Skip to content

Latest commit

 

History

History
197 lines (144 loc) · 4.65 KB

Ubuntu-install-cuda-tensorflow.md

File metadata and controls

197 lines (144 loc) · 4.65 KB
Raw

STEP 1: Install Ubuntu 16.04

  1. 到 Ubuntu官网 下载ISO镜像, 刻录进U盘
  2. 进入Win10，打开磁盘管理，压缩出足够的的磁盘空间(40GB or more)
  3. Reboot，进入BIOS，关闭 Security boot 及 Win10 的 Fast boot (Important)
  4. Reboot，从USB 引导进入安装界面
  5. 选择Ubuntu与Windows共存，一路安装到底

STEP 2: Install NVIDIA Driver

  • 更新源和必要的软件，如果在国内请自行更换合适的source
sudo apt-get update
sudo apt-get upgrade
  • 禁用Nouveau
sudo vi /etc/modprobe.d/disable-nouveau.conf

//加入如下两行
blacklist nouveau
options nouveau modeset=0
  • 重建kernel initramfs并重新启动
sudo update-initramfs -u
sudo reboot
  • 安装NVIDIA 驱动

重启进入登录界面，切换到tty1(ctrl+alt+f1), 关闭lightdm图形界面

sudo service lightdm stop

增加 Nvidia 的 ppa 源

sudo apt-get purge nvidia-*
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update

安装Nvidia

sudo apt-get install nvidia-375
reboot

重启，再次进入tty1，执行如下命令，没问题则ok了

sudo apt-get update && sudo apt-get upgrade

最后，用 nvidia-smi 查看GPU的信息

> nvidia-smi
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 375.82                 Driver Version: 375.82                    |
|-------------------------------+----------------------+----------------------+
| 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 980 Ti  Off  | 0000:01:00.0      On |                  N/A |
|  0%   56C    P8    17W / 250W |    471MiB /  6076MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0       437    G   fcitx-qimpanel                                  11MiB |
|    0     32510    G   compiz                                         122MiB |
+-----------------------------------------------------------------------------+

STEP 3: Install CUDA 8.0

  • 从官网下载CUDA文件(以cuda_8.0.61_375.26_linux.run为例)
  • 加执行权限并安装
  • 安装时会询问是否安装显卡驱动，务必选择No，前面已自行安装
cd Downloads/
sudo chmod a+x cuda_8.0.61_375.26_linux.run
sudo ./cuda_8.0.61_375.26_linux.run
  • 设置环境变量
sudo vi ~/.bashrc  
//加入这两行， LD_LIBRARY_PATH 和 CUDA_HOME 都不能少
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64
export CUDA_HOME=/usr/local/cuda-8.0
  • 保存后，务必使环境变量生效
source ~/.bashrc
  • 测试CUDA的Sample
cd /usr/local/cuda-8.0/samples/1_Utilities/deviceQuery
sudo make
sudo ./deviceQuery

STEP 4: Install Cudnn

  • 到 官网 下载(需注册账号)， 解压 & 复制文件 & 加执行权限
cd Downloads/
tar -zxvf cudnn-8.0-linux-x64-v5.1.tgz 
sudo cp cuda/include/cudnn.h /usr/local/cuda-8.0/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda-8.0/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda-8.0/lib64/libcudnn*

STEP 5: Install Tensorflow via pip

  • 安装必要依赖
sudo apt-get install libcupti-dev
  • 安装python-pip python-dev 并更新到最新版
sudo apt-get install python-pip python-dev
pip install -U pip
//or python3
sudo apt-get install python3-pip
pip3 install --upgrade pip
  • 安装tensorflow
//python
sudo pip2 install tensorflow-gpu
//or python3
sudo pip3 install tensorflow-gpu
  • 测试

开启一个terminal,这里我测试py3下的tensorflow。

bg@cgilab:~$ python3
>>> import tensorflow as tf
>>> hello = tf.constant('Hello, TensorFlow!')
>>> sess = tf.Session()
>>> print(sess.run(hello))

输出，成功！

Hello, TensorFlow!

Take it easy!