Where is `apply` defined in torch.autograd.Function?

autograd
1 
import torch
from torch.autograd import Function

class ExpTest(Function):
    @staticmethod
    def forward(ctx, i):
        print(ctx)
        result = i.exp()
        return result


x=torch.tensor([1,2,3.], requires_grad=True)
output = ExpTest.apply(x)
print(output.grad_fn)

# <torch.autograd.function.ExpTestBackward object at 0x7f0f376d84f0>
# <torch.autograd.function.ExpTestBackward object at 0x7f0f376d84f0>

As shown above, looks like the apply method first instantiated an ExpTestBackward object, then pass it to forward. but where is this apply defined? There is no such method in torch.autograd.function.py file

2

It’s defined in cpp, see python_function.cpp. The Function class in python inherits from the one defined in cpp.