Where is the `torch.empty` defined?

Question

(I am not a Python veteran so pardon me if this is a naive question.)

Following the PyTorch tutorial, below code works fine.

import torch
x= torch.empty(5,3) # <========= HERE
print(x)

But the Visual Studio Code complains that:

E1101:Module 'torch' has no 'empty' member

Eclipse PyDev has a similar warning.

So I use below code to check what torch offers:

for (k, v) in torch.__dict__.items():
    print(k, v)

And I find that it is a built-in:

empty

Since it is a method, I searched all the *.py file in the torch package folder for something like def empty and empty, no luck.

So where is this empty method defined? Some pyd/DLL?

How is it imported into the namespace?

Or put it another way, how to create a built-in function for my own package?

ADD 1 -- 10:38 AM 11/12/2018

Extending Python Built-ins with C/C++

I guess it's the torch::empty method. I have found a lot of matches in the github PyTorch codebase. But I haven't nailed the definition yet.