Bug fixes for unflatten conv2d and keras to hls conversion
Created by: benjaminkreis
A few fixes for bugs uncovered trying to Csim @pierinim's model:
- gets around appended paths to weights in h5 file when using jupyter
- fix typo in unflatten
- fix some bugs in hls-writer for choosing input names and types
We can Csim @pierinim's "NoConcatenate" model now, but Csynth will require the sublayer functionality.