tensorflow-gpu环境搭建

Posted on | In ,
Symbols count in article: 3.7k | Reading time ≈ 3 mins.

tensorflow-gpu环境搭建

环境依赖

1
2
3
4
5
Ubuntu-16.04.3-desktop-amd64
cuda_8.0.61_375.26_linux.run
cudnn-8.0-linux-x64-v7.tgz
Anaconda3-4.2.0-Linux-x86_64.sh
nvidia driver 384.111

换apt-get源

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
deb http://mirrors.aliyun.com/ubuntu/ xenial main restricted universe multiverse  
deb http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted universe multiverse  
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted universe multiverse  
deb http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse  

deb http://mirrors.aliyun.com/ubuntu/ xenial-proposed main restricted universe multiverse  

deb-src http://mirrors.aliyun.com/ubuntu/ xenial main restricted universe multiverse  
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted universe multiverse  
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted universe multiverse  
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse  

deb-src http://mirrors.aliyun.com/ubuntu/ xenial-proposed main restricted universe multiverse  

deb http://archive.canonical.com/ubuntu/ xenial partner  
deb http://extras.ubuntu.com/ubuntu/ xenial main

安装vim

1
2
sudo apt-get remove vim-common
sudo apt-get install vim

安装gdebi

1
sudo apt-get install gdebi

禁用nouveau

1
sudo vim /etc/modprobe.d/blacklist.conf

在文件最后插入以下两行内容

1
2
blacklist nouveau
options nouveau modeset=0
1
2
sudo update-initramfs -u
sudo reboot
1
2
lsmod | grep nouveau
没有输出则为成功

gcc降版本

ubuntu的gcc编译器是5.4.0，然而cuda8.0不支持5.0以上的编译器，因此需要降级，把编译器版本降到4.9

1
2
3
4
5
6
7
8
9
sudo apt-get install g++-4.9
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.9 20
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-5 10
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.9 20
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-5 10
sudo update-alternatives --install /usr/bin/cc cc /usr/bin/gcc 30
sudo update-alternatives --set cc /usr/bin/gcc
sudo update-alternatives --install /usr/bin/c++ c++ /usr/bin/g++ 30
sudo update-alternatives --set c++ /usr/bin/g++

安装缺失的包

1
2
sudo apt-get install libglu1-mesa-dev libx11-dev libxi-dev libxmu-dev libgl1-mesa-dev
sudo updatedb

安装 cuda8.0

1
2
3
sudo service lightdm stop
alt+ctrl+F1
./cuda_8.0.61_375.26_linux.run

不要安装驱动和 opencl 和 X-config 其他默认,完成后

1
sudo service lightdm start

在确保 apt-get 源已经更新后，设置里 GUI 安装 nvidia 驱动384.111。

添加环境变量

1
sudo vi ~/.bashrc

最后加入以下几行

1
2
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda-8.0/lib64:/usr/local/cuda-8.0/extras/CUPTI/lib64"
export CUDA_HOME=/usr/local/cuda-8.0

安装 nvidia 工具包

1
2
3
sudo apt install nvidia-cuda-toolkit
sudo reboot
nvcc -V   //输出 nvidia 驱动的版本信息

测试 cuda 是否安装成功

1
2
3
cd /usr/local/cuda/samples/1_Utilities/deviceQuery
make .
./deviceQuery 输出 GPU 信息

安装 cuDNN

1
2
3
4
5
tar xvzf cudnn-8.0-linux-x64-v7.tgz

sudo cp cuda/include/cudnn.h /usr/local/cuda-8.0/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda-8.0/lib64
sudo chmod a+r /usr/local/cuda-8.0/include/cudnn.h /usr/local/cuda-8.0/lib64/libcudnn*

安装 Anaconda

1
./Anaconda-4.2.0 不要用 sudo

conda 更新源

1
2
3
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --set show_channel_urls yes
vi ~/.condarc 把 default 去掉

安装 tensorflow

不用更新 conda 包列表,直接

1
2
conda install tensorflow-gpu
sudo reboot

测试 tensorflow

1
2
3
4
5
6
import tensorflow as tf
a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
c = tf.matmul(a, b)
sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
print(sess.run(c))

References

http://blog.csdn.net/zhaoyu106/article/details/52793183

https://segmentfault.com/a/1190000008234390