Abstract: A manufacturing method for a semiconductor device includes: forming an etching termination layer, a first dielectric layer, an auxiliary dielectric layer and a second dielectric layer which are successively stacked from bottom to top; by taking a photoresist layer as an etching barrier layer, patterning the second dielectric layer to obtain a first opening pattern, the bottom of the first opening being provided with a second opening pattern exposing part of the auxiliary dielectric layer; forming a first trench passing through the second dielectric layer and the auxiliary dielectric layer and extending to the first dielectric layer, and forming a second trench passing through the first dielectric layer from the bottom of the first trench and extending to the etching termination layer; and forming a conductive layer in the first and second trenches.

Abstract: The present disclosure relates to the technical field of plasma cleaning, and provides a plasma cleaning head for an optical member and a plasma cleaning device, including a plasma circulation assembly having a hollow shape and a lens frame mounted on the plasma circulation assembly, wherein the plasma circulation assembly is provided with an outlet port which is larger than an ejection port adapted to eject a plasma and is adapted to diffuse the plasma, and the lens frame is in communication with and surrounds the outlet port, and the lens frame is adapted to embed an optical lens body so that the optical lens body is arranged opposite to the outlet port to help improve dispersion and stability of the plasma flowing from the outlet port to the optical lens body, and helps to promote the energy absorption of the plasma by the optical lens body.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Methods and apparatus for reducing a size of a neural network model, the method including: compressing data of the neural network model; identifying structure information of a vector register, wherein the structure information includes a number of registers included in the vector register; comparing a number of elements in the compressed data with a first condition, wherein the first condition is determined based on the number of registers in the vector register; and in response to the number of elements satisfying the first condition, associating the compressed data with the vector register to enable loading the compressed data to the vector register.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Methods and apparatus for reducing a size of a neural network model, the method including: compressing data of the neural network model; identifying structure information of a vector register, wherein the structure information includes a number of registers included in the vector register; comparing a number of elements in the compressed data with a first condition, wherein the first condition is determined based on the number of registers in the vector register; and in response to the number of elements satisfying the first condition, associating the compressed data with the vector register to enable loading the compressed data to the vector register.

Abstract: A device and system for detecting a touch panel are used for detecting a metal mesh on the touch panel, where row direction channels and column direction channels on the metal meshes each include a plurality of touch units which are connected with each other. In the device for detecting a touch panel, a signal transceiving component includes a signal generating unit and a signal receiving unit, which are configured to be arranged on the same side of the metal meshes in preset relative positions, and send and receive a signal through capacitive coupling with the metal mesh, respectively; a defect detecting unit is configured to detect each of a plurality of touch units in the metal meshes based on a detection signal received by the signal receiving unit.

Abstract: A spatial prediction method of rice stable isotope based on environmental similarity is provided. The method comprises: describing environmental characteristics of the rice stable isotope; measuring an environmental similarity between a site to be predicted and a sample site; measuring a reliability of the sample site; and carrying out a spatial prediction on the rice stable isotope according to the environmental similarity between the site to be predicted and the sample site and the reliability of the sample site. This method can improve the accuracy of the prediction result.

Abstract: The present application relates to a method, device, and system for detecting welding spot quality abnormalities based on deep learning. The method includes: acquiring a dynamic welding parameter in a welding process corresponding to any target welding spot; inputting the dynamic welding parameter into a pre-trained dynamic welding parameter simulation model for simulation, and acquiring a welding simulation parameter output by the dynamic welding parameter simulation model; determining a deviation of the dynamic welding parameter from the welding simulation parameter, and determining that the target welding spot is an abnormal welding spot when the deviation is greater than a preset threshold. The solution of the present application can reduce the frequency of manual tearing down and batches for abnormality detection, which has a faster abnormality detection speed and may cover all welding spots.

Abstract: Systems, methods, and apparatuses relating to operations in a configurable spatial accelerator are described. In one embodiment, a configurable spatial accelerator includes a first processing element that includes a configuration register within the first processing element to store a configuration value that causes the first processing element to perform an operation according to the configuration value, a plurality of input queues, an input controller to control enqueue and dequeue of values into the plurality of input queues according to the configuration value, a plurality of output queues, and an output controller to control enqueue and dequeue of values into the plurality of output queues according to the configuration value.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Systems, methods, and apparatuses relating to operations in a configurable spatial accelerator are described. In one embodiment, a configurable spatial accelerator includes a first processing element that includes a configuration register within the first processing element to store a configuration value that causes the first processing element to perform an operation according to the configuration value, a plurality of input queues, an input controller to control enqueue and dequeue of values into the plurality of input queues according to the configuration value, a plurality of output queues, and an output controller to control enqueue and dequeue of values into the plurality of output queues according to the configuration value.

Abstract: A device and system for detecting a touch panel are used for detecting a metal mesh on the touch panel, where row direction channels and column direction channels on the metal meshes each include a plurality of touch units which are connected with each other. In the device for detecting a touch panel, a signal transceiving component includes a signal generating unit and a signal receiving unit, which are configured to be arranged on the same side of the metal meshes in preset relative positions, and send and receive a signal through capacitive coupling with the metal mesh, respectively; a defect detecting unit is configured to detect each of a plurality of touch units in the metal meshes based on a detection signal received by the signal receiving unit.

Abstract: Systems, methods, and apparatuses relating to operations in a configurable spatial accelerator are described. In one embodiment, a configurable spatial accelerator includes a first processing element that includes a configuration register within the first processing element to store a configuration value that causes the first processing element to perform an operation according to the configuration value, a plurality of input queues, an input controller to control enqueue and dequeue of values into the plurality of input queues according to the configuration value, a plurality of output queues, and an output controller to control enqueue and dequeue of values into the plurality of output queues according to the configuration value.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Methods and apparatus for reducing a size of a neural network model, the method including: compressing data of the neural network model; identifying structure information of a vector register, wherein the structure information includes a number of registers included in the vector register; comparing a number of elements in the compressed data with a first condition, wherein the first condition is determined based on the number of registers in the vector register; and in response to the number of elements satisfying the first condition, associating the compressed data with the vector register to enable loading the compressed data to the vector register.

Abstract: Embodiments are directed to techniques to determine dataflow graph instructions comprising one or more pick/switch instruction pairs and generate a reverse static single assignment graph based on the dataflow graph instructions, the reverse static single assignment graph comprising strongly connected components, each of the strongly connected components associated with at least one of the one or more pick/switch instruction pairs. Embodiments also include traversing the reverse static single assignment graph depth-first, and replace pick/switch instructions associated with strongly connected components having configuration values with compound instructions.

Abstract: A verification request is received. In response to receiving the verification request, a first character string is obtained. The first character string comprises one or more variable characters. At least one of the one or more variable characters in the first character string is replaced with at least one backup character to generate a second character string based on multiple pre-established corresponding relationships. A verification code is generated based on the second character string. A user corresponding to the verification request is verified based on a user input corresponding to the verification code.

Abstract: A verification request is received. In response to receiving the verification request, a first character string is obtained. The first character string comprises one or more variable characters. At least one of the one or more variable characters in the first character string is replaced with at least one backup character to generate a second character string based on multiple pre-established corresponding relationships. A verification code is generated based on the second character string. A user corresponding to the verification request is verified based on a user input corresponding to the verification code.