Modifying `_modules` is equivalent to modify member variable in a model?

1 
from torch import nn

class MyModel(nn.Module):
    def __init__(self):
        super().__init__()
        self._add_module()

    def _add_module(self):
        self._modules['test'] = nn.Linear(3, 5)

m = MyModel()
print(m.test)

The result of the code above is exactly

Linear(in_features=3, out_features=5, bias=True)

I found this interesting. Why is this happening？