Processing

Please wait...

Settings

Settings

Goto Application

1. CN112101515 - Acceleration method and device of deep learning model

Office
China
Application Number 202010745500.3
Application Date 29.07.2020
Publication Number 112101515
Publication Date 18.12.2020
Publication Kind A
IPC
G06N 3/04
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
G06N 3/063
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
06Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
063using electronic means
G06N 3/08
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
CPC
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06N 3/063
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
06Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
063using electronic means
G06N 3/08
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
Applicants BEIJING IDRIVERPLUS TECHNOLOGY CO., LTD.
北京智行者科技有限公司
Inventors FU JIAWEI
付家为
CHEN DONG
陈东
ZHANG FANG
张放
LI XIAOFEI
李晓飞
ZHANG DEZHAO
张德兆
WANG XIAO
王肖
HUO SHUHAO
霍舒豪
Agents 北京慧诚智道知识产权代理事务所(特殊普通合伙) 11539
Title
(EN) Acceleration method and device of deep learning model
(ZH) 深度学习模型的加速方法及装置
Abstract
(EN) The invention provides an acceleration method of a deep learning model. The acceleration method comprises the steps of obtaining a contribution value of each channel in a plurality of channels in eachconvolution layer in the model; according to the contribution values of the channels in all the convolution layers, cutting the channels in the convolution layers in the model, and obtaining a cut model; respectively training the model and the cut model; evaluating the trained model and the cut trained model to obtain a first evaluation value and a second evaluation value; and according to the first evaluation value and the second evaluation value, determining whether to output the cut and trained model as a new model. Therefore, under the condition that the reasoning precision is not lost orthe loss is very small, the reasoning speed is greatly increased.
(ZH) 本发明提供了一种深度学习模型的加速方法,包括:获取模型中的每个卷积层中的多个通道中每个通道的贡献值;根据全部卷积层中各通道的贡献值,对模型中的卷积层中的通道进行裁剪,得到裁剪后的模型;对模型和裁剪后的模型分别进行训练;对训练后的模型和裁剪后训练的模型分别进行评估,得到第一评估值和第二评估值;根据第一评估值和第二评估值,确定是否将裁剪后训练的模型作为新模型输出。由此,在推理精度不损失或者损失非常小的情况下,实现推理速度的大大提升。
Related patent documents