I decide to install tensorflow and document all the process, from the installation to running some Machine Learning models.
Everything explained here is run under Ubuntu and with python 2.7. I download Anaconda Python since it install all the libraries necessary to run ML models (scipy, sklearn, numpy etc)
First go to Tensor Flow we are going to install Tensor Flow with CPU support because its easier. If you have a Nvidia GPU then you can install Tensor Flow with GPU support and accelerate the computations.
To install TensorFlow we open a terminal (ALT + CTL + T) and execute the command:
$conda create -n tensorflow
We are creating an enviroment, an enviroment is a directory with a collection of python packages. The idea of enviroments is to prove our programs with different versions of the package and maintaining previous versions.
We can switch enviroments using the command source activate. For example to use our new enviroment we use
$source activate tensorflow
If you then run the following command, you will se all the enviroments and the * marks the enviroment we are using
$conda-env list
Once we are in the new enviroment we install tensorflow with the command:
$pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.2.1-cp27-none-linux_x86_64.whl
After 3 minutes the installation is done. To check if everything is runnung run the following python script
import tensorflow as tf
hello = tf.constant('Hola, TensorFlow!')
sess = tf.Session()
print(sess.run(hello))
You should see Hello in the terminal. As simple as that!