先anaconda
open anaconda prompt
$ conda create -n tensorflow python=3.5
確保名叫tensorflow的環境已經被成功添加：conda info --envs
$ activate tensorflow.
$ conda install tensorflow
Verifying the Installation
$python
import tensorflow as tf
錯誤為ImportError: cannot import name 'weakref'，那么請在tensorflow的的安裝位置中找到名為的tf_should_use.py文件，將第28行的from backports import weakref替換為import weakref
open pycharm
interpreter->add python interpreter-> existing environment->LIJINGYU070/Anaconda3/envs/tensorflow/python.exe
package + ->tensorflow->install
package 'tensorflow' installed successfully
==============
安裝了scikit learn之后，numpy報錯 multiarray找不到, 連numpy都用不了了
pip uninstall numpy
pip install -U numpy
用conda install老不行
pandas總是報錯cant determine bottleneck
conda install bottleneck 就好了
在create一個envtensorflow之后，install tensorflow，發現import numpy報錯multiarray找不到，
pip install numpy --upgrade就好了
裝pandas 一裝就壞，報錯numexpr缺少__version__屬性
pip install pandas --upgrade就好了
嚇得我只敢 pip install scikit-learn --upgrade
Finally!!!
Done!!!