Dependency issue with old torch version

Hi everyone i’m new to the community. We have been working on this github project which is based on NSGA-net, for model compression. (refer : GitHub - mikelzc1990/nsganetv2: [ECCV2020] NSGANetV2: Evolutionary Multi-Objective Surrogate-Assisted Neural Architecture Search). I have tried to run the code following the instructions as given. But i seem to not able to get the correct versions. It uses an old torch version - PyTorch 1.5.1. These two libraries are also having problems: * [OnceForAll] 0.0.4,
, [timm] - 0.1.30. I tried using the version 1.13.0 and cuda10.2 as given by some people to whom i have asked before but this still seems to not work. When i use another version there seems to be updates regarding the functions and libraries used by the project, it is either replaced or removed completely in the new torch versions. So it throws various errors.
Could anyone please give some suggestions on how to work through this problem

The install instruction fails since torch==1.5.1 wasn’t built with CUDA 11.7, which was released later.
Could you share the errors you are seeing when updating to the latest stable or nightly PyTorch release?