The first blog, the main record of the installation of deep Learning Framework (TensorFlow), installed very simple (a word), started I thought it was troublesome, no n card, do not know how to start. Okay, here we are.
I generally like the pydev of Eclipse (personal feeling is very useful), then Python is mainly Anconda (Tsinghua Mirror), loading TensorFlow
1. Download Anconda3, Baidu ancond Tsinghua Mirror, download anconda3-cp35-cp35m-xxxx, download the installation, download eclipse (free installation), and then configure Pydev, not related to this article, not described here.
2. Then Anconda prompt (start there), enter Conda install TensorFlow, enter Y, etc. will be installed.
3. Test TensorFlow New Pydev module, head up add #-*-coding:utf-8-*-, enter import TensorFlow as TF below, run without problem on installation succeeded.
4. Run down the official CNN mnist data, only to say that the CPU is very slow (6G), and so the money to change N card run down.
Note: TensorFlow Upgrade 1.0 will appear some bugs, here is a brief description of the solution.
1. Out Attributeerror:module ' pandas ' has no attribute ' computation '
Way:conda update dask
2.AttributeError: ' Module ' object has no attribute ' mul '
Way:tf.multiply替代tf.mul
3.AttributeError: ' Module ' object has no attribute ' sub '
Way:用tf.subtract替代tf.sub
4.AttributeError: ' Module ' object has no attribute ' neg '
Way:用tf.negative替代tf.neg
5. There are some other problems can be Baidu, not listed here. What to modify global variables (just different names)
Win7 64bit installation TensorFlow small essay