TensorFlow入门01：环境搭建

1、CPU版本安装

1.1 安装tensorflow

pip3 install --upgrade tensorflow

1.2 Python验证，看到版本信息就可以了

python3
>>> import tensorflow as tf
>>> print('Tensorflow version ', tf.__version__)

Tensorflow version  1.12.0

2、GPU版本安装（需要NVIDIA显卡）

2.1 检查驱动信息

nvidia-smi

Fri Nov 16 21:22:13 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.77                 Driver Version: 390.77                    |
|-------------------------------+----------------------+----------------------+
| 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 106...  Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   52C    P2    27W /  N/A |   5938MiB /  6078MiB |     22%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      7953      G   /usr/lib/xorg/Xorg                           126MiB |
|    0      8215      G   /usr/bin/gnome-shell                         109MiB |
|    0     13578    C+G   python3                                     5689MiB |
+-----------------------------------------------------------------------------+

2.2 安装CUDA

# 查看网站 https://developer.nvidia.com/cuda-90-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1704&target_type=runfilelocal
# 选择下载这个版本 Linux x86_64 Ubuntu 17.04 runfile
# 安装，但注意不要更新驱动
sudo chmod +x cuda_9.0.176_384.81_linux.run
./cuda_9.0.176_384.81_linux.run --override

2.3 安装CUDNN

# 查看网站 https://developer.nvidia.com/rdp/cudnn-download
# 选择下载这个版本 9.0 cuDNN Library for Linux
# 解压
tar -zxvf cudnn-9.0-linux-x64-v7.tgz
# 手工安装
sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-9.0/lib64/
sudo cp  cuda/include/cudnn.h /usr/local/cuda-9.0/include/
# 调整权限
sudo chmod a+r /usr/local/cuda-9.0/include/cudnn.h /usr/local/cuda-9.0/lib64/libcudnn*

2.3 安装libcupti-dev

sudo apt-get install libcupti-dev

2.4 修改.bashrc

# 增加下面两行
export PATH=/usr/local/cuda-9.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

2.5 安装tensorflow-gpu

pip3 install --upgrade tensorflow-gpu

2.6 Python验证，看到GPU就可以啦

Python3
>>> from tensorflow.python.client import device_lib
>>> device_lib.list_local_devices()

[name: "/device:CPU:0"
device_type: "CPU"
memory_limit: 268435456
locality {
}
...
incarnation: 2559160109308400478
physical_device_desc: "device: 0, name: GeForce GTX 1060 with Max-Q Design, pci bus id: 0000:01:00.0, compute capability: 6.1"
]

3、Docker方式安装
3.1 CPU版

# 运行tensorflow
docker run -it -p 8888:8888 gcr.io/tensorflow/tensorflow

3.2 GPU版

# 安装nvidia-docker
wget https://github.com/NVIDIA/nvidia-docker/releases/download/v1.0.1/nvidia-docker_1.0.1-1_amd64.deb
sudo dpkg -i nvidia-docker*.deb

# 测试nvidia-docker，执行nvidia-smi命令
nvidia-docker run --rm nvidia/cuda nvidia-smi

# 运行tensorflow
nvidia-docker run -it -p 8888:8888 gcr.io/tensorflow/tensorflow:latest-gpu

4、编译CUDA Demo（非必须）

# 咱们选用的版本只支持到gcc6
apt-get install gcc-6 g++-6
ln -s /bin/gcc /bin/gcc-6

# 安装libmpich-dev
sudo apt-get install libmpich-dev


# 切换路径
cd PATH_TO_DEMO

# 编译
make