How to make a sequential CNN+DNC?

[aryopg]

I'm struggling to understand how tensorflow graph work in this DNC implementation. I'm trying to achieve this:

INPUT --> CONV --> Fully connected layer (DNC) --> output

i'm trying to make them sequentially (not transfer learning whatsoever), so they will be updated together with the same backpropagation. Any suggestion? Or this implementation cannot be utilized to achieve my goal? Thanks

EDIT:
I'm also considering to make a stacked DNC, is that possible?