Ubuntu 16.04 CUDA_10.0及cuDNN的安装与卸载

(cuda版本：cuda_10.0.130_410.48_linux.run
补丁：cuda_10.0.130.1_linux.run
cudnn版本：cudnn-10.0-linux-x64-v7.3.1.20.tgz)

cuda

下载

cuda官网下载(https://developer.nvidia.com/cuda-10.0-download-archive) :选择runfile安装方式
下面还有个补丁3.3兆一块下了

安装

#当前目录下打开终端输入
sudo chmod +x ./cuda_10.0.130_410.48_linux.run
sudo ./cuda_10.0.130_410.48_linux.run

添加PATH

gedit ~/.bashrc
最后添加

# added by cuda 10.0 installer
export PATH="/usr/local/cuda-10.0/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH"

soucer ~/.bashrc

补丁

补丁cuda_10.0.130.1_linux.run 同上一样安装

sudo chmod +x ./cuda_10.0.130.1_linux.run
sudo ./cuda_10.0.130.1_linux.run

验证

nvcc -V

cd /usr/local/cuda-10.0/samples/1_Utilities/deviceQuery
sudo make && ./deviceQuery

若最后显示Result = PASS，表明查询显卡信息成功
最后执行sudo make clean清除垃圾文件

cudnn

cudnn的安装文件：https://developer.nvidia.com/rdp/cudnn-archive (其他7.6.4等版本类似如下安装)

解压

tar -xvf cudnn-10.0-linux-x64-v7.3.1.20.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
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

查看

cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

卸载cudnn

报错：Loaded runtime CuDNN library: 7.3.1 but source was compiled with: 7.4.2
tensorflow1.14需要 7.4.2及以上版本

sudo rm -rf /usr/local/cuda/include/cudnn.h
sudo rm -rf /usr/local/cuda/lib64/libcudnn*

卸载后重复上面安装cudnn步骤

个版本tensorflow 镜像下载

http://mirrors.aliyun.com/pypi/simple/tensorflow-gpu/

更新软链接

sudo chmod +r libcudnn.so.7.4.2
sudo ln -sf libcudnn.so.7.4.2 libcudnn.so.7  
sudo ln -sf libcudnn.so.7 libcudnn.so     
sudo ldconfig

最后再验证

cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2