If someone has the model running already, could you please share the captions generated for the images mentioned in page 25 of http://cims.nyu.edu/~brenden/1604.00289v2.pdf .