Deep Learning Virtual Machine

Sets up a VirtualBox VM with Keras, TensorFlow and Theano installed. A Vagrant file is used to set up this VM, which runs on Ubuntu 14.04.

Getting Started

I assume you already have VirtualBox (version 5+) installed, if you don't, please download and install it.

1. Download and install Vagrant if you haven't previously done so.

2. Change into the deep-learning-vm directory start the VM by running:

   $ vagrant up

3. Once the vm is up run

   $vagrant ssh

4. Jupyter starts automatically. Open http://localhost:8100 on the host machine to see it.

What's Installed

• Deep Learning,
  • Keras, 1.0.5
  • TensorFlow, 0.10.0
  • Theano, 0.8.2
• Python, 3.5.2
  • H5py, 2.6.0
  • numpy, 1.11.1
  • pandas, 0.18.1
  • scipy, 0.18.1

In addition, an Ipython Jupyter notebook server is also installed. You can view it from the host's browser at http://localhost:8100.

Testing Keras

This VM comes with a few notebooks (with required datasets).

If you'd prefer the obligatory handwritten digits example,

$ vagrant ssh
$ curl -sSL https://raw.githubusercontent.com/fchollet/keras/master/examples/mnist_mlp.py | python

The script above, mnist_mlp.py, trains a simple deep neural network on the MNIST dataset. If all goes well, you should see a test accuracy close to 98.40% after 20 epochs.

Testing Theano (Optional)

Create a new notebook, and key in the following

import theano
theano.test()

then Run. It took ~2 hours for the 2995 tests to run in the VM on my laptop.