I’ve learned that PyTorch provides bindings for C++ and Python. After looking up the PyTorch documentation and various posts, I found that the two filespytorch/tools/autograd/gen_python_functions.py and pytorch/aten/src/ATen/native/native_functions.yaml are related to the bindings .
However, it appears that I need to manually inspect pytorch/aten/src/ATen/native/native_functions.yaml and *.h files to deduce the mapping between Python API and C++ implementation.
I’m curious if there’s a more automated method to obtain this mapping in PyTorch. For example, given torch.zeros_like , can we easily determine its implementation location in C++, such as pytorch/aten/src/ATen/native/TensorFactories.cpp ?