Environmental piecemeal take three or four days, although encountered various problems, but still a good build, organize their own record at the same time be considered to give people in need of some guidance bar
One. Install dual system
- Win10 Disk Management Assistant compressed or the direct use of unused space, custom space, will free up disk space to delete the volume becomes unallocated state.
2. Select ultraISO make Ubuntu system disk, namely by means ultraIOS will burn the ISO system image to the U disk, U disk to back up important files to achieve.
3. When finished system disk, install the system, restart the computer, press F12 to enter the BIOS, choose U disk, it will enter the installation interface Note: When installing the system into English language options, try not to choose Chinese as the latter might work An unknown error occurred inevitable
4. Note that the main mounting step two systems by selecting another option is selected when installing the installation path, custom partitioning
/ / Home / boot and divided into swap area details Baidu
Recommends that the final distribution of the primary partition /
two. Before you begin the installation environment
- Install linux system based on NVIDIA graphics drivers, NVIDIA Baidu search drive or direct access to the head loss Sheriff network, download the latest drivers, pay attention to choose the drive system is installed linux x86 64 bit general about 384.111 version can be, not too low version
- The right mouse button to open the plug-in terminal, it will improve efficiency
sudo apt-get install nautilus-open-terminal
Can execute the command
Note: you have just installed Ubuntu system default resolution will be low, it is recommended you download the win10 linux system NVIDIA drivers, mounted directly on Ubuntu
Remember to switch the graphics driver is installed using the NVIDIA
sudo service lightdm stop
After the black press ctrl + alt + F1 enters mode tty1
Remember not familiar with linux environment Note drive storage address, enter the tty1 mode, access the folder into the drive, install the driver
.run file details installation instructions Baidu
After the installation is completed sudo service lightdm start
Change in the resolution or the terminal input nvidia-smi, graphics information representative of successful installation appears
three. Cuda8.0 installation
Also recommend downloading under win10 good cuda8.0 version, installed directly into the terminal, the method is similar to driver installation.
Select Configure cuda installation refer to other Installation Guide install (previously issued documents), select the default installation path path, asking whether to install NVIDIA drivers select No, because already have them.
After installation configuration variables, see Baidu hair before or install
four. Cudnn5.1 installation
The cudnn decompression, covering run command
tar xvzf cudnn-7.5-linux-x64-v4.tgz # to note here is that you download the version you need to unzip the downloaded file corresponding version
After extracting the file folder name is cuda #
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
Cuda test instructions
cd /usr/local/cuda-8.0/samples/1_Utilities/deviceQuery
make
./deviceQuery
Graphics information after completion, and finally show that is successful PASS
Note: make may appear insufficient permissions problems that may be due to excessive lead terminal open, click the leftmost terminal display terminal all running, close all terminal window, re-run.
If an invalid attempt to modify user permissions
Fives. TensorFlow installation
TensorFlow visit the official website
Installation Python3.4
Baidu detailed way, after the installation is complete Python3.4 set as the default version, must not and can not uninstall the previous version of Python, otherwise it will be a serious bug
Note: Please note that different versions TensorFlow corresponding python version when you download should correspond with each other, to avoid not run.
Install python3.4
Sudo add-apt-repository ppa:fkrull/deadsnakes
Sudo apt-get update
Sudo aptget –y install python3.4
Python3.4 set as the default version, comes with more general version of ubuntu python
Sudo update-alternatives –install /usr/bin/python python /usr/bin/python2 100
Sudo update-alternatives –install /usr/bin/python python /usr/bin/python3 150
Input terminal can view the default python version
TensorFlow installation
According to the official website in English are sequentially arranged installation guide, install using pip
Configuration pip
Sudo apt-get install python3-pip python3-dev
Installation tensorflow
Sudo pip3 install --upgrade https://storage.googleapis.com/tensorflow/linux/gup/tensorflow_gpu-1.2.1-cp34-cp34m-linux_x86_64.whl
Complete the installation execution (six) were tested python
Note: libcusolver.so.8.0:: can not open shared object file: No such file or directory in error, indicating that the path is not set correctly ImportError appear if the test
Set as follows:
Sudo vim ~/.bashrc
Add at the end of file
export LD_LIBRARY_PATH=”$LD_LIBRARY_PATH:/usr/local/cuda/lib64”
export CUDA_HOME=/usr/local/cuda
Execution is completed
source ~ / .bashrc
Details Reference http://blog.csdn.net/appleml/article/details/76922476
six. TensorFlow test
The command input terminal
python
>>> import tensorflow as tf
>>> hello = tf.constant('Hello,Tensorflow')
>>> sess=tf.Session()
>>> print(sess.run(hello))
If you can show Hello, Tensor, shows no problem!
Try a
>>> a=tf.constant(10)
>>> b=tf.constant(22)
>>> print sess.run(a+b)
Output should be 32
reference:
[1] Ubuntu 16.04 Win10 dual system with dual-drive installation diagram
https://blog.csdn.net/fesdgasdgasdg/article/details/54183577
[2] windows10 install ubuntu dual system tutorial (absolute history of the most detailed)
https://www.cnblogs.com/masbay/p/10745170.html
[3] Win10 Ubuntu16.04 / Ubuntu18.04 perfect dual system installation
https://blog.csdn.net/qq_24624539/article/details/81775635
[4] Win10 + Ubuntu dual system repair system boot Ubuntu
https://blog.csdn.net/chekongfu/article/details/84872841
[5] Ubuntu 16.04 Nvidia driver installation (run mode)
https://blog.csdn.net/lihe4151021/article/details/90083431
[6] Ubuntu 16.04 installation CUDA8.0
https://www.linuxidc.com/Linux/2018-09/154292.htm
[7] Ubuntu16.04 mounted cuda and cudnn of three methods (all effective pro-test)
https://blog.csdn.net/wanzhen4330/article/details/81699769
[8] mounted on Ubuntu16.04 cudnn5.1 detailed tutorial
https://blog.csdn.net/QLULIBIN/article/details/80729570
[9] Ubuntu16.04 mounted tensorflow (GPU acceleration)