PYNQ: PYTHON PRODUCTIVITY FOR ZYNQ

Doubt on overlays

I had a doubt , if i totally change the code behind the hardware architecture , can i still use it?
I have successfully built the pynq/lstm ip ,now i wish to add 2 more lstm layers by removing the output , softmax , concatenate etc layers present after the LSTM layer in the model. I would then change the weights in the modelfw_bw.hpp files.

I will also change the input since my problem is different.

So can i use the hidden layer of the lstm module and build over it as necessary?

You might be better posting a comment to the LSTM GitHub.

Cathal

Problem is they don’t care to reply.