Coming from c++, pip and python dependency management is the bane of my life. How do you make a python software leveraging pytorch that will ship as a single .exe and be able to target whatever gpu the user has without downloads?

Funnily enough a lot of the challenges in this particular case is related to PyTorch and CUDA being native code (mostly in C++). Of course combined with the fact that pip is not really adequate as a native/C++ code package manager.

Perhaps if C++ had a decent standardized package manager, the Python package system reuse that? ;p

just wait for next week and python will get a better package manager!