Hi, I want some ops that is similar to combinatoric iterators in itertools. These functions do the same thing as in itertools. For example:

>>> A = torch.tensor([1,2,3])
>>> B = torch.tensor([4,5,6])
>>> cartesian_prod(A,B)
tensor([[ 1,  1,  1,  2,  2,  2,  3,  3,  3],
        [ 4,  5,  6,  4,  5,  6,  4,  5,  6]])
>>> combinations(A, 2)
tensor([[ 1,  1,  2],
        [ 2,  3,  3]])
>>> combinations_with_replacement(A, 2)
tensor([[ 1,  1,  1,  2,  2,  3],
        [ 1,  2,  3,  2,  3,  3]])

I already have a python implementation of these ops, supporting tensors of arbitrary rank, implemented using torch.unbind, torch.stack, and torch.cat. But this implementation is very slow, and is a bottleneck of my application.

So I'm just asking, is pytorch interested in supporting these natively? If yes, I'm happy to contribute.