pynq-z2

[gioregg]

Hi,

I have tried to implement a model on a zcu board but I run in some issues. After seeing your tutorial on part 7 I am thinking to switch to use the pynq-z2. Before doing so I would like to ask some questions regarding the general setup of the tutorial part 7.
After creating the ip block with axi stream by using the vivado accelerator, did you have to change some file inside the ip folder like configuration.xml or some other verilog or vhdl files?
When creating the tutorial of part 7 did you run into issue when running the nn.predict() in jupyter? ( I am asking this because with my last board I could not move further because the nn.predict() never terminated and was giving a dma error)
Did you use for the whole procedure vivado version 2020.1?
Which version of the pynq image did you use? Can I use the latest version(3.0.1) or do I need to use ealier version?

I hope to hear back from you.

Thank you,

Gio

[phucducnguyen]

Hi,
I used docker provided with the tutorial
docker run -p 8888:8888 ghcr.io/fastmachinelearning/hls4ml-tutorial/hls4ml-0.8.0-vivado-2019.1:latest
I did not change anything inside the ip folder.
For the image on Pynq-z2, I use this one here and set up normally
I faced one issue that was solved here
Edit: depend on the version, you can hit another problem relate to "The optimizer cannot recognize variable conv_0/kernel:0" which can be solved by changing Adam() to legacy.Adam() here