I’m trying to use a precompiled library for custom implementations of standard routines. I’ve been looking on the internet but have not had much luck finding documentation. Does anyone know how to start/ where to look for more information?
What kind of library are you looking for and how does PyTorch come into play here? 
Could you explain your use case a bit more, please?
I am building library to execute deep learning models with limited memory. This library has custom implementations of standard routines(forward_pass, etc) written in C++. I have attempted to extend the files as explained from this tutorial. However, I have been having issues with dependencies between the library and Pytorch. As an alternate approach, I’m trying to utilize the library precompiled. I know that Pytorch uses the cuDNN library for CUDA. Is it possible to replicate that process for a custom library? I hope that shed some light on my question.
Thank you for your help!