ubuntu16+tensorflow-gpu安装cuda

1. 初始化环境

sudo apt-get update 

sudo apt-get upgrade 

sudo apt-get install -y build-essential git python-pip libfreetype6-dev libxft-dev libncurses-dev libopenblas-dev gfortran python-matplotlib libblas-dev liblapack-dev libatlas-base-dev python-dev python-pydot linux-headers-generic linux-image-extra-virtual unzip python-numpy swig python-pandas python-sklearn unzip wget pkg-config zip g++ zlib1g-dev 

1. 下载 cuda [注:tensorflow-gpu当前pip安装版本1.9，支持cuda9.0+cudnn7.1.4for cuda9.0

http://developer.download.nvidia.com/compute/cuda/repos/ 选择对应系统版本，本次使用cuda9

sudo dpkg -i  cuda-repo-ubuntu1410_XXXX_amd64.deb 

rm cuda-repo-ubuntu1410_XXXX_amd64.deb 

sudo apt-get update 

sudo apt-get install -y cuda

安装错误卸载 sudo dpkg -P packagename, sudo apt-get autoremove cuda

2. 下载cudnn

https://developer.nvidia.com/rdp/cudnn-archive
选择 cuDNN vXXX Library for Linux，目前下载cudnn7.1.4 for cuda9.0

cp cudnn-XXX.solitairetheme8 cudnn-linux-x64.tgz
tar -xvf cudnn-linux-x64.tgz

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*

3. 配置环境变量

sudo vi ~/.bashrc
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
export CUDA_HOME=/usr/local/cuda
export PATH="$CUDA_HOME/bin:$PATH"
source ~/.bashrc

4. 安装tensorflow-gpu == 1.9

pip install keras
pip install tensorflow-gpu==1.9

5. 安装运行

import tensorflow as tf

提示

h5py numpy警告
pip install h5py==2.8.0rc1