Regarding builtin_function_or_method code

1

I want to see the code for nn.Dropout, while staying in google colab, I ctrl+click on nn.Dropout, it takes me to dropout.py, which has Dropout class, and returns

F.dropout(input, self.p, self.training, self.inplace)

so I ctrl+click on F.dropout, it takes me to functional.py
where something like this is returned

return (_VF.dropout_(input, p, training)
            if inplace
            else _VF.dropout(input, p, training))

now I am stuck, because this is builtin, and no option to go beyond this.
only way is to search for this in github.
this happens multiple times for a lot of builtin_function_or_method, for example searching for view, permute, all are builtin_function_or_method, so cannot access their implementation from colab notebook.

2

I don’t think there is an easy method using a Python notebook / REPL. :confused:
However, why would you like to see the underlying (optimized) C++ code?

4

to see how dropout is implemented.
this happens not only for dropout, but for other builtin also, for example torch.group_norm, I want to see how this is implemented, but it is builtin.