1, the installation anaconda
(1) download ananconda * .sh files go anaonda official or Tsinghua mirror source
(2) into the directory where you downloaded the file, execute ananconda * .sh bash
(3) all the way yes, and finally the installation of input vs NO
(4) restart, In other source
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes
2, the installation keras, to the rear end of TensorFlow-GPU
(1) create a virtual environment and activate
conda create -n py36-keras python=3.6
source activate py36-keras
Py36-keras which is the name of a virtual environment
(2) Installation tensorflow-gpu
conda install tensorflow-gpu
Of course, you can also specify the version here, where I installed the default version of the
note, installed after the installation tensorflow-gpu cuda and cudnn not need to go alone, conda when installing tensorflow-gpu has helped us install better matching and cuda cudnn
(. 3) mounted keras
conda install keras
So ok friends ~