Multiprocessing of models in C++

HI, I’m quite new to C++, but would like to rewrite my python code to C++, because of speed and better deployment.

Now what I do in Python I use multiprocessing, each process holds a model and I do inference on each model for the images that are loaded into a queue.

My question: can you guide me to how I would do that in C++ with Pytorch? What would be the best practices? Would multithreading be an option? Any pointers are appreciated!