Do I need to install `llibtorch` to load my triained model with `torchscript` in C++?

I have read the tutorials here and here.
I know that I can save my pytorch model with torchscript, but If I want to embed my trained model in my C++ program, I need to have libtorch installed in my own program in order to use torchscript api, which sounds not reasonable, since my own program only has 100m size big.
Is there any solution for this?