Adding (almost) all torchvision models to imgnet
Tested all the TV models. All of them work, except for squeezenet.
I added them to the supported models list in the imagenet train script.
I had to change some parts for inception as its preprocessing requires
special handling.
