We are writing libtorch bindings for Elixir’s Nx - thank you for making it easily accessible - and we have noticed the precompiled libtorch archives, for both Linux and Mac, are missing the file “include/tensor/tensor.h”.
Am I missing some crucial step in setting this up? For now we are using aten.
EDIT: We are using a Makefile to build libtorch and perhaps that’s the root cause? We should be using CMake to build certain headers for us?