Torch needs to adjust models for each dataset

[lukeyeager]

When using Caffe, DIGITS adjusts the last layer of the network to match the number of classes in the dataset. This is pretty easy, just layer[-1].num_output = num_classes.

It's more complicated in Torch because the network is defined in a piece of lua source code instead of a text file. We need to find some kind of solution for this. Maybe after main.lua reads the model, it can edit that object and adjust the output of the last layer?

[Sravan2j]

@lukeyeager : Currently, I am changing the output of the last layer, if and only if it is "Linear" layer. For all other layers, I am displaying warning message. This function can be improved in future to support all other layers. Please test it once and close issue if fix looks good. Thanks for your time.

[lukeyeager]

Closed in #324.