Quantization approaches and good to knows

[RicDen]

Hello,

I am aiming to deploy and benchmark various models with FINN and I have trained them in standard pytorch.
However, as far as I understand, I would have to rewrite the networks in Brevitas and include quantization layers, training, etc to be compatible with FINN. Maybe I also have misconceptions about this workflow and make my work more complex than needed.
Currently, this seems to be a significant effort besides the hardware deployment, and my focus is on the hardware deployment. Are there any best practices or starting points to handle the transfer into Brevitas with maximum benefits in FINN the best way?

Any input, references, repositories, or other forms of suggestions would be highly appreciated!
Thank you so much in advance!