annotations of the plurality of assessment annotations may be determined based at least on one or more of the first ground truth annotation or the second ground truth annotation. In addition, one or more annotator adjustment operations may be performed based at least on the one or more assessments.
Apparatuses, systems, and techniques to perform third party management of operational thread memory requests within one or more processors by use of a server to manage said memory and memory requests to perform dynamic memory allocation during runtime without overwrite risk, according to at least one embodiment. In at least one embodiment, processor scomprising one or more circuits to cause one or more threads to use one or more virtual storage location addresses to access one or more physical storage locations based, at least in part, on information indicating whether the one or more physical storage locations are allocated to the one or more threads.
Apparatuses, systems, and techniques to optimize an agentic artificial intelligence (AI) system, based at least in part on one or more user-configurable parameters. In at least one embodiment, at least one user-configurable parameter is used by one or more compilers to cause generation of at least one optimized graph representation of at least a portion of an agentic AI systems based, at least in part, on at least one performance metric associated with the one or more portions of the one or more agentic AI systems.
A power control circuit to control data center power delivery is disclosed. In at least one embodiment, the power control circuit may be configured to selectively cause at least a portion of a supplied power to be directed to one or more server racks or to charge one or more batteries based, at least in part on, one or more charge levels corresponding to the one or more batteries.
H02J 7/02 - Circuit arrangements for charging or depolarising batteries or for supplying loads from batteries for charging batteries from AC mains by converters
H02J 9/06 - Circuit arrangements for emergency or stand-by power supply, e.g. for emergency lighting in which the distribution system is disconnected from the normal source and connected to a standby source with automatic change-over
In various examples, systems and methods are disclosed relating to part-invariant peak power management. One or more circuits can receive a plurality of instructions for a graphics processing device. The plurality of instructions can correspond to a respective plurality of power consumption values. The one or more circuits can determine that the respective plurality of power consumption values cause a threshold to be exceeded during a time period. The one or more circuits can generate a control signal to control a clock signal for the graphics processing device responsive to determining that the respective plurality of power consumption values cause the threshold to be exceeded.
Apparatuses, systems, methods, and/or techniques to identify anomalies in a computer network topology. In at least one embodiment, a first portion of a computer network is used to train a model and the train model is applied to a second portion of the network to predict link connections in the second portion. In at least one embodiment, such predictions are used to identify topology amomalies.
H04L 41/0631 - Management of faults, events, alarms or notifications using root cause analysisManagement of faults, events, alarms or notifications using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
H04L 41/12 - Discovery or management of network topologies
H04L 41/16 - Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
76.
MAPPING TABLES FOR APPLICATION PROGRAMMING INTERFACES CALLED FROM NEURAL NETWORK AGENTS
Apparatuses, systems, and techniques to cause mapping of tokens generated by neural networks for application programming interfaces (APIs). In at least one embodiment, a processor includes one or more circuits which cause one or more tables to map one or more tokens generated by one or more neural networks into one or more tokens to be used by one or more APIs.
Apparatuses, systems, and techniques to use one or more nueral networks to image object features are described. In at least one embodiment, one or more neural networks are used to generate one or more image object features based, at least in part, on a varaible sample rate of one or more audio signals.
G10L 19/04 - Speech or audio signal analysis-synthesis techniques for redundancy reduction, e.g. in vocodersCoding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
Apparatuses, systems, and techniques to simulate communicated between wireless devices and base stations. In at least one embodiment, communication between wireless devices and base stations are simulated based on a grouping of wireless devices according to shared communication frequency bands.
Systems and methods are disclosed that curate synthetic datasets using a stereo model. For example, embodiments of the present disclosure may include an iterative process that generates a first synthetic dataset using a replicator composer, uses the generated dataset to train a stereo model such as a Foundational Stereo model, and curates a second synthetic dataset using the trained stereo model. Additionally, and/or alternatively, the synthetic datasets that were generated by the replicator composer may include multiple categories of realism such as realistic style synthetic data and chaotic style synthetic data as well as multiple categories of use cases such as navigation, driving, and manipulation. Additionally, and/or alternatively, the replicator composer may generate a scene based on determining a center of mass of objects that are generated for the scene and using the center of mass to orient the camera.
In various examples, a technique for generating a hierarchical representation of a document includes generating, via a first machine learning model, a hierarchical structure associated with a document, wherein the hierarchical structure includes one or more headings and one or more paragraphs. The technique also includes identifying, via the first machine learning model, heading text included in the document and associated with each of the one or more headings and paragraph text included in the document and associated with each of the one or more paragraphs. The technique further includes generating, via a second machine learning model and based at least on the identified heading text included in the document, a formatted listing including the heading text associated with each of the one or more headings and generating a hierarchical document based at least on the formatted listing and the paragraph text associated with the one or more paragraphs.
In various examples, a technique for querying a hierarchical representation of a document includes receiving a hierarchical document tree including one or more parent and child nodes each associated with a portion of a source document and generating, using a machine learning model and based at least on a query prompt and contents of each of the one or more parent and child nodes, a similarity score for each of the one or more parent and child nodes. The technique also includes calculating combined similarity scores associated with each of one or more child nodes based at least on the similarity score associated with the child node and one or more similarity scores associated with one or more nodes having a parent relationship to the child node and generating a query result based at least on the combined similarity scores associated with the one or more child nodes.
Scene reconstruction is a computer vision process that creates a model of a scene from a given input, usually including creating a three-dimensional (3D) scene model from one or more input two-dimensional (2D) images of the scene. High-quality scene reconstruction and rendering is useful for various applications, such as autonomous agent applications and scene editing applications. Existing scene reconstruction methods encounter limitations in dynamic scenes where moving objects are not consistent across views from different times, and these methods generally lack the motion cues essential for effective object-environment decomposition and further lead to incomplete environment reconstruction of novel views from persistent occlusion of environmental structures. The present disclosure integrates spatial memory from prior traversals of a scene when learning a representation of a temporally changing scene, which can provide observations of occluded areas and contextual information for traffic participants for more comprehensive and efficient scene representation learning.
Systems and methods are disclosed for training and using a digital human foundational model (DHFM) comprising a generative adversarial network (GAN) generator. For instance, the method may include processing training inputs using the GAN generator to generate texel-aligned Gaussian maps that align Gaussian attributes to a coarse mesh template of the human and rendering a synthetic human representation of the human based on the texel-aligned Gaussian maps. The synthetic human representation comprises a full-bodied representation of the human indicating facial and hand features of the human. The method also includes processing the synthetic human representation using one or more discriminators to generate one or more discriminator outputs, computing one or more losses based on the texel-aligned Gaussian maps and the one or more discriminator outputs, and training the GAN generator using the one or more losses.
G06T 7/246 - Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
G06T 7/73 - Determining position or orientation of objects or cameras using feature-based methods
G06T 13/40 - 3D [Three Dimensional] animation of characters, e.g. humans, animals or virtual beings
G06V 10/72 - Data preparation, e.g. statistical preprocessing of image or video features
G06V 10/764 - Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
G06V 10/774 - Generating sets of training patternsBootstrap methods, e.g. bagging or boosting
Apparatuses, systems, and techniques to generate one or more data items related to an object from a set of multiple data items related to the same object using a generative adversarial network. In at least one embodiment, data objects in a set are disentangled into common and unique components and encoded such that they may be used to train one or more neural networks to generate missing data from the data objects in the set, such as generating missing medical images from a set of related medical images.
Various embodiments include techniques for rendering an image. The techniques include allocating a plurality of voxels to represent a scene, sorting the plurality of voxels based on a plurality of associated Morton codes to obtain a rendering order, and rendering the plurality of voxels based on the rendering order to generate an image.
Systems including a pelvis depth estimation model configured to generate a pelvis depth estimate of a person depicted in an image, a human mesh estimation model configured to generate a body mesh corresponding to the person depicted in the image given the estimated pelvis depth, and a camera solver configured to apply differentiable rasterization to derive camera parameters for the image.
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.
Techniques for performing low-rank self-calibration of 3D geometric foundation models include receiving a plurality of unlabeled images of a scene, generating a pair of point maps and a pair of confidence maps for a first pair of unlabeled images of the plurality of unlabeled images, determining intrinsic camera parameters for the first pair of unlabeled images, refining the pair of point maps based on the intrinsic camera parameters to generate a refined pair of point maps, generating pseudo-labels for the first pair of unlabeled images based on the refined pair of point maps and the pair of confidence maps, and fine tuning a pretrained machine learning model based on the pseudo-labels to generate a fine-tuned machine learning model.
Multimodal large language models (MLLMs) have evolved to interpret visual elements, progressing from text prompts for holistic image understanding to sophisticated approaches for region-level understanding. However, a key limitation of existing methods is the reliance on representations that may not consistently capture regions across frames, particularly when aiming for a unified solution for both images and videos. The present disclosure unifies image and video region-level understanding by an LLM via token marks.
Spatio-temporal reconstruction modeling includes receiving images of a scene, dividing each of the images into patches; generating an image token for each patch; appending one or more motion tokens to the image tokens to generate an input token vector; processing the input token vector with a machine learning (ML) model to generate an output token vector with output image and motion tokens; decoding each output image token to generate a 3D Gaussian and a motion key; decoding each output motion token to generate a velocity basis and a motion query; generating of velocity vectors based on the motion queries and the motion keys; generating a 2D image for a first timestep based on the 3D Gaussians and the velocity vectors; training the ML model based on the 2D image; generating optimized 3D Gaussians using the trained ML model; and generating a dynamic reconstructed 3D scene from the optimized 3D Gaussians.
One embodiment of a method for generating scene representations includes processing a first image using a first trained machine learning model to generate one or more second images, processing the one or more second images using a second trained machine learning model to generate three-dimensional (3D) geometry and camera information, and generating a four-dimensional (4D) scene representation based on the 3D geometry and the camera information.
Apparatuses, systems, and techniques to perform a matrix multiplication using parallel processing. In at least one embodiment, a matrix multiplication is divided into a set of tiles, with each tile processed with a prolog task, a calculation task, and an epilog task. The prolog tasks are performed by a dedicated set of threads, with the remaining tasks performed in an interleaved manner using two or more thread groups.
The disclosure provides an apparatus, system, and method that augments the typical processing by a storage controller for obtaining data. An augmented data controller is disclosed that provides augmented instructions to a storage controller to improve the processing of data requests, such as fine-grained data requests. The augmented data controller provides improved features for processing data requests, such as a request format that groups a set of commands from a requesting agent to the storage controller, and/or groups a set of responses from the storage controller for the requesting agent to process. Semantic guidance for data storage is provided and interaction with the storage controller is extended to support a hierarchical protocol that aggregates across multiple requests. With the augmented storage controller, additional information can be tracked, such as a set of memory regions and sets of data requests.
The disclosure provides an apparatus, system, and method that augments the typical processing by a storage controller for storage access requests. An augmented data controller is disclosed that provides completion guidance to improve the processing of storage access requests, such as fine-grained data requests. The completion guidance can be generated based on semantic information provided by requesting agents and may be sent with storage access requests that are for a data read or write. The completion guidance allows software-defined completion notifications that are beneficial for accessing data, such as for multiple fine-grained (e.g. 4 KB or less), random sparse accesses that are emerging from processors, such as from GPU threads. The completion guidance is a result of the requesting agents providing additional contextual information for storage access requests to improve the intelligence of storage controllers and improve the efficiency of accessing data storage for obtaining or writing data.
The disclosure provides an apparatus, system, and method that augments the typical processing by a storage controller for storage access requests. An augmented data controller is disclosed that provides completion guidance to improve the processing of storage access requests, such as fine-grained data requests. The completion guidance can be generated based on semantic information provided by requesting agents and may be sent with storage access requests that are for a data read or write. The completion guidance allows software-defined completion notifications that are beneficial for accessing data, such as for multiple fine-grained (e.g. 4KB or less), random sparse accesses that are emerging from processors, such as from GPU threads. The completion guidance is a result of the requesting agents providing additional contextual information for storage access requests to improve the intelligence of storage controllers and improve the efficiency of accessing data storage for obtaining or writing data.
The disclosure provides an apparatus, system, and method that augments the typical processing by a storage controller for storage access requests. An augmented data controller is disclosed that provides completion guidance to improve the processing of storage access requests, such as fine-grained data requests. The completion guidance can be generated based on semantic information provided by requesting agents and may be sent with storage access requests that are for a data read or write. The completion guidance allows software-defined completion notifications that are beneficial for accessing data, such as for multiple fine-grained (e.g. 4 KB or less), random sparse accesses that are emerging from processors, such as from GPU threads. The completion guidance is a result of the requesting agents providing additional contextual information for storage access requests to improve the intelligence of storage controllers and improve the efficiency of accessing data storage for obtaining or writing data.
A processor includes one or more circuits to determine that a measurement of a pressure of a liquid-phase of a refrigerant in a high pressure portion of a cooling loop exceeds a first threshold and to determine that a temperature associated with one or more cold plates associated with the cooling loop is below a second threshold. The processor is to enable a pressure control system of an in-rack refrigerant distribution unit (RDU) including a pressure management valve to provide pressure of the liquid-phase of the refrigerant from the high pressure portion of the cooling loop to a low pressure portion of the cooling loop to cause a pressure-drop, before an expansion valve, for the liquid-phase of the refrigerant.
Systems and methods herein are for hose management in a computing module. A rotatable support structure includes a moveable inset structure and is to support rotation movement of cooling hoses that are to cool underlying devices in the computing module, where the rotation movement is about an axis of the computing module to allow the cooling hoses to be moved away from the underlying devices of the computing module and where the movable inset structure is movable within the rotatable support structure to receive tension on the cooling hoses from being coupled to a manifold of the computing module.
In various examples, methods and systems are provided for sampling and transmitting the most useful information from a radar signal representing a scene while staying within the computational and storage confines of a standard automotive radar sensor and the bandwidth constraints of a standard communication link between a radar sensor and processing unit. Disclosed approaches may select a patch of frequency bins that correspond to radar signals based at least on proximities of the frequency bins to one or more frequency bins corresponding to at least one peak and/or detection point in the radar signals. Data representing samples corresponding to the patch of frequency bins may be transmitted to the processing unit and applied to one or more machine learning models in order to accurately classify, identify, and/or track objects.
A cluster storage system including servers containing storage processing units (SPUs) can create synchronized snapshot sets for the volumes that the SPUs maintain and can report the snapshot sets to a cloud-based service. Each snapshot in a set reflects the state a corresponding volume had at a rollback point corresponding to the set. A user of the storage system contacts the cloud-based service about recovery of the storage system, and the cloud-based service may present the user with a list of rollback points corresponding to the synchronized snapshot sets. The user may select to recover the storage system to any of the rollback points, and the SPUs promote the selected snapshots to replace the volumes for storage services.
