Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

1434 results about "Floating point" patented technology

In computing, floating-point arithmetic (FP) is arithmetic using formulaic representation of real numbers as an approximation to support a trade-off between range and precision. For this reason, floating-point computation is often found in systems which include very small and very large real numbers, which require fast processing times. A number is, in general, represented approximately to a fixed number of significant digits (the significand) and scaled using an exponent in some fixed base; the base for the scaling is normally two, ten, or sixteen. A number that can be represented exactly is of the following form...

Processor and electronic equipment

The invention discloses a processor and electronic equipment. The processor includes a computing unit including a tensor core configured to perform a matrix multiplication operation using a scaling factor, and a memory, the computing unit further including a scaling factor processing module configured to determine and cache a scaling factor for each tensor associated with the matrix multiplication operation, the computing unit further comprises at least one storage module arranged on a data path between the tensor core and the memory, and the at least one storage module is exclusively occupied by the tensor core when the tensor core executes tensor related operation. And the scaling factor processing module is arranged on the at least one storage module. At present, scaling factors and floating-point number quantization are completed by a vector calculation core, so that performance is reduced, and delay becomes high, and a scaling factor processing module arranged on a storage module in a calculation unit can improve the overall execution efficiency of low-precision matrix multiplication using the scaling factors.
Owner:SHANGHAI BIREN TECH CO LTD

Single-epoch GNSS-oriented PAR ambiguity precision factor determination method

The invention discloses a single-epoch GNSS-oriented PAR ambiguity precision factor determination method, and relates to the technical field of satellite navigation and positioning. Comprising the following steps: dividing pseudo-range and carrier wave observed quantities in a GNSS double-difference mathematical model into two groups, and constructing a function relationship between a floating point solution # imgabs1 # and a floating point solution # imgabs2 # variance-covariance matrix of an ambiguity vector # imgabs0 #; constructing a single-frequency single-epoch single-system P-ADOP expression and establishing a shrinkage factor k de-inversion transformation analytic expression, and simultaneously obtaining a single-frequency single-epoch single-system GNSS non-inverse transformation P-ADOP approximation formula; and performing analogy derivation on the single-frequency single-epoch single-system non-inverse transformation P-ADOP approximate formula to obtain a single-frequency multi-system or multi-frequency single-epoch GNSS non-inverse transformation P-ADOP general approximate formula. The method has important theoretical value and practical guiding significance for GNSS optimal ambiguity subset rapid selection and positioning and partial ambiguity fixing success rate influence mechanism analysis.
Owner:CHINA UNIV OF MINING & TECH

Frequency difference compressor based on precision perception, gradient compression method, equipment and medium

The invention provides a frequency difference compressor based on precision perception, a gradient compression method, equipment and a medium, which are used for carrying out data compression and transmission between a server and a client so as to reduce communication overhead in personalized federated learning, and relates to the technical field of data compression. The frequency difference compressor comprises an information bottleneck rarefaction unit, a frequency domain compression unit, a dynamic quantization unit and a differential coding unit. Non-key gradient redundant components of the original gradient data are removed through an information bottleneck rarefaction unit to generate sparse gradient data; generating a metadata packet containing the first N high-energy frequency domain coefficients and the positions and the number of the first N high-energy frequency domain coefficients through a frequency domain compression unit; performing adaptive bit width mapping on the high-precision floating point gradient data into low-order integer representation through a dynamic quantization unit to generate dynamic quantization gradient data; and the differential gradient is transmitted to the client through the differential coding unit, so that data compression and transmission are completed. The method solves the problems that the communication overhead is large and gradient information cannot be reserved as far as possible in gradient transmission.
Owner:XIAMEN UNIV OF TECH

Mixed data precision matrix multiplication and addition unit and calculation method

The invention provides a mixed data precision matrix multiplication and addition unit and a calculation method, the matrix multiplication and addition unit comprises a calculation unit, and the calculation unit comprises a format division module, a multiplication array module, an addition tree module, an accumulator module, a normalization module and a shift register module. The calculation unit converts the first input matrix and the second input matrix into input data in a middle floating point format; executing parallel multiplication operation on the input data to generate an intermediate product result; performing index alignment and accumulation on the intermediate product result to generate an intermediate accumulated value; accumulating the intermediate product result and the value of the third input matrix in a form of accumulating an intermediate accumulated value, and outputting an accumulated result; and converting an accumulation result into a normalized result and outputting the normalized result. The format division module supports various precisions and converts data with different widths into an intermediate floating point format, so that other hardware units can be reused, and the problems that hardware resources are complex and different model reasoning scenes are difficult to meet are solved.
Owner:NANJING UNIV

Dynamic adjustment of floating point exponent bias for exponent compression

Approaches for compressing exponents of floating point values include accumulating a distribution of values of exponents of the first set of floating point values, and compressing the exponents of the first set of floating point values into a compressed exponent bit-width as a function of a compressed exponent bias. The compressed exponent bit-width and the compressed exponent bias are adjusted based on the distribution of values of exponents of the first set of floating point values. The distribution of values of exponents of the first set of floating point values is accumulated with values of exponents of a second set of floating point values that is input in subsequent time period. The exponents of second set of floating point values are compressed into the compressed exponent bit-width as a function of the compressed exponent bias after the adjusting of the compressed exponent bit-width and the compressed exponent bias.
Owner:XILINX INC

Energy-efficient datapath for vector-scaled hierarchical codebook quantization

Vector-scaled hierarchical codebook quantization reduces precision (bitwidth) vectors of parameters and may enable energy-efficient acceleration of deep neural networks. A vector (block array) comprises one or more parameters within a single dimension of a multi-dimensional tensor (or kernel). For example, block array comprises 4 sub-vectors (blocks) and each sub-vector comprises 8 parameters. The parameters may be represented in integer, floating-point, or any other suitable format. A vector cluster quantization technique is used to quantize blocks of parameters in real-time. Hardware circuitry within a datapath identifies an optimal codebook of a plurality of codebooks for quantizing each block of parameters and the block is encoded using the identified codebook. During processing, the identified codebook is used to obtain the quantized parameter and perform computations at the reduced precision.
Owner:NVIDIA CORP

Tensor core, processor, data processing method, electronic device and storage medium

The invention discloses a tensor core, a processor, a data processing method, electronic equipment and a storage medium, and is applied to the field of tensor processing. The tensor kernel comprises a first dot multiplication unit and a scaling factor matrix multiplication processing module, the scaling factor matrix multiplication processing module comprises a second dot multiplication unit, and the first dot multiplication unit and the second dot multiplication unit support dot multiplication operations of different floating-point number precisions. The tensor core is configured to receive a first tensor, a second tensor, a scaling factor of the first tensor, a scaling factor of the second tensor, an offset term of the first tensor, and an offset term of the second tensor, perform a matrix multiplication operation using the scaling factor and the offset term using a first dot multiplication unit and a scaling factor matrix multiplication processing module, and obtaining a matrix multiplication operation result of the first tensor and the second tensor. Matrix multiplication operation using scaling factors is executed by multiplexing dot multiplication units with different precisions in a tensor kernel, extra hardware area cost is reduced, and existing hardware resources are fully multiplexed.
Owner:SHANGHAI BIREN TECH CO LTD

Picture occlusion relation processing method and device and storage medium

The invention discloses a picture occlusion relation processing method and device and a storage medium, and relates to the technical field of rendering, and the method comprises the steps: determining a vertical distance between a target element in a to-be-rendered frame and a preset base point; performing normalization processing on the vertical distance to generate a floating point value corresponding to each target element; encoding the floating point value to generate at least one integer channel value corresponding to the element; storing the integer channel value corresponding to the target element to vertex color information of the grid data corresponding to the target element to generate a rendering grid; and rendering the target element based on the rendering grid. According to the method, the normalized and coded depth value is borne through the vertex color information, so that the technical problem that the rendering performance is low due to the fact that the node tree needs to be traversed in each frame and massive elements need to be sorted in the prior art is solved, seamless fusion of the depth data of the rendered elements and the vertex colors is realized, and the rendering efficiency of the elements is improved.
Owner:SHENZHEN ZIXIAO INTERACTIVE TECH CO LTD

Multi-dimensional collaborative AI training processor benchmark evaluation device and method

PendingCN120315982AHardware monitoringProcessor modelFloating point
The invention discloses a multi-dimensional collaborative AI training processor benchmark evaluation device and method. According to end-to-end performance evaluation, TTA and throughput indexes are obtained through a standardized training process; the operator support degree evaluation adopts a dynamic parameter generation technology to evaluate the speed-up ratio and the floating point performance utilization rate of calculation-intensive and memory-access-intensive operators; the energy consumption evaluation calculates a power consumption delay product based on steady-state power consumption monitoring. And finally, integrating multi-dimensional indexes through an analytic hierarchy process model and a fuzzy theory, realizing level mapping by adopting a four-section membership function, and finishing multi-level aggregation evaluation in combination with matrix operation. According to the method, three dimensions of performance, operator support degree and energy efficiency are creatively integrated, the problems of single evaluation index and high environment dependence of the heterogeneous AI processor are solved through automatic container deployment, dynamic parameter generation and steady-state power consumption monitoring technologies, and a comprehensive and quantifiable evaluation basis is provided for processor model selection.
Owner:XIDIAN UNIV

Nonlinear equalization method and device, equipment and storage medium

The invention discloses a nonlinear equalization method, device and equipment and a storage medium, and relates to the technical field of communication, and the method comprises the steps: obtaining a floating point tap coefficient of an original equalizer, and carrying out the low-order quantization of the floating point tap coefficient, and obtaining an integer tap coefficient; constructing a target equalizer based on the integer tap coefficient; and compensating nonlinear distortion of a to-be-processed signal based on the target equalizer to obtain a recovered signal. Through the mode, the low-order quantization is performed on the tap coefficient of the equalizer, and the traditional floating-point number is converted into the low-order integer for representation, so that the storage space and the calculation complexity are remarkably reduced, the hardware implementation efficiency can be optimized and the power consumption can be reduced while the transmission performance is ensured.
Owner:PENG CHENG LAB

Operation method, processor, and computing device

PCT designated stage expiredWO2024244476A1Digital data processing detailsComputer hardwareSign bit
Embodiments disclosed in the present application relate to the technical field of computing, and in particular to an operation method, a processor, and a computing device. The method is executed by a processor comprising a data transfer unit and a data operation unit, and comprises: the data transfer unit acquiring an activation vector and a quantized weight vector in an AI model, the weight vector being an integer, and the activation vector being a floating point number; the data transfer unit inputting to the data operation unit a sign bit and a mantissa bit of each element in the weight vector as a sign bit and a mantissa bit of the floating point number having an index of 0, and inputting each element in the activation vector to the data operation unit; the data operation unit performing a vector operation on the inputted weight vector and activation vector, to obtain an operation result. Using the present application, operations can be performed on an integer number and a floating point number in the data operation unit. Thus, when an AI model is quantitatively compressed, it is possible to only compress the weight data, and the impact of quantitative compression on the precision of the AI model can be reduced.
Owner:HUAWEI TECH CO LTD

GNSS integer ambiguity estimation method and device based on improved particle swarm optimization

The invention provides a GNSS (Global Navigation Satellite System) integer ambiguity estimation method and equipment based on improved particle swarm optimization, which are used for realizing rapid and accurate estimation of high-dimensional integer ambiguity and comprises the following steps of: performing integer Gaussian transformation parameterization on floating point ambiguity and a variance-covariance matrix thereof; taking a minimum optimal integer candidate vector residual quadratic form as fitness, and establishing an ambiguity search target function; performing direct rounding estimation on the floating point ambiguity to obtain a search initial value; determining a search space of each ambiguity component by combining the floating point ambiguity and the median error after integer Gaussian transformation; setting an improved particle swarm optimization method, and setting a dynamic inertia weight, a dynamic learning factor, a boundary detection and control model, an adaptive iteration model and a scale factor improvement model; and performing optimal and suboptimal integer ambiguity candidate vector estimation, and obtaining a corresponding residual quadratic form. The method can be used for ambiguity calculation of non-difference ambiguity, dual-frequency and ionosphere-free combination models, and can also be popularized to other GNSS integer ambiguity estimation problems.
Owner:WUHAN UNIV

Supporting 8-bit floating point format operands in computing architecture

An apparatus for facilitating support of 8-bit floating point format operands in a computing architecture is disclosed. In one embodiment, the apparatus includes a processor comprising: a decoder to decode an instruction fetched for execution into a decoded instruction, where the decoded instruction is a matrix instruction that operates on an 8-bit floating point operand to cause the processor to perform a parallel dot product operation; a controller to schedule the decoded instruction and provide input data for an 8-bit floating point operand according to an 8-bit floating point data format indicated by the decoded instruction; and a pulsation dot product circuit to execute the decoded instruction using pulsation layers, each pulsation layer comprising one or more sets of interconnected multipliers, shifters and adders, each set of multipliers, shifters and adders to generate a dot product of 8-bit floating point operands.
Owner:INTEL CORP

Instruction sequence generation method for verifying floating point operation of processor

The invention relates to the technical field of processor design and test, and discloses an instruction sequence generation method for verifying floating point operation of a processor, which comprises the following steps: constructing a floating point test coverage model, generating a floating point test database, and constructing a basic RISC-V instruction sequence generator; adding a register updating mechanism bound with a memory on the basis of the basic instruction RISC-V sequence generator, and generating initialized data of the memory; and injecting the generated instruction machine code and the memory initialization data into the memory of the processor to be verified for simulation, and comparing the execution track obtained by simulation with the execution track of the instruction set simulator to find potential errors in the processor. According to the method, an efficient coverage model is constructed to generate representative floating point test data, a floating point data path of the processor can be verified more comprehensively, the defects of an existing method in the aspect of floating points in the RISC-V instruction generation process are overcome, and potential design defects in the processor are found.
Owner:ZHEJIANG UNIV +1

Quantization Error Compensation for Vector Computing

A method for performing a computing task includes: extracting one or more features from a user content; converting the features to a floating point query vector; quantizing the floating point query vector; obtaining a database vector including one or more floating point feature vectors; determining a compensation vector based on a data distribution of the floating point query vector; quantizing the floating point feature vectors; determining an error function based on a difference between data distributions of i) the quantized query vector compensated with the compensation vector, and ii) the floating point query vector; determining, based on the error function, values of the compensation vector corresponding to the quantized feature vectors; combining the quantized query vectors and the values of the compensation vector to obtain one or more compensated query vectors; and performing the computing task using the compensated query vectors and the quantized feature vectors to obtain an output.
Owner:MACRONIX INTERNATIONAL CO LTD

Model Compression Method and Apparatus, and Related Device

A model compression method includes obtaining a first weight of each layer of a neural network model, where the first weight of each layer is a value of a floating-point type; and quantizing the first weight of each layer based on a quantization parameter to obtain a second weight of each layer. The second weight of each layer is a multi-bit integer. Quantities of bits of second weights of at least a part of layers are different. A quantity of quantization bits of a second weight of a layer with high sensitivity to a quantization error is greater than a quantity of quantization bits of a second weight of a layer with low sensitivity to the quantization error.
Owner:HUAWEI TECH CO LTD

V-DMC signalling improvements in displacement sub-bitstream for wavelet coefficient inter prediction with fixed-point inverse quantization

A device for decoding encoded dynamic mesh data determines a set of quantized integer coefficient values for displacement vectors of the encoded dynamic mesh data; determines a quantization parameter based on one or more syntax elements included in a displacement sub-bitstream of the encoded dynamic mesh data; inverse quantizes, based on the quantization parameter, the set of quantized integer coefficient values to determine a set of fixed-point dequantized coefficient values; determines a set of fixed-point transformed coefficient values based on the set of fixed-point dequantized coefficient values; converts the set of fixed-point transformed coefficient values to a set of floating-point transformed coefficient values; inverse transforms the set of floating-point transformed coefficient values to determine a set of reconstructed displacement vectors; and determines a reconstructed deformed mesh based on the set of reconstructed displacement vectors.
Owner:QUALCOMM INC

Data processing method based on large language model, and large language model and electronic device

Disclosed in the present application are a data processing method based on a large language model, and a large language model, an electronic device, a computer-readable storage medium and a computer program product. The method is applied to a user terminal, wherein a large language model is deployed on the user terminal, and weight parameters of linear calculation layers of the large language model are pre-quantized into format data of an integer data type. The method comprises: acquiring input data; performing vector conversion on the input data by means of an embedding layer of a large language model, so as to obtain a floating-point query vector of a floating-point data type corresponding to the input data; converting the floating-point query vector into an integer query vector of an integer data type; and performing an operation by means of weight parameters of linear calculation layers and the integer query vector, so as to obtain a query result corresponding to the input data. By means of the solution provided in the present application, a large language model can be smoothly run on a user terminal, such that the user terminal can provide services for users without needing network connectivity, and can better ensure the privacy of the users.
Owner:TAOBAO CHINA SOFTWARE

Quantization for neural networks

Disclosed herein are systems and methods for performing post training quantization. A processor obtains fixed-point output values from a layer of an artificial neural network (ANN) wherein the layer includes fixed-point weights determined based on floating-point weights and a weight scaling factor determined based on an output scaling factor. Next, the processor converts the fixed-point output values to floating-point output values based on the output scaling factor. Then, the processor expands a range of floating-point values. Next, the processor calculates a new output scaling factor based on the expanded range of floating-point output values. Finally, the processor stores the new output scaling factor in an associated memory.
Owner:TEXAS INSTRUMENTS INC

Floating point addition implementation method and device for mixed precision operation and electronic equipment

The invention discloses a floating point addition implementation method and device for mixed precision operation and electronic equipment. The method comprises the steps of performing mild preprocessing on two low-precision operands of floating point addition operation; if the improved near-path condition is met, leading zero prediction and mantissa addition operation are carried out on the two operands in parallel, a near-path operation result is obtained, and the number of leading zeros comprises the number of leading zeros subjected to floating-point number addition mantissa high-order offset and the number of non-normalized leading zeros; if the improved near-path condition is not met, performing far-path order shifting and mantissa addition operation on the two operands to obtain a far-path operation result; and selecting one as an operation result to carry out mantissa rounding and normalized shifting to obtain a high-precision floating point addition operation result. Through the above mode, according to the embodiment of the invention, normalized calculation with improved precision and existing near-path calculation can be combined into one, floating point additive operation resource reuse and delay elimination are realized, and the execution efficiency is improved.
Owner:CHAORUI TECH (SHANGHAI) CO LTD

BLAS3 structured operator accelerated computing system based on Hopper architecture GPU

The invention provides a BLAS3 structured operator accelerated computing system based on a Hopper architecture GPU, and relates to the technical field of computers. The system comprises: a calculation unit discrimination module for determining a calculation unit used by a current operator during operation, and estimating the maximum row dimension upper bound of the current operator in a tensor core execution path; an instruction sensing block parameter determination module dynamically determines the optimal block size and number of the input matrix in real time; the block matrix loading and aligning module divides an input matrix and a matrix to be updated into sub-matrixes by taking the block size as a basic block and completes loading of the corresponding sub-matrixes; the operator kernel function execution module completes shared memory structured parallel loading and storage of a double-precision floating-point number array of a sub-matrix corresponding to the input matrix, and calls a tensor core to carry out multiply-add accumulation calculation; and the assembly line and concurrent scheduling module adds the block calculation tasks into corresponding task sets and performs multi-stream concurrent scheduling on the task sets.
Owner:NORTHEASTERN UNIV CHINA

Methods and systems for quantization of large language models

A method and an apparatus for storing data points are provided. The method comprises: receiving the plurality of data points, each data point of the plurality of data points being represented in a floating-point representation; quantizing each one of the first plurality of data points, by: executing, during a first quantization phase: converting each data point of the plurality of data points into a corresponding first data point of a plurality of first data points; executing, during a second quantization phase: applying, to each first data point of the plurality of first data points a clamping function, thereby converting each first data point of the plurality of first data points into a corresponding second data point of a plurality of second data points; and storing the plurality of second points for further calculations instead of the first plurality of data points.
Owner:HUAWEI TECH CO LTD

Low-bit-width high-energy-efficiency floating point storage and calculation integrated circuit based on partial pre-alignment architecture

The invention belongs to the technical field of storage and calculation integration, and particularly relates to a low-bit-width and high-energy-efficiency floating point storage and calculation integrated circuit based on a partial pre-alignment framework. The circuit comprises a memory array, a pre-calculation unit, an adder tree, a configurable arithmetic unit and a normalization unit, and supports mixed precision operation of FP8MACFP4 and FP8MACFP8. The method is characterized in that a partial pre-alignment strategy dominated by an activation value is adopted, the maximum index of the activation value is dynamically counted, the mantissa of the maximum index is aligned, and multiple partial pre-alignment intermediate results are pre-calculated and latched for reuse; in combination with a customized lookup table and a multiplexer, a pre-calculation result is directly selected to replace real-time multiplication and displacement; and through the reconfigurable hardware, the FP8MACFP8 high-precision operation is realized by utilizing the FP8MACFP4 unit combination. According to the method, complete online floating point multiplication and addition operation is realized, and excellent energy efficiency ratio and operation speed are obtained while high precision is kept.
Owner:FUDAN UNIVERSITY

Ray-Box Intersection Circuitry

Techniques are disclosed relating to ray intersection tests for ray tracing in graphics processors. In some embodiments, a ray cache stores, transformed ray direction corresponding to a transform of a ray based on its dominant direction axis. Ray intersect acceleration circuitry may determine whether a ray intersects a bounding volume of a bounding volume hierarchy (BVH) data structure. Ray-plane test circuitry may perform a set of six ray-plane tests for the bounding volume using at most four floating-point multiplication operations by four multiplier circuits. The floating-point operations may operate on the transformed ray direction components from the ray cache circuitry and coordinate information for the bounding volume. Disclosed techniques may reduce area and power consumption by avoiding multiplier circuitry for some ray-plane tests.
Owner:APPLE INC

Data processing method and device, computer equipment and storage medium

The invention relates to a data processing method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring input original data and a corresponding data type; under the condition that the data type is a floating point format, obtaining an index and a mantissa of the original data; according to the index, determining an integer significant digit of a fixed-point format corresponding to the original data; according to the index and the mantissa, determining mantissa significant digits of a fixed-point format corresponding to the original data; determining a fixed-point format structure corresponding to the original data according to the integer significant bit and the mantissa significant bit of the fixed-point format; converting the original data into fixed-point format data, and storing the fixed-point format data into a fixed-point format structure to obtain converted data; and under the condition that the data type of the original data is a fixed-point format or the converted data is obtained, inputting the original data and / or the converted data into the normalized exponential function to obtain target data. By adopting the method, the data processing efficiency can be improved.
Owner:CCORE TECH CO LTD

Accelerator for operations between floating point matrix and integer matrix and operation method thereof

An operation accelerator that performs an operation between a floating point matrix and an integer matrix includes a first buffer storing integer matrix data; a second buffer storing floating point matrix data; a data converter to convert the floating point matrix data into an integer; and an operator to perform multiplication on the integer matrix data and integer operation target matrix data output from the data converter, wherein the data converter includes a pre-aligner to find a maximum exponent value among multiple floating point values included in the floating point matrix data, perform pre-alignment for moving a mantissa of each of floating points by a difference between the maximum exponent value and an exponent value of each of the multiple floating point values, and generate the integer operation target matrix data based on mantissas of a preset number of high-order bits extracted from among mantissas of pre-aligned floating point values.
Owner:SEOUL NATIONAL UNIVERSITY R&DB FOUNDATION

Method and device for measuring computing power of intelligent computing center by 1 degree computing power

The present invention provides a method and device for measuring the computing power of an intelligent computing center through 1 degree computing power, which relates to the field of computing power technology. The method comprises: step S1, obtaining the initial half-precision floating-point computing power number, initial video memory bandwidth and initial video memory capacity of a processor of a first intelligent computing center; step S2, calculating the first ratio of the initial half-precision floating-point computing power number to the preset half-precision floating-point computing power number, calculating the second ratio of the initial video memory bandwidth to the preset video memory bandwidth, and calculating the third ratio of the initial video memory capacity to the preset video memory capacity; step S3, setting the weighted sum of the first ratio, the second ratio and the third ratio, the preset floating-point operation number unit and the preset time unit as the initial 1 degree computing power value of the first intelligent computing center. The present invention can accurately measure the computing power of the intelligent computing center through the 1 degree computing power value.
Owner:DATACANVAS LTD

Multi-target neural architecture search method and system based on gradient similarity hypernetwork

The invention provides a multi-target neural architecture search method and system based on a gradient similarity hypernet, and the method comprises the steps: constructing a hypernet, and generating a shared weight through training; in the training process, a gradient norm importance sampling strategy is adopted, the sampling probability of paths and data is dynamically optimized, and architecture and data which have great influence on weight updating are preferentially selected, so that the efficiency is improved. For gradient conflicts in multi-objective optimization, a PCGrad projection method is introduced, gradient similarity dynamic dominant direction selection and gradient projection are calculated, conflicts are reduced, and the optimization process is stabilized. After super-network training, population inheritance sharing weight is initialized, performance is obtained on a target data set, non-dominated sorting and optimization are carried out through an evolutionary algorithm, and it is ensured that balanced performance is achieved on the targets of precision, floating point calculation times, parameter quantity and the like. According to the method, on the premise of ensuring multi-target balance optimization, the calculation cost is effectively reduced, and the efficiency and the performance of neural network architecture search are improved.
Owner:NANJING UNIV OF INFORMATION SCI & TECH

Floating point multiply-accumulate unit facilitating variable data precision

A fused dot-product multiply-accumulate (MAC) circuit may support variable precision of floating-point data elements to perform computations in deep learning operations (e.g., MAC operations). The operating mode of the circuit may be selected based on the accuracy of the input element. The mode of operation may be an FP16 mode or an FP8 mode. In the FP8 mode, a product index may be calculated based on an index of a floating point input element. A maximum index may be selected from the one or more product indexes. A global maximum index may be selected from a plurality of maximum indexes. A product mantissa may be calculated based on a difference between the global maximum exponent and a corresponding maximum exponent and aligned with another product mantissa. The adder tree may accumulate the aligned product mantissas and compute the partial and mantissas. The portions and mantissas may be normalized using a global maximum index.
Owner:INTEL CORP

Multiply-accumulate operation method and device for extremely low precision training

The invention provides a multiply-accumulate operation method and device for extremely low precision training, and the method comprises the steps: obtaining an operated part and to-be-operated data, and the to-be-operated data comprises one or more of floating point, fixed point and logarithm data formats; if the high-precision floating-point number exists in the data to be calculated, decomposing the high-precision floating-point number into the sum of two low-precision floating-point numbers; floating-point, fixed-point and logarithm data formats in the data to be calculated are decoded to be uniform formats of symbols, indexes and mantissas, and input data in the uniform formats are generated; performing multi-term dot product on the input data with the unified format to realize combined calculation of data formats with different precision; and adding the combined calculation result and the calculated part sum by adopting an advanced random rounding method. The method can support the combined calculation among various formats, reduces the bit width of the accumulator on the premise of maintaining the precision requirement by using an advanced random rounding method, and reduces the operation cost and improves the training efficiency under extremely low precision compared with the existing operation hardware.
Owner:TSINGHUA UNIVERSITY