安装CUDA8.0 cuDNN5.1
2016-09-10 14:35
405 查看
环境:GTX 1080, Ubuntu 16.04
下载Nvidia-367.18-driver.run file
删除旧的驱动 (sudo apt remove –purge nvidia*)
删除旧的nvidia.deb packages
删除cuda 文件夹(/usr/local/cuda*) (sudo apt remove –purge libcuda*)
reboot
sudo service ligdm stop
run the Nvidia-367.18.run
run the cuda-8.run file.(注意:在某一步不要安装自带的361版本的驱动)
reboot
sudo service ligdm start
在最后加入以下两行,保存
export PATH=/usr/local/cuda-8.0/bin:
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:
然后使其生效, source /etc/profile
或者下面更潇洒(不过还没试验)
echo ‘export PATH=/usr/local/cuda/bin:
echo ‘export LD_LIBRARY_PATH=/usr/local/cuda/lib64:
source ~/.bashrc
2)nvcc –version
3) 测试cuda的sample
cd /home/zhyj3038/NVIDIA_CUDA-8.0_Samples
sudo make
然后
cd /NVIDIA_CUDA-8.0_Samples
1)./bin/x86_64/linux/release/deviceQuery
2)./5_Simulations/nbody/nbody
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*
准备
下载cuda-8.run file(保证MD5值和官网一致,为5639ffeb939ee58a81554d06bd084e15 cuda_8.0.27_linux.run)下载Nvidia-367.18-driver.run file
删除旧的驱动 (sudo apt remove –purge nvidia*)
删除旧的nvidia.deb packages
删除cuda 文件夹(/usr/local/cuda*) (sudo apt remove –purge libcuda*)
reboot
安装驱动和cuda
进入文字界面 tty (ctrl-alt-1) ,sudo service ligdm stop
run the Nvidia-367.18.run
run the cuda-8.run file.(注意:在某一步不要安装自带的361版本的驱动)
reboot
sudo service ligdm start
设置环境变量
sudo gedit /etc/profile在最后加入以下两行,保存
export PATH=/usr/local/cuda-8.0/bin:
$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:
$LD_LIBRARY_PATH
然后使其生效, source /etc/profile
或者下面更潇洒(不过还没试验)
echo ‘export PATH=/usr/local/cuda/bin:
$PATH’ >> ~/.bashrc
echo ‘export LD_LIBRARY_PATH=/usr/local/cuda/lib64:
$LD_LIBRARY_PATH’ >> ~/.bashrc
source ~/.bashrc
校验
1)nvidia-smi2)nvcc –version
3) 测试cuda的sample
cd /home/zhyj3038/NVIDIA_CUDA-8.0_Samples
sudo make
然后
cd /NVIDIA_CUDA-8.0_Samples
1)./bin/x86_64/linux/release/deviceQuery
2)./5_Simulations/nbody/nbody
cuDNN5.1
tar -zxvf cudnn-8.0-linux-x64-v5.1.tgzsudo 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*
相关文章推荐
- [深度学习]Ubuntu16.04 + GTX 1050 + cuda8.0 + cuDNN5.1 + caffe安装详解
- ubuntu16.04 安装CUDA 8.0 和 cuDNN 5.1 /cudnn6.0,可适用于gpu版本的(tensorflow,caffe,mxnet)
- Ubuntu 16.04+Gtx1050Ti+cuda 8.0+cudnn 5.1 tensorflow 安装
- Ubuntu 14.04 安装 CUDA8.0 cudnn 5.1 tensorflow1.2.1GPU
- [深度学习]Ubuntu16.04 + GTX 1050 + cuda8.0 + cuDNN5.1 + caffe安装详解
- Ubuntu 14.04环境下安装CUDA 8.0及Cudnn 5.1
- Ubuntu16.04 安装 CUDA8.0 + cudnn5.1 + TensorFlow(GPU) 详细过程
- Matconvnet安装:win10+VS2015(pro)+Matlab 2017a+cuda8.0+cudnn 5.1
- ubuntu16.04安装CUDA8.0和CUDNN5.1
- Ubuntu 16.04安装CUDA8.0 + cuDNN5.1 + OpenCV3.1.0
- caffe安装 Ubuntu14.04 cuda 8.0 cudnn 5.1
- GTX1080 安装 cuda 8.0 + cuDNN5.1
- ubuntu14.04+caffe+cuda8.0+cudnn-8.0-v5.1(gtx1070)安装与测试
- 深度学习主机环境配置: Ubuntu16.04 + GeForce GTX 1070 + CUDA8.0 + cuDNN5.1 + TensorFlow
- ubuntu16.04 64位安装tensorflow+cuda8.0+cudnn7.0
- Ubuntu14.04 & CUDA8.0 & Theano & Tensorflow & TensorLayer & Cudnn安装血泪史
- Ubuntu16.04 (WIN10)+ cuda8.0 + cudnn5.1 + tensorflow1.0.1配置
- Ubuntu16.04+tensorflow(gpu)+Cuda(8.0)+cudnn(5.1)
- Ubuntu 14.04 64位系统安装cuda8.0+cudnn7.5+opencv+caffe 血泪教程
- Ubuntu16.04安装cuda8.0,cudnn6.0,并在conda里安tensorflow1.3(与cuda8.0匹配),pycharm的gpu加速环境变量配置