Antoni Parellada
Antoni Parellada

Reputation: 4791

Python 2, Python 3 and Anaconda on Windows 10. Now attempting to get Tensorflow to run

I have Windows 10 (64-bit computer) with Python 2. Here is the output on the cmd >:

Python 2.7.13 (v2.7.13:a06454b1afa1, Dec 17 2016, 20:42:59) 
[MSC v.1500 32 bit (Intel)] on win32

which is visible in C:\Python27.

In the same folder there is Python 3: C:\Python36-32.

In my path I have C:\ProgramData\Anaconda2\Scripts which seems to default to Project Interpreter: 2.7.13 (C:\ProgramData\Anaconda2\python.exe).

So after much hair pulling and retrofitting code in Python 3 (you can check my previous posts to get an idea), I get Python 2 up and running.

Only that what I am most interested in is Google's choice of Python for ML - in particular Tensorflow. And, alas, Tensorflow runs on Python 3!

No problem, right? After all I had downloaded and installed it in C:\... Wrong!

If I go to PyCharms and try installing Tensorflow after selecting Python3, this is what I get:

enter image description here

OK... So I go to the terminal prompt, and try there:

C:\Users\Toni>pip install tensorflow
Collecting tensorflow
  Could not find a version that satisfies the requirement tensorflow (from versions: )
No matching distribution found for tensorflow

But this may be because I have Python2?

So I go to Google, and find this promising post, which makes me think that going to cmd > and entering conda install python=3.5.0 will do the trick.

The problem is that if I run this, I am asked to un-install a bunch of stuff that I have painfully gotten to work, for example the autograd package...

Can I get some help as to how to get to run Tensorflow from this point in Dante's np.inferno?

Upvotes: 0

Views: 642

Answers (1)

P-Gn
P-Gn

Reputation: 24581

That is precisely to avoid those conflicts that conda has environments, in which you can install a fresh and separate conda distribution.

In your case, you could start creating an environment (named tensorflow here, could be anything else) specifically for tensorflow by calling

conda create -n tensorflow python=3.5

then put yourself in the environment with

activate tensorflow

and from there install tensorflow as you mentionned.

I must say that all of this is actually pretty well explained in tensorflow's installation tutorial.

Upvotes: 1

Related Questions