fixed point deployment for pynq

[ryorise23]

Hi,

I am considering deploying a model quantized with fixed-point arithmetic using Brevitas onto a PYNQ-Z1 board. According to the following notebook, only integer operations are converted to hardware. Does this mean that it is impossible to implement operations involving decimals in hardware?
https://github.com/Xilinx/finn/blob/main/notebooks/advanced/4_advanced_builder_settings.ipynb

In my experiments, I quantized a model using Brevitas with Q3.3 and built it from the ONNX model to deploy on a PYNQ board. However, I observed that the accuracy was significantly lower than expected. Additionally, upon inspecting the driver.py file, I noticed that the data type of the input is UINT6. While the bit width is correctly recognized, it seems that the fractional part of the values is being ignored.

Thank you for your assistance.