Abstract: Methods, apparatus, and processor-readable storage media for generating 3D images using machine learning and generative artificial intelligence are provided herein. An example computer-implemented method includes obtaining at least one 2D image; determining one or more features of the at least one 2D image by processing at least a portion of the at least one 2D image using a first machine learning technique; generating multiple 3D visualizations associated with the at least a portion of the at least one 2D image by processing at least a portion of the one or more features using one or more generative artificial intelligence techniques; selecting at least one of the multiple 3D visualizations by processing at least a portion of the multiple 3D visualizations using a second machine learning technique different from the first machine learning technique; and performing one or more automated actions based on the at least one selected 3D visualization.

Abstract: The present invention pertains to a thermoelectric material which exhibits a ZT of ?0.3 and which is exposed to an external magnetic field of 0.01 T-2 T resulting in a ZT of ?1.3 at a temperature of ?300 K. The present invention pertains to thermoelectric materials at a temperature of ?300 K under an applied external magnetic field of 0.01 T-2 T, in which the material includes: a three-dimensional topological insulator, or topological semimetal having a carrier mobility of ?104 cm2/Vs at 20K and a carrier concentration of 1017-1020/cm3, and an effective mass?0.04 free electron mass, and a Fermi energy of ?100 meV. The invention further relates to methods of making such thermoelectric material at a low magnetic field of less than 2 T.

Abstract: The present disclosure relates to a method for interaction operation control of a deformable object based on a vision-tactile-language-action multimodal model, including: encoding a visual image, tactile data, and language data for the deformable object to obtain a visual feature, a tactile feature, and a language feature, performing cross-modal feature alignment processing on the visual feature, the tactile feature, and the language feature to obtain a multimodal fusion feature, inputting the multimodal fusion feature into a large model for environment understanding, adopting a planning manner of ‘thinking-decision’ to iteratively perform action planning and execution, and repeating the above operation steps until an interaction operation task of the deformable object is completed.

Abstract: The present disclosure relates to a training method of a large model for wire harness operation of a humanoid robot based on a meta-action, which includes: constructing wire-harness-operation meta-actions based on human wire harness operations; constructing a wire-harness-operation meta-action dataset; training the large model for the wire harness operation of the humanoid robot using reinforcement learning based on the wire-harness-operation meta-action dataset; acquiring input data, outputting joint-motor parameters using the trained large model for the wire harness operation of the humanoid robot and updating the input data in real time based on the joint-motor parameters; acquiring a basic strategy and acquiring a residual strategy based on the basic strategy; and performing the wire harness operation of the humanoid robot based on the basic strategy and the residual strategy.

Abstract: Provided is a real-time control method for robot manipulation actions based on collaboration between a large model and a small model.

Abstract: A processing system and method for executing arithmetic and conversion operations involving 16-bit brain floating-point (BF16)-formatted data are described. An instruction specifying either an arithmetic or conversion operation and a first data element in BF16 data format are received. For arithmetic operations, the exponents of the data elements are aligned, and a result is generated using the aligned exponents. For conversion operations, the mantissa of the BF16 data element is scaled based on its exponent, and the element is converted to a second data format, such as FP32 or a reduced-precision format, using precision-aware scaling and rounding. The result is stored in operations registers, such as for additional processing.

Abstract: Systems and techniques for providing mixed-precision matrix multiplication in multi-chiplet processors recognize different precision formats of matrices to be multiplied based on, e.g., parameters provided with instructions or start and end memory locations of the matrices. A plurality of different multiplication chains are provided for different formats such that mixed-precision matrix multiplication can be performed using multiplication chains configured to handle multiplication of different precision formats. The multiplication chains are automatically selected based on the precision formats of the matrices to be multiplied, enabling programmers to utilize the chains without having to directly access the individual multiplication chains.

Abstract: Techniques are disclosed for transposing and loading 6-bit floating-point matrix data into operations registers of a processing unit. A matrix comprising 6-bit floating-point data elements is received from memory, in which the matrix is stored in either a row-major or column-major layout. A matrix transposition loading operation is performed during the loading process, rearranging the matrix elements into a transposed layout (e.g., converting column-major to row-major). The transposed matrix is stored in operations registers for use in parallel processing tasks. The process may include caching partially stored data elements from memory and combining them with subsequently retrieved data to complete the transposition.

Abstract: Systems and techniques for providing co-issue of instructions utilize a scheduler associated with a compute unit to select one double-precision (i.e., 64-bit) instruction and one single-precision (i.e., 32-bit) instruction for issue to and execution at a compute unit. Each compute unit includes or is associated with one or more pairs of double-precision arithmetic logic units (ALUs) and single-precision ALUs. The selected double-precision and single-precision instructions are associated with different threads or waves such that no dependency can exist between the two instructions. The selected single-precision instruction may perform address calculations or other memory tasks while the double-precision instruction may perform data computations that may be required for, e.g., matrix multiplication, or other machine learning functionality.

Abstract: An accelerator unit (AU) including vector registers and one or more processor cores is configured to schedule instructions for execution that share one or more matrix block. To this end, the AU maintains tracking entries for the hardware buffers of one or more of these processor cores with each tracking entry indicating vector register addresses associated with the corresponding matrix block loaded into the hardware buffer. When the AU schedules an instruction indicating a matrix multiplication operation for execution, the AU compares the vector register addresses indicated in the instruction to the tracking entries of the hardware buffers. In response to the vector register addresses in the instruction matching a tracking entry, the AU suppresses a read request to the vector registers and uses data from a corresponding hardware buffer to perform the matrix multiplication operation.

Abstract: One or more cache lines are configured to store a lookup table (LUT) having a plurality of floating-point numbers that are represented by a first number of bits. One or more multiplexers are connected to the cache line(s). The multiplexer(s) is/are configured to select one of the floating-point numbers based on an integer index having a second number of bits that is less than the first number of bits. The integer index is a quantized representation of the selected one of the floating-point numbers. In some cases, a memory is configured to store a matrix of integer indices that are generated by quantizing floating-point numbers that represent data associated with a machine learning algorithm. The integer index is provided by the matrix of the integer indices.

Abstract: The present invention relates to a vision-language large model-based multi-robot collaborative navigation method and system. The method includes steps of: according to a task instruction input by a user, combined with a semantic map constructed by each robot, allocating subtask instructions to robots after parsing by a large language model; acquiring, by each robot, an environmental image in real time, according to the allocated subtask instruction, parsing and predicting a next action using a vision-language navigation large model, and executing a corresponding action and updating a state by the robot; and monitoring a robot state and checking task progress in real time through a human-machine interface and adjusting a task or handling an exception during execution when needed by an operator, and dynamically adjusting, by the robot, an execution strategy according to feedback information to optimize navigation for task completion.

Abstract: The present invention relates to a robot action generation method and system combining general and specialized models, where the method includes: constructing the general model and the specialized model, pre-training the general model, performing parameter fine-tuning on a pre-trained general model, and training the specialized model based on a fine-tuned general model; acquiring a task instruction and real-time visual information, inputting the task instruction and real-time visual information into the fine-tuned general model, and outputting an action sequence and a task latent feature; and acquiring real-time point cloud perception data, and inputting the real-time point cloud perception data, together with the action sequence and the task latent feature, into a trained specialized model, and outputting continuous robot actions. Compared with the prior art, the present invention improves the speed of robot action generation and enhances the generalization of robot action generation.

Abstract: Systems and techniques for providing pseudorandom number generation include generating a pseudorandom number based on a bitwise exclusive-OR, or XOR, function that compares a shifted version of an input seed with one of a plurality of values, where one of the values is selected based on whether a first bit of the input seed is set or unset. An initial seed value is stored in hardware or generated based on an unknown value stored in memory, and successive outputs of the pseudorandom number generation are used as seeds in subsequent pseudorandom number generation operations.

Abstract: An accelerator unit (AU) including vector registers and one or more processor cores is configured to schedule instructions for execution that share one or more matrix block. To this end, the AU maintains multiple tracking entries for each of one or more hardware buffers of the processor cores that indicate the vector register addresses associated with the corresponding matrix blocks loaded into the entries of the hardware buffers. When the AU schedules an instruction indicating a matrix multiplication operation for execution, the AU compares the vector register addresses indicated in the instruction to the tracking entries of a hardware buffer. In response to the vector register addresses of the instruction matching a tracking entry for the hardware buffer, the AU suppresses a read request to the vector registers and uses data from a corresponding entry of the hardware buffer to perform the matrix multiplication operation.

Abstract: The present disclosure relates to a method, a device, and a computer program product. The method includes: in response to receiving a calling command from a client, determining whether the calling command includes a security identifier. The method further includes, in response to determining that the calling command includes a security identifier, determining a security service corresponding to the security identifier in a firmware trusted platform module (FTPM). The method further includes performing security isolation and encryption processing on data related to the calling command by using the security service in the FTPM. In this way, the real intention of the calling command is concealed by the security identifier, which can reduce the possibility of attackers attacking or tampering with the calling command. Therefore, embodiments of the present disclosure enhance the security of the system, protect the device from potential attacks and damage, and prevent unauthorized accesses and malicious operations.

Abstract: A method for training a robot action generation model and a method for generating robot actions. The training method includes: constructing an automotive wire harness operation dataset and extracting human poses; constructing an instruction set based on the dataset and the human poses; pre-training a motion tokenizer; obtaining a text-motion vocabulary; pre-training a language model based on the dataset, the instruction set, and the text-motion vocabulary; and constructing an operation dataset and fine-tuning the pre-trained language model based on the operation dataset to complete training. The method for generating robot actions utilizes the robot action generation model trained by the training method to output a humanoid action sequence and redirects the humanoid action sequence to a robot to generate robot actions. The present disclosure enhances generalizability of the generation model, makes the generation of robot actions more aligned with automotive wire harness operations.

Abstract: A structure detection method and system for a low-light confined space based on mobile-fixed collaborative guidance are provided. The method is implemented by a fixed sensor node network and an unmanned aerial vehicle detection system in a mobile-fixed collaborative system. The method includes: the fixed sensor node network collecting image data of the low-light confined space, generating a basic disease data map, transmitting the basic disease data map to the unmanned aerial vehicle detection system, performing global path planning through a dual-mode adaptive confined space disease detection path planning algorithm to generate a precise detection path and a supplementary detection path; and the unmanned aerial vehicle detection system detecting diseases in the low-light confined space according to the precise detection path and the supplementary detection path through a machine learning algorithm, and obtaining a disease map result.

Abstract: Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for determining a navigation path. The method may include acquiring a source geographical location and a destination geographical location received from a user side device. In addition, the method may include determining a navigation path from the source geographical location to the destination geographical location based on a communication resource heat database, the communication resource heat database including at least a plurality of geographical regions associated with the navigation path and communication resource heat of each of the plurality of geographical regions, the communication resource heat including signal quality, signal strength, and a remaining resource capacity. Then, the method may include sending the determined navigation path to the user side device.

Abstract: Methods and systems are disclosed for executing operations on single-instruction-multiple-data (SIMD) units. Techniques disclosed perform a dot product operation on input data during one computer cycle, including convolving the input data, generating intermediate data, and applying one or more transitional operations to the intermediate data to generate output data. Aspects described, wherein the input data is an input to a layer of a convolutional neural network and the generated output data is the output of the layer.