This was a comment in our code that has been transferred to issues.

[sdake]

TODO: move this comment to an issue tracker, where alteast it will be tracked and can then

be removed from the code. I am fairly certain we will launch with CUDA 12.2.2.

NVIDIA CUDA-11.8's nvcc (parallel C compiler) does not integrate with gcc-12. For the moment

use gcc-11. CUDA-12.2 is compatible, so using that for now. Using gcc-11 is quite a challenge

facebookresearch/maskrcnn-benchmark#25 (comment).

#RUN apt install -y gcc-11
#RUN apt install -y g++-11
#RUN rm -f /usr/bin/gcc
#RUN rm -f /usr/bin/g++
#RUN ln -s /usr/bin/gcc-11 gcc
#RUN ln -s /usr/bin/g++-11 g++