1. Create a TensorFlow environment
conda create -n tensorflow python=3.6
activate tensorflow
2. Installation tensorflow-gpu
conda install tensorflow-gpu
It's that simple line of code you can get! ! ! In fact, I wonder why some people would be here a long time somersaults, the feeling is divided into several main reasons for it:
(1) network instability. This is certainly have an impact, it is best to do one with a 4G hotspot or wired network, and second, can a little over the wall (for this reason I do not say it)
(2) When installing the software did not choose a "auspicious." The official server is the time in the United States, the domestic daytime, sleeping foreigner, relatively smooth network, install the software at this time is more likely to be successful; at night, foreigners there during the day, they got up access to the Internet, we access the server time, the network is relatively blocked, this time-line installation is more prone to failure. (This is my own imagination, a little metaphysical sense, and if not, then, when you are a fun treat it)
（3）
