Darknet may have non-free CUDA support or language code in it.
There looks like there is non-free CUDA language
https://en.wikipedia.org/wiki/CUDA
support in
Package: darknet (0.0.0+git20180914.61c9d02e-2build4)
https://packages.trisquel.org/aramo/darknet
These are at
darknet-0.0.0+git20180914.61c9d02e/src/cuda.c
and
``darknet-0.0.0+git20180914.61c9d02e/src/cuda.h
in the source that apt-get downloads from Trisquels repository.
The https://packages.trisquel.org/aramo/darknet
also shows that this program is written in C and CUDA.
I do not yet know if there are many other CUDA parts in this, or if CUDA can be removed from this source in Trisquel's repository without breaking the ability to build Darknet.
https://trisquel.info/en/forum/why-doesnt-trisquel-forum-use-libre-chatgpt
and
https://trisquel.info/en/forum/darknet-floss-program
show more information about CUDA or/and CUDA support being in Trisquel's Darknet.
The https://www.gnu.org/distros/free-system-distribution-guidelines.html
shows in part
A free system distribution must not steer users towards obtaining any nonfree information for practical use, or encourage them to do so. The system should have no repositories for nonfree software and no specific recipes for installation of particular nonfree programs. Nor should the distribution refer to third-party repositories that are not committed to only including free software; even if they only have free software today, that may not be true tomorrow. Programs in the system should not suggest installing nonfree plugins, documentation, and so on.
and https://pjreddie.com/darknet/install/
shows information about "Compiling With CUDA"
The README.md file shows the
website, so this information, and it's CUDA support may "steer users towards obtaining any nonfree information for practical use" as the https://pjreddie.com/darknet/install/
page shows in part
Darknet on the CPU is fast but it's like 500 times faster on GPU! You'll have to have an Nvidia GPU and you'll have to install CUDA.
so this may "steer users towards" using non-free CUDA.
Using grep -r cuda
in a terminal also shows more CUDA things.
I do not know how many of these CUDA things are part of the non-free language or support non-free CUDA, as I do not know if these are also written in CUDA or somehow use CUDA.