View Torch model weights and activations

[lukeyeager]

When testing a Caffe model, I use the python bindings which loads the caffemodel in the same memory space as the webserver. Thus, it's trivial to return the weights and activations of the model as images on a webpage (code here).

For Torch, the testing is performed in a separate process. In order to view the weights and activations, we'll either have to:

1. Store them as temporary image files on disk (boo!)
2. Or return the data to the server through a pipe or socket

[lukeyeager]

Closed in #324.