在win10 / deepin 15.6 / Ubuntu上配置CUDA 和cuDNN

windows 10 + cuda9 + cuDNN7.1

##1. 安装GeForce Experience更新最新显卡驱动
##2. 安装CUDA 9.0
Step 1: 下载cuda 9.0
https://developer.nvidia.com/cuda-toolkit-archive
Step 2: 安装cuda 9.0
不要勾选driver component和GeForce Experience
Step 3: 安装cuDNN 7.1
将bin, include, lib复制到cuda文件夹C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0
Step 4: 添加环境变量

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0; C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin; C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\include; C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\lib; C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\lib\x64

Step 5: 测试

---

---

Ubuntu + CUDA9.0 + cuDNN7.1.4

##1. 安装最新驱动
方法一：
参考

查看最新驱动版本：https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa

sudo apt-get purge nvidia*
sudo add-apt-repository ppa:graphics-drivers
sudo apt-get update
sudo apt-get install nvidia-driver-535

blocking minor version updates
sudo apt-mark hold nvidia-390
重启，检查nvidia-smi。
此方法有时会有显卡驱动错误，导致nvidia-smi无法使用，原因未明

方法二：
参考

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update

查看推荐驱动

xin@xin-MS-7B48:~$ sudo ubuntu-drivers devices
== /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0 ==
modalias : pci:v000010DEd00001B81sv00001458sd00003701bc03sc00i00
vendor   : NVIDIA Corporation
model    : GP104 [GeForce GTX 1070]
driver   : nvidia-driver-396 - third-party free recommended
driver   : nvidia-driver-390 - third-party free
driver   : xserver-xorg-video-nouveau - distro free builtin

sudo apt install nvidia-driver-396

##2. 安装CUDA9.0
下载
https://developer.nvidia.com/cuda-90-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1704&target_type=runfilelocal
xianjiangji
先降级gcc g++

#降级g++
sudo apt install g++-4.8 gcc-4.8
cd /usr/bin
sudo rm gcc g++
sudo ln -s g++-4.8 g++
sudo ln -s gcc-4.8 gcc

安装cuda
sudo sh cuda_9.0.176_384.81_linux.run

##3. 安装cuDNN 7.1.4
下载
https://developer.nvidia.com/rdp/cudnn-archive
解压

tar xzvf cudnn-9.0-linux-x64-v7.1.tgz
cd cuda

将cuda文件夹中lib64和include复制到/usr/local/cuda-9.0中

cd /home/xin/Downloads/cuda/include
sudo cp cudnn.h /usr/local/cuda-9.0/include
cd /home/xin/Downloads/cuda/lib64
sudo cp -a . /usr/local/cuda-9.0/lib64

---

---

deepin 15.6 + cuda 8 + cuDNN7.1

##1. 安装最新驱动
sudo apt-get install nvidia-driver

##2. 安装cuda
参考

sudo apt update
sudo apt install bumblebee bumblebee-nvidia nvidia-smi

#降级g++
sudo apt install g++-4.8 gcc-4.8
cd /usr/bin
sudo rm gcc g++
sudo ln -s g++-4.8 g++
sudo ln -s gcc-4.8 gcc

sudo apt install nvidia-cuda-dev nvidia-cuda-toolkit nvidia-nsight nvidia-visual-profiler

##3. 安装cuDNN 7.1
参考：https://blog.csdn.net/luzhanbo207/article/details/78036278
###3.1. 下载cuDNN
https://developer.nvidia.com/rdp/cudnn-archive
https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.1.3/prod/8.0_20180414/cudnn-8.0-linux-x64-v7.1
###3.2. 解压并将文件复制到相应文件夹
由于cuda是用使用apt安装，系统将cuda直接安装在/usr中

tar xzvf cudnn-8.0-linux-x64-v7.1.tgz
cd ./cuda/include
sudo cp *.h /usr/include (这里是你安装cuda的地址，由于使用apt安装，系统将cuda直接安装在/usr目录)
cd ./cuda/lib64
sudo cp libcudnn.so /usr/lib
sudo cp libcudnn.so.7 /usr/lib
sudo cp libcudnn.so.7.1.3 /usr/lib
sudo cp libcudnn_static.a /usr/lib
sudo ln -s /usr/libcudnn.so.7.1.3 /usr/libcudnn.so.7
sudo ln -s /usr/libcudnn.so.7 /usr/libcudnn.so