How to create block diagram using IP for Pynq?

[dustinjoe]

Hello. Thank you for this exciting project! I am trying to use the ip generated from a Keras model and want to use it in Vivado to generate block diagram for my Pynq-Z1 board. But not quite sure how to connect to the Zynq block. Wondering if I can get some help here?

Another little issue is that I am building a model arch that has more than one Conv1D, I found that if I inserted a dense between two Conv1D layers, the 'keras-to-hls.py' script cannot recognize the dimension of the second Conv1D correctly. But only one Conv1D or two consequtive Conv1D layers seem good.

[signorgelato]

Oh, you can just manually edit the 'keras-to-hls.py' file by changing current_shape at the right places.

[dustinjoe]

Thanks for your reply. I actually currently just use one conv1d and seems that the model is not that different.

[jmduarte]

Consider also using the new VivadoAccelerator backend #349, which specifically supports Pynq-Z2/Z1

@nicologhielmetti also has a tutorial which he will publish soon