Unable to deploy example ML model on RFSOC academia board

As in this announcement I tred to run git clone commands on jupyter terminal of RFSOC Academia board. but it gave fatal error: Unable to acess website certificate verification failed.
I am a rookie here, Is it alright to run the commands given in announcements on pynq board? PLease help me out here.

I’m assuming you don’t have internet access on the board. I also noticed your other post so I assume you are addressing it there:


I was able to run the git clone to download packages but rest all commands are not working, pip3 install pynq-dpu gives error:
ERROR: Could not install packages due to an OSError: HTTPSConnectionPool(host=‘files.pythonhosted.org’, port=443): Max retries exceeded with url: /packages/c3/53/e46c48a4aaee96e8d6db9e32614ae8e0733dd5ed8916e3e9a71e9d7b7bd1/pynq_dpu-1.4.0.tar.gz (Caused by SSLError(SSLCertVerificationError(1, ‘[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: certificate is not yet valid (_ssl.c:1108)’)))
Please guide me with the flow.

Hi there,

I just want to quickly point out that DPU-PYNQ does not support the RFSoC2x2 out of the box. You can pip install the package, but it won’t deliver the bitstreams – you have to generate these yourself. There are instructions how to add your own hardware design for your board on the DPU-PYNQ readme.

On the issue you’re seeing, looks like an SSL certificate problem, if you google around for keywords SSLError + pip, you can find some relevant posts on stackoverflow and the like.



Based on my current understanding, we use hardware platforms like zcu104 (MPSOC) and add vitis to it. Do I need to create base platform for RFSOC and then add vitis or just copying some files listed on DPY-PYNQ page enough which are for zcu104? Is there any pros/cons between these methods. I am totally new to these methods.
Also can you tell me after running make command what does the 3 files generated signify?

You need to provide a platform for the Vitis AI flow, which will add a DPU kernel to it.

It depends on what you want to do. If you just want the DPU running on your RFSoC2x2 board, the simplest way is likely to modify the part of the platform project tcl. And add additional folders for your board (alongside supported ones like ZCU104, KV260, etc…). You could copy the ZCU104 config as that DPU config should fit on the RFSoC part.

However, if you want to have the DPU running alongside your custom design, that might include RFDC to be able to pull in radio data, you will need to build your own platform with those IP included in it. There’s a tutorial on building a basic platform you could check out here.

The generated .xclbin is a necessary file for VART (Vitis AI Runtime) to work, it contains the bitstream as well, it gets stored in /usr/lib on the board if I remember correctly. We use the .bit and .hwh for the Overlay class, to load the design on the board.


I want something like BPSK/QPSK on hardware and I found designs wrapped as overlays.If I want to change parameters like the number of symbols transmitted per second, modify it to transmit images and other system configurations. Do I need vivado enterprise edition license and redesign the RF END for RFSOC and then create custom overlays followed by adding dpu to it? Sorry if this sounds more like a hardware design query.

You will certainly need a license for an RFSoC part (table 1 of ug973).

If DPU is what you want to use, you would have to build a platform design with your RF design first, then add the DPU using the Vitis flows I described earlier.

You could also potentially look at projects like FINN with Brevitas, that help you train and deploy a custom neural network model, but I am not quite as knowledgeable on those tools.