一 安装nvidia驱动410.93
1.禁止nouveau驱动
$ sudo vi /etc/modprobe.d/blacklist.conf 写入下面5行代码,保存并退出
blacklist nouveau
blacklist lbm-nouveau
options nouveau modeset=0
alias nouveau off
alias lbm-nouveau off
$ echo options nouveau modeset=0 | sudo tee -a /etc/modprobe.d/nouveau-kms.conf
$ sudo update-initramfs -u
$ sudo reboot
$ ls mod | grep nouveau
2.开始安装nvidia驱动
$ sudo apt install gcc make //提前配置清华源
$ sudo init 3
按ctrl + alt+ F1 进入tty模式
$ sudo service lightdm stop
$ sudo init 3
$ chmod +x NVIDIA-Linux-x86_64-410.93.run
$ sudo sh ./NVIDIA-Linux-x86_64-410.93.run --no-opengl-files
“continue installation”+“no”
$ nvidia-smi
二 安装cuda10
$ sudo sh cuda_10.0.130_410.48_linux.run
不选安装驱动
$ sudo gedit ~/.bashrc 将下面三行代码写入~/.bashrc文件保存后退出
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
export CUDA_HOME=/usr/local/cuda
export PATH=/usr/local/cuda-10.0/bin:$PATH
$ source ~/.bashrc
$ nvcc --version
三 装cudnn7.3.1
$ tar -xzvf 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 +r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
$ cd /usr/local/cuda/lib64
$ sudo ln -sf libcudnn.so.7.3.1 libcudnn.so.7
$ sudo ln -sf libcudnn.so.7 libcudnn.so
$ ll
$ sudo vi /etc/ld.so.conf.d/cuda.conf 将下面一行写入文件
/usr/local/cuda/lib64
$ sudo ldconfig
四 装python3.6.4
$ sudo apt-get install gcc make zlib1g-dev
$ sudo apt-get install libbz2-dev
$ sudo apt-get install libsqlite3-dev
$ sudo apt-get install python3-dev libxml2-dev libffi-dev libssl-dev libxslt1-dev
$ tar -zxvf Python-3.6.4.tgz
$ mv Python-3.6.4 /usr/local
$ cd /usr/local/Python-3.6.4/
$ ./configure
$ make
$ make install
$ ll /usr/bin | grep python
$ rm -rf /usr/bin/python
$ ln -s /usr/local/bin/python3.6 /usr/bin/python
$ python -V
$ cd /usr/bin
$ sudo mv pip pip2
$ sudo ln -s /usr/local/bin/pip3 /usr/bin/pip
$ pip --version
五 装openmpi-2.0.0
$ tar -zxvf openmpi-2.0.0.tar.gz
$ cd openmpi-2.0.0
$ ./configure --prefix="/usr/local/openmpi"
$ make
$ sudo make install
$ sudo gedit ~/.bashrc //写入下面两行,这两行需添加在原路径的后面
export PATH="$PATH:/usr/local/openmpi/bin"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/openmpi/lib/"
$ sudo ldconfig
六 装tensorflow-1.12和mxnet-cu100
pip安装不支持cuda10版tensorflow,去github搜 tensorflow wheel 下载大神编译的tensorflow.
注意:千万不要自己编译,会累死的,珍惜生命,远离编译.
$ sudo pip install tensorflow-1.12.0-cp36-cp36m-linux_x86_64.whl
$ sudo pip install mxnet-cu100
若ubuntu16.04安装报错:ImportError: /usr/lib/x86_64-linux-gnu/libstdc++.so.6: version GLIBCXX_3.4.22 not found (required by /usr/local/lib/python3.6/site-packages/tensorflow/python/../libtensorflow_framework.so)
或ImportError: /lib/x86_64-linux-gnu/libm.so.6: version GLIBC_2.27 not found (required by /usr/local/lib/python3.6/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so)
解决:升级ubuntu18.04系统
附:
卸载 cuda9.0及cudnn
$ sudo /usr/local/cuda-9.0/bin/uninstall_cuda_9.0.pl
$ sudo rm -rf /usr/local/cuda-9.0