Currently I understand most of the torch operators are registered in native_functions.yaml. With their implementations in folders like aten/src/Aten/native/, aten/src/Aten/native/cpu/ and aten/src/Aten/native/cuda* .
But I am still confused about HOW exactly PyTorch calls its own operators.
Now I have an image like this, which is from website.
Could anybody please show the specific code that are doing these dispatching things?