在 Ubuntu 18.04 上安装 TensorFlow GPU 的过程中遇到几个问题，在此将过程记录下来。
前言
本文写于 2018 年 7 月，此时的 TensorFlow 最新版本为 1.9，此版本支持 CUDA Toolkit 9.0 和 cuDNN SDK v7。同时由于本人所安装的 Ubuntu 为英文版本，中文语言版本用户请寻找对应的程序位置。
NVIDIA 显卡驱动
首先打开 Software & Updates 中的 Additional Drivers 选项卡，将显卡驱动更改为 Using NVIDIA driver metapackage from nvidia-driver-390(proprietary, tested)
当前最新版本为 390，版本可能不同，请选择对应的版本。
CUDA 9.0
由于 TensorFlow 二进制包仅支持 CUDA 9.0，因此便选择 CUDA 9.0 进行安装。
下载地址：https://developer.nvidia.com/cuda-90-download-archive
选择 Linux - x86_64 - Ubuntu - 17.04 - runfile(local) 进行安装。
安装依赖
下载完成后安装依赖：
$ sudo apt install freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa libglu1-mesa-dev
由于 CUDA 9.0 不支持 GCC 6 以后的版本，本文这里将 GCC 降级为 6。
$ sudo apt install gcc-6 g++-6
$ sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-6 60
$ sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-6 60
运行安装脚本
$ sudo chmod a+x cuda_9.0.176_384.81_linux.run # 给予运行权限
$ sudo ./cuda_9.0.176_384.81_linux.run -toolkit -samples -silent -override
将 cuda_9.0.176_384.81_linux.run 修改为对应的文件名（一般来说是相同的）
设置环境变量
将下列环境变量添加到你的系统中，如 ~/.bashrc，并运行 source ~/.bashrc
# CUDA PATH
export PATH=/usr/local/cuda-9.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+${LD_LIBRARY_PATH}:}/usr/local/cuda-9.0/lib64
验证 CUDA 9.0 安装
首先运行 nvcc -V 查看 CUDA 版本，运行结果应如下所示：
$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Sep__1_21:08:03_CDT_2017
Cuda compilation tools, release 9.0, V9.0.176
之后进入 ~/NVIDIA_CUDA-9.2_Samples 下运行 make 命令进行编译。编译完成后进入 ~/NVIDIA_CUDA-9.0_Samples/bin/x86_64/linux/release 运行 ./deviceQuery 命令。运行结果应如下所示：
$ ./deviceQuery
./deviceQuery Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GeForce GTX 1070"
CUDA Driver Version / Runtime Version 9.1 / 9.0
CUDA Capability Major/Minor version number: 6.1
Total amount of global memory: 8111 MBytes (8505131008 bytes)
(15) Multiprocessors, (128) CUDA Cores/MP: 1920 CUDA Cores
GPU Max Clock rate: 1683 MHz (1.68 GHz)
Memory Clock rate: 4004 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
Maximum Layered 1D Texture Size, (num) layers 1D=(32768), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(32768, 32768), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: Yes
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Supports Cooperative Kernel Launch: Yes
Supports MultiDevice Co-op Kernel Launch: Yes
Device PCI Domain ID / Bus ID / location ID: 0 / 1 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 9.1, CUDA Runtime Version = 9.0, NumDevs = 1
Result = PASS
以及 ./bandwidthTest 命令：
$ ./bandwidthTest
[CUDA Bandwidth Test] - Starting...
Running on...
Device 0: GeForce GTX 1070
Quick Mode
Host to Device Bandwidth, 1 Device(s)
PINNED Memory Transfers
Transfer Size (Bytes) Bandwidth(MB/s)
33554432 12510.8
Device to Host Bandwidth, 1 Device(s)
PINNED Memory Transfers
Transfer Size (Bytes) Bandwidth(MB/s)
33554432 12739.0
Device to Device Bandwidth, 1 Device(s)
PINNED Memory Transfers
Transfer Size (Bytes) Bandwidth(MB/s)
33554432 181601.2
Result = PASS
NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.
cuDNN v7
前往 https://developer.nvidia.com/rdp/cudnn-archive 下载对应的 cuDNN，点击 “Download cuDNN v7.0.5 (Dec 5, 2017), for CUDA 9.0”，下载以下三个链接。
- cuDNN v7.0.5 Runtime Library for Ubuntu16.04 (Deb)
- cuDNN v7.0.5 Developer Library for Ubuntu16.04 (Deb)
- cuDNN v7.0.5 Code Samples and User Guide for Ubuntu16.04 (Deb)
下载完成后进行安装：
$ sudo dpkg -i libcudnn7_7.0.5.11-1+cuda9.0_amd64.deb
$ sudo dpkg -i libcudnn7-dev_7.0.5.11-1+cuda9.0_amd64.deb
$ sudo dpkg -i libcudnn7-doc_7.0.5.11-1+cuda9.0_amd64.deb
安装 TensorFlow
安装 libcupti-dev 依赖
sudo apt-get install libcupti-dev
并将以下环境变量加入系统：
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+${LD_LIBRARY_PATH}:}/usr/local/cuda/extras/CUPTI/lib64
（可选）固定 cuDNN 版本：
运行以下命令固定 cuDNN 版本
$ sudo apt-mark hold libcudnn7 libcudnn7-dev libcudnn7-doc
取消固定：
$ sudo apt-mark unhold libcudnn7 libcudnn7-dev libcudnn7-doc
TensorFlow GPU
运行以下命令：
$ pip3 install tensorflow-gpu
在 Python 中验证安装：
import tensorflow as tf
with tf.Session() as sess:
print(sess.run(tf.constant('Hello, TensorFlow!')))
运行结果应如下所示：
/home/ubuntu/.pyenv/versions/3.6.6/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88
return f(*args, **kwds)
/home/ubuntu/.pyenv/versions/3.6.6/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: numpy.dtype size changed, may indicate binary incompatibility. Expected 96, got 88
return f(*args, **kwds)
2018-07-31 03:55:51.119408: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2018-07-31 03:55:51.218358: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-07-31 03:55:51.218692: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1392] Found device 0 with properties:
name: GeForce GTX 1070 major: 6 minor: 1 memoryClockRate(GHz): 1.683
pciBusID: 0000:01:00.0
totalMemory: 7.92GiB freeMemory: 7.33GiB
2018-07-31 03:55:51.218705: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2018-07-31 03:55:52.587414: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-07-31 03:55:52.587483: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958] 0
2018-07-31 03:55:52.587518: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0: N
2018-07-31 03:55:52.587926: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 7077 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1070, pci bus id: 0000:01:00.0, compute capability: 6.1)
b'Hello, TensorFlow!'