Machine vision series (02) -- tensorflow2 3 + win10 + GPU installation

In short video technology , Visual effects are currently the most widely used , Visual effects belong to the combination of machine vision and artificial intelligence technology .

 

Short video technology development , Computer vision is a factor that must be considered , Computer vision AI Handle , At present, China mainly depends on TensorFlow and Pytorch To achieve .

 

Tensorflow windows The installation of the , A lot of information is misleading , Beginners stop at a glance . This article helps beginners quickly build TensorFlow windows Application environment .Linux and Mac See the following chapters for the installation of the upper environment .

 

Most installation materials are installed first anaconda, cause TensorFlow Installation must be done with anaconda The illusion of . This installation does not use anaconda, Reduce the coupling of the installation .

 

A large number of subsequent chapters describe machine vision and related algorithms .

1、 Preposition

A、 If it is Win7, Be sure to patch ( this paper Win7 Also applicable ).

B、 See if your computer graphics card is N card , If it is N card , To install GPU edition , Otherwise, you can only install CPU edition .

C、 If GPU Version is too low , Need to upgrade driver , Otherwise, the installation version is too early ( Upgrading is described in the following topics ).

 

D、 Some are too old N card , Unable to upgrade driver , Can't install TensorFlow  GPU Version or can only be installed TensorFlow GPU Early versions .

2、 Upgrade the graphics card driver , Also upgraded CUDA

2.1、 website

Go to the website below ：

https://www.nvidia.com/download/index.aspx?lang=en-us

2.2、 Interface

Choose according to the personal computer system .

2.3、 download

Choose the right version , Download the graphics card driver .

2.4、 install

Just go to the next step . If the graphics card is too early , Installation may fail .

3、 View personal computers CUDA edition

3.1、 open NVIDIA Control panel

Right click on the desktop , Right click menu , choice NVIDIA Control panel .

3.2、 Click on the component , see CUDA edition

4、 View the following correspondence

4.1、 Refer to the website ：

https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

https://tensorflow.google.cn/install/source_windows

4.2、CUDA Version and name

Component Name	Version Information	Supported Architectures
CUDA Runtime (cudart)	11.2.72	x86_64, POWER, Arm64
Cuobjdump	11.2.67	x86_64, POWER, Arm64
CUPTI	11.2.67	x86_64, POWER, Arm64
CUDA cuxxfilt (demangler)	11.2.67	x86_64, POWER, Arm64
CUDA Demo Suite	11.2.67	x86_64
CUDA GDB	11.2.67	x86_64, POWER, Arm64
CUDA Memcheck	11.2.67	x86_64, POWER
CUDA NVCC	11.2.67	x86_64, POWER, Arm64
CUDA nvdisasm	11.2.67	x86_64, POWER, Arm64
CUDA NVML Headers	11.2.67	x86_64, POWER, Arm64
CUDA nvprof	11.2.67	x86_64, POWER, Arm64
CUDA nvprune	11.2.67	x86_64, POWER, Arm64
CUDA NVRTC	11.2.67	x86_64, POWER, Arm64
CUDA NVTX	11.2.67	x86_64, POWER, Arm64
CUDA NVVP	11.2.67	x86_64, POWER
CUDA Samples	11.2.67	x86_64, POWER, Arm64
CUDA Compute Sanitizer API	11.2.67	x86_64, POWER, Arm64
CUDA cuBLAS	11.3.1.68	x86_64, POWER, Arm64
CUDA cuFFT	10.4.0.72	x86_64, POWER, Arm64
CUDA cuRAND	10.2.3.68	x86_64, POWER, Arm64
CUDA cuSOLVER	11.0.2.68	x86_64, POWER, Arm64
CUDA cuSPARSE	11.3.1.68	x86_64, POWER, Arm64
CUDA NPP	11.2.1.68	x86_64, POWER, Arm64
CUDA nvJPEG	11.3.1.68	x86_64, POWER, Arm64
Nsight Eclipse Plugins	11.2.67	x86_64, POWER
Nsight Compute	2020.3.0.18	x86_64, POWER, Arm64
Nsight Windows NVTX	1.21018621	x86_64, POWER, Arm64
Nsight Systems	2020.4.3.7	x86_64, POWER, Arm64
Nsight Visual Studio Edition (VSE)	2020.3.0.20315	x86_64 (Windows)
NVIDIA Linux Driver	460.27.04	x86_64, POWER, Arm64
NVIDIA Windows Driver	460.89	x86_64 (Windows)

 

 

 

 

 

4.3、 Graphics driver and CUDA Version selection

4.4、TensorFlow GPU Version and CUDA And Python And VS The relationship between

4.5、TensorFlow CPU Version and VS And Python The relationship between

5、 According to the above correspondence , Download the following file

5.1、 file

A、VS     （ Microsoft visual studio development environment ）

B、Python  （TensorFlow Official development language ）

C、CUDA   （ NVIDIA data parallel computing platform ）

D、CUDNN （ Neural network acceleration library ）

5.2、 After downloading, the figure shows

6、 install VS

6.1、 explain

             It is recommended that all tools be installed , need 40 Multiple G（ Personal test ）.

             NVIDIA says it just needs to install C++ Desktop Develop Kit installation （ No test ）.

6.2、 install

             All the way to the next step .

7、 install Python

7.1、 Environmental variables

 

7.2、 Be careful （ During installation , You may have to wait ）

above 5 Put a box on it “√”, And then click “Browse”, Custom installation path , And then click “Install”.

 

PS： The installation path should be in English , Don't have Chinese and spaces .

8、 modify Python Tools Pip Source

             The installation process needs to be downloaded from the Internet , default Python Source abroad , Slow download .

 

Modified to domestic Ali's Python Source , Speed up the installation of the following steps .

8.1、PyPI  Mirror image

PyPI (Python Package Index)  yes  Python  Software repository for programming languages . Developers can use  PyPI  Find and install by  Python  Software developed and shared by the community , You can also upload your own library to  PyPI .

8.2、Linux Next change pip Source （ It has nothing to do with this installation 、 Just skip ）

Manual replacement python Of pip/pip3 Source .

 

1. stay Linux Found in the system ~/.pip/pip.conf This file , If you don't have this file, create a file with the same name . If the system is connected ~./pip There is no such thing in this directory , Create a directory and create a pip.conf file .

 

1. Add or modify the following to the file .

 

[global]

index-url = https://mirrors.aliyun.com/pypi/simple/

 

[install]

trusted-host=mirrors.aliyun.com

8.3、windows Next change pip Source

A、 Enter... In the address field of the file manager %appdata% Press enter to enter the current user's home directory .

B、 Create a new folder under the current folder pip.

C、 stay pip Create a new file in the folder pip.ini file .

D、 stay pip.ini Enter the following in the file .

 

[global]

index-url = https://mirrors.aliyun.com/pypi/simple/

 

[install]

trusted-host=mirrors.aliyun.com

9、 Install two plug-ins

9.1、 If during the following installation , Installation failed

If installation fails , Re executing the command can .

9.2、scipy library

stay power shell Under the window , Copy the following command to execute ：

 

pip3 install scipy

9.3、numpy library

stay pwer shell Under the window , Copy the following command to execute ：

 

pip3 install numpy

10、 install cuda

10.1、 install

Double click the... Downloaded above cuda file .

 

Just go to the next step , You can change the installation path , But it is not recommended to modify .

10.2、 test

open powershell, Execute the following command ：

 

nvcc -V

 

Success will return to cuda Version number .

11、 install cudnn

11.1、 decompression

decompression cuDNN Compressed package , You can see bin、include、lib Catalog ：

11.2、 Open Directory

Path is ：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1

11.3、 Copy

Find the version directory you installed , open , find bin、include、lib Catalog ;

 

take 11.1 cuDNN The corresponding files in the compressed package are copied to cuda Of bin、include、lib Catalog .

 

Be careful ： Is to copy files to bin、include、lib Catalog , Not a copy directory .

12、 environment variable

12.1、 Several variables

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\libnvvp

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\lib\x64

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin

12.2、 Set up （ Less , Will be submitted to the dll abnormal ）

13、 test CUDA and CUDNN Is the installation successful

13.1、 Open Directory

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\extras\demo_suite

13.2、 start-up PowerShell

13.3、 test

Execute the following command ：

 

.\bandWidthTest.exe

appear Pass On behalf of success .

14、 install tensorflow-gpu

stay PowerShell Window input the following command ：

 

pip3 install tensorflow-gpu

15、 test

15.1、 open Python Environmental Science

stay PowerShell Enter the following command ：

 

python

15.2、 Run the following code

import tensorflow as tf

a = tf.constant(2.0)

b = tf.constant(4.0)

print(a + b)

 

The following results represent success .

16、 Other

             And the cloud platform 、 Big data related environment construction ,TensorFlow The installation of is simple .

 

But for beginners , It will take some time . The content of this article is , Hope to give beginners some useful enlightenment .

 

             Pay attention to American Photography Technology , Feel the visual feast .

 

 

17、 Learning website

 

https://tensorflow.google.cn/install/source_windows