Thomas Huang

I think that the closest you can get to doing regression for traditionally classification based tasks is by binning the range of continuous values that you have into uniform size...

You're supposed to run `train_imagenet.sh` to begin training the model. If you look inside `train_imagenet.py` then you can see in lines 14-25 the pre-defined flags with their corresponding descriptions.

For python 3, I found a fix that worked for me on both prints: Change your sorted to `sorted(items, key=operator.itemgetter(1))` and it should sort by values and then print correctly....