在安装好nvidia驱动的基础上安装 CUDA9.0 + cudnn7.0 + TensorFlow-gpu1.7.0 这三个是匹配的版本
别的匹配(CUDA8.0 + cudnn6.0 + TensorFlow-gpu1.4.0),更高版本没有了解,日后补充
1.安装CUDA9.0
下载CUDA9.0 runfile 文件,执行
sudo sh cuda_9.0.run
一路accept,yes,设置cuda_toolkit 和cuda_samples路径
然后设置环境变量
sudo gedit ~/.bash_profile export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64" export CUDA_HOME=/usr/local/cuda
export PATH=$PATH:/usr/local/cuda/bin
source ~/.bash_profile
2.安装cudnn
这里的路径是自己解压cudnn压缩包后的路径
tar xvzf cudnn-8.0-linux-x64-v6.0-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*
在 ~/.bashrc文件中配置环境变量
export CUDNN_HOME=/home/hhb/local/cudnn/cuda export LD_LIBRARY_PATH=${CUDNN_HOME}/lib64:$LD_LIBRARY_PATH export CPLUS_INCLUDE_PATH=${CUDNN_HOME}/include:$CPLUS_INCLUDE_PATH
source ~/.bashrc
3.安装TensorFlow 1.7GPU版
sudo pip install tensorflow-gpu==1.7.0
4.查看cuda版本和cudnn版本
cat /usr/local/cuda/version.txt cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
查看cuda_toolkit是否正确配置
nvcc --version