Unlock instant, AI-driven research and patent intelligence for your innovation.

Computing method, device and related products

A computing device and computing method technology, applied in the field of information processing, can solve the problems of complex and changeable programming languages ​​of neural network algorithms and high complexity of neural network algorithms

Active Publication Date: 2022-03-11
SHANGHAI CAMBRICON INFORMATION TECH CO LTD
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the complexity of the neural network algorithm itself is high, and an algorithm may need to be split and executed on a system on a chip.
The programming languages ​​used to implement neural network algorithms are also complex and changeable
When the neural network algorithms implemented in different programming languages ​​are executed on different types of SoCs, it is necessary to set specific interfaces for specific language algorithms and specific SoCs, making the neural network algorithms highly complex when implemented in SoCs

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Computing method, device and related products
  • Computing method, device and related products
  • Computing method, device and related products

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0096] In a possible implementation manner, the data of the convolution algorithm is segmented according to the segmentation information to obtain segment data of the data, and the data includes input data and output data of the convolution algorithm;

[0097] determining a first intermediate representation of said data based on segmented data of said data;

[0098] determining the storage space corresponding to the first intermediate representation of the data in the on-chip cache;

[0099] generating a second intermediate representation of the data according to the storage space;

[0100] A second intermediate representation of the convolution algorithm is generated based on the second intermediate representation of the data and the second intermediate representation of the operator of the convolution algorithm.

[0101] In a possible implementation manner, the input data of the convolution algorithm includes input neurons, weights, and biases, and the output data of the co...

Embodiment 2

[0121] In a possible implementation manner, the data of the convolution algorithm is segmented according to the segmentation information to obtain segment data of the data, and the data includes input data and output data of the convolution algorithm;

[0122] determining a first intermediate representation of said data based on segmented data of said data;

[0123] A first intermediate representation of the convolution algorithm is generated according to the first intermediate representation of the data and the first intermediate representation of the operator of the convolution algorithm.

[0124] In a possible implementation manner, the data is multi-dimensional data, and the data of the convolution algorithm is segmented according to the segmentation information to obtain the segmented data of the data, including:

[0125] Segmenting each dimension data of the data according to the segmentation information to obtain dimension segmentation data of each dimension data of the...

Embodiment 3

[0131] In a possible implementation manner, the data of the pooling algorithm may be segmented according to the segmentation information to obtain segmented data of the data, where the data includes input data and output data of the pooling algorithm;

[0132] determining a first intermediate representation of said data based on segmented data of said data;

[0133] determining the storage space corresponding to the first intermediate representation of the data in the on-chip cache;

[0134] generating a second intermediate representation of the data according to the storage space;

[0135] A second intermediate expression of the pooling algorithm is generated according to the second intermediate expression of the data and the second intermediate expression of the operator of the pooling algorithm.

[0136] In a possible implementation manner, the data of the pooling algorithm is multi-dimensional data, and the data of the pooling algorithm is segmented according to the segme...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The present disclosure relates to an operation method, device and related products. The method includes: segmenting the data of the convolution algorithm according to the segmentation information to obtain the segmented data of the data, the data including the input data of the convolution algorithm and output data; determine the first intermediate representation of the data according to the segmentation data of the data; generate the convolution according to the first intermediate representation of the data and the first intermediate representation of the convolution algorithm operator The first intermediate expression of the algorithm. The embodiment of the present disclosure obtains the first intermediate expression of the algorithm through the first intermediate expression of the data and the first intermediate expression of the operator. When different algorithms are implemented in different on-chip systems, no specific interface needs to be set, which improves the algorithm and on-chip Compatibility between systems reduces the difficulty of algorithm development.

Description

technical field [0001] The present disclosure relates to the technical field of information processing, and in particular to an operation method, device and related products. Background technique [0002] Different SoCs can be configured with different language expressions, that is, different codes that can be executed on the SoC can be configured. However, the complexity of the neural network algorithm itself is high, and an algorithm may need to be split and executed on a system-on-chip. The programming languages ​​used to implement neural network algorithms are also complex and varied. When neural network algorithms implemented in different programming languages ​​are executed on different types of SoCs, it is necessary to set specific interfaces for specific language algorithms and specific SoCs, making the neural network algorithms highly complex when implemented in SoCs . Contents of the invention [0003] In view of this, the present disclosure proposes a computi...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06F17/15G06F9/30G06N3/04G06N3/06
CPCG06F9/30007G06F17/153G06N3/061G06N3/045
Inventor 不公告发明人
Owner SHANGHAI CAMBRICON INFORMATION TECH CO LTD