-
Notifications
You must be signed in to change notification settings - Fork 0
Installation and Usage
This page should give you some basic guidelines about how to get started.
- Install Python 3 (3.6.5): Don't install Anaconda or whatever. Theoretically, Anaconda gives you many advantages, but in practice it leads to lots of conflicts with differing python versions. Install Python 3 on your system. On mac, I keep the installation of Python 3 separate from 2 because that's what the system uses. Maybe you'll have to do that too.
- Install Jupyter (notebook frontend),
ipython(nice interactive python shell) - Once python is installed, you can use the following to install packages.
sudo pip install PACKAGE_NAME
Packages you'll need are:
- Tensorflow: They recommend installing TensorFlow in a VirtualEnv to keep it even more separate from your system. I haven't done that (and the VirtualEnv stuff has caused me a lot of headache for other things in the past), but maybe it's a good idea
- Keras
That's what you'll need to get started. These will install many other packages as dependencies, like numpy, scipy, and so on. If that doesn't happen automatically, you'll have to read errors and do them one by one.
Facebook, Microsoft, Google, and (I think) Amazon have all released open source frameworks for Deep Learning. Keras is an open-source interface to these libraries, so theoretically, we can switch the lower level framework later. But, the maintainer of Keras, an opinionated engineer called Francois Challet, now works at Google, so who knows?
Anyway, Tensorflow vs Keras vs Theano vs the-other-ones is not a debate that is important to us yet. So we choose Tensorflow. Here is a detailed comparison for when we need to make a choice.