深度学习系列之平台安装：TensorFlow安装（gpu版本，ubuntu16.04）

1.安装驱动:

参考：https://blog.csdn.net/sinat_40276791/article/details/80403784

• 查看驱动推荐：
  $ ubuntu-drivers devices
  == /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0 ==
  modalias : pci:v000010DEd00001C81sv00001028sd000011C0bc03sc00i00
  vendor : NVIDIA Corporation
  model : GP107 [GeForce GTX 1050]
  driver : nvidia-driver-396 - third-party free recommended
  driver : nvidia-driver-390 - third-party free
  driver : xserver-xorg-video-nouveau - distro free builtin

安装：

$ sudo ubuntu-drivers autoinstall

按照推荐的驱动安装，如果失败，请添加NVIDIA driver源仓库然后在安装，安装命令如下：

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

在添加源并update之后可通过setting/software update安装

测试安装是否成功

$ nvidia-smi

Tue May 22 12:14:56 2018
±----------------------------------------------------------------------------+
| NVIDIA-SMI 390.59 Driver Version: 390.59 |
|-------------------------------±---------------------±---------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|=++==============|
| 0 GeForce GTX 1050 Off | 00000000:01:00.0 Off | N/A |
| 20% 39C P0 N/A / 75W | 444MiB / 2000MiB | 0% Default |
±------------------------------±---------------------±---------------------+

±----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1155 G /usr/lib/xorg/Xorg 28MiB |
| 0 1252 G /usr/bin/gnome-shell 48MiB |
| 0 1539 G /usr/lib/xorg/Xorg 163MiB |
| 0 1713 G /usr/bin/gnome-shell 107MiB |
| 0 2442 G …-token=755D9B9E74B9E412D1116412B89F4DC2 93MiB |
±----------------------------------------------------------------------------+

2.安装cuda:

• 下载地址（注意cuda版本和cudnn与tensorflow官网编译时使用版本一致，见文章末尾链接）
  https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1604&target_type=deblocal

• 下载完成之后通过以下命令安装

    sudo dpkg -i cuda-repo-ubuntu1604-9-2-local_9.2.88-1_amd64.deb
    sudo apt-key add /var/cuda-repo-<version>/7fa2af80.pub
    sudo apt-get update
    sudo apt-get install cuda

• 如果安装版本不对，卸载方法如下:

sudo apt-get purge cuda
dpkg -l | grep ^rc | cut -d’ ’ -f3 | sudo xargs dpkg --purge

3.安装cudnn:

下载地址同上，版本注意与cuda对应

tar -xvf cudnn-9.0-linux-x64-v7.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*

• 添加环境变量:

echo 'export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"' >> ~/.bashrc
echo 'export CUDA_HOME=/usr/local/cuda' >> ~/.bashrc
source ~/.bashrc

4.安装tensorflow1.14:

pip install tensorflow-gpu==1.14.0

• check it:

python input：

>>> from tensorflow.python.client import device_lib
>>> device_lib.list_local_devices()

list:
[name: “/device:CPU:0”
device_type: “CPU”
memory_limit: 268435456
locality {
}
incarnation: 8754742150252655686
, name: “/device:GPU:0”
device_type: “GPU”
memory_limit: 1276313600
locality {
bus_id: 1
links {
}
}
incarnation: 17576461877219777618
physical_device_desc: “device: 0, name: GeForce GTX 1050, pci bus id: 0000:01:00.0, compute capability: 6.1”
]

官方教程中安装出错问题

使用

`pip install https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.5.0-cp27-none-linux_x86_64.whl`

提示：

Could not install packages due to an EnvironmentError: [Errno 13] Permission denied: '/usr/local/lib/python2.7/dist-packages/numpy-1.14.3.dist-info'
Consider using the `--user` option or check the permissions.

在最后加上–user即可

• 按照turorials安装后example无法执行

tensorflow版本太低，为0.5.0，需要更新：

pip install --upgrade tensorflow --user

或者在安装tensorflow时直接指定安装版本（gpu版本）：

pip install pip install tensorflow-gpu==1.8.0

tensorflow安装后gpu无法启用，检查cuda版本和cudnn版本是否与官网编译环境完全一致：https://www.tensorflow.org/install/source_windows