DenseNet with Keras crashes "floating point exception"

Hey guys,

So I’m trying to train Jeremey’s Keras DenseNet implementation on a simple 80,80 image dataset similarly to the CIFAR example done in class.
I’m using AWS GPU instance (Tesla K80) with Keras 2 and Tensorflow 1.1

Once I run the model.fit() my notebook kernel dies with:
[1] 2371 floating point exception (core dumped) ipython

Any ideas on how to go about this?
Should I try the PyTorch implementation ?