Ubuntu install TensorFlow 1.10 + CUDA 9.2 + cuDNN 7.2

 

為了裝TensorFlow 1.10 下面升級一下系統的軟件環境

NVIDIA DRIVER

去官網下載最新的linux驅動    http://www.nvidia.com/Download/index.aspx 

 

直接運行會報錯

 

sudo bash NVIDIA-Linux-x86_64-390.87.run

 

ERROR: You appear to be running an X server; please exit X before
         installing. For further details, please see the section INSTALLING
         THE NVIDIA DRIVER in the README available on the Linux driver
         download page at www.nvidia.com.

需要先關閉圖形界面，在另一台電腦上用ssh登錄這台電腦然后運行

sudo init 3 
sudo killall Xorg

然后再運行 

sudo bash NVIDIA-Linux-x86_64-390.87.run

裝好后運行 

nvidia-smi

出現下圖結果說明成功安裝

再運行下面命令恢復圖形界面

sudo init 5

可以重啟一下確認顯卡驅動是否正常

 

如果需要改gcc 或g++版本 請參考上一篇博文 https://www.cnblogs.com/jins-note/p/9597210.html

 

 

CUDA 9.2

由於TensorFlow 1.10 支持cuda 9.2 

去官網下載最新版本 

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1710&target_type=runfilelocal

先安裝一些推薦庫

sudo apt-get install freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa libglu1-mesa-dev libglfw3-dev libgles2-mesa-dev

 

這里注意：cuda里帶的驅動比剛從官網下的新，那么用cuda里帶的驅動（居然比官網下的顯卡驅動新？？）

不然會報錯 

cudaerrorinsufficientdriver

然后安裝

sudo init 3
sudo killall Xorg
sudo bash cuda_9.2.148_396.37_linux.run

安裝過程如下

 

Description

The NVIDIA CUDA Toolkit provides command-line and graphical
Do you accept the previously read EULA?
accept/decline/quit: accept

Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 396.37?
(y)es/(n)o/(q)uit: y

Install the CUDA 9.2 Toolkit?
(y)es/(n)o/(q)uit: y

Enter Toolkit Location
 [ default is /usr/local/cuda-9.2 ]:

Do you want to install a symbolic link at /usr/local/cuda?
(y)es/(n)o/(q)uit: y

Install the CUDA 9.2 Samples?
(y)es/(n)o/(q)uit: y

Enter CUDA Samples Location
 [ default is /home/whatever ]:

安裝 補丁

sudo bash cuda_9.2.148.1_linux.run

裝好后修改環境變量 ~/.bashrc 在末尾添加

export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64"
export CUDA_HOME=/usr/local/cuda
export PATH="$CUDA_HOME/bin:$PATH"

修改完畢之后執行一下使其生效：

source ~/.bashrc

恢復顯示

sudo init 5

 裝好后去 CUDA Samples  目錄編譯一些例子看看能不能運行，能運行就ok  

cd ~/NVIDIA_CUDA-9.2_Samples/
make -j8

編譯好后去下面目錄里運行

cd bin/x86_64/linux/release

 

cuDNN

去官網下載對應版本  https://developer.nvidia.com/rdp/cudnn-download 需要登錄才能下載

 

下載后的文件后綴名應該是 *.tgz 如果 是 .solitairetheme8 那就改成 .tgz

安裝很簡單，解壓復制到對應目錄就好

tar -zxvf cudnn-9.2-linux-x64-v7.2.1.38.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/ -d
sudo chmod a+r /usr/local/cuda/include/cudnn.h
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

 

TensorFlow 1.10

安裝 Anaconda   從這里下載 https://www.anaconda.com/download/

更換為國內源 https://mirrors.ustc.edu.cn/help/anaconda.html

conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/pkgs/main/
conda config --add channels https://mirrors.ustc.edu.cn/anaconda/cloud/conda-forge/
conda config --set show_channel_urls yes

然后安裝

conda install tensorflow-gpu==1.10

裝好后測試

import tensorflow as tf
tf.__version__

 

 

 

 

 

 

 

參考： https://cuiqingcai.com/5822.html