Mario Lezcano Casado

After discussing with Nikita, we'll keep this PR simple, as it's very close to being mergeable, and then we'll put together a follow-up PR with the necessary ATen changes to...

@ngimel this one should be ready for review. As discussed above, this one implements a forward-looking `addcdiv`, implementing also the behaviour for integers with the correct division. We will submit...

My two cents from the PyTorch end of things. We implement the algorithm mentioned in the cited paper in PyTorch. It works fine for the first derivative, but for the...

I just had to choose one and I went for implementing the equivalent of `matmul` for vectors (as you cannot do this with `matmul`). Now, according to the specification, the...

You can indeed replicate the this behaviour resizing the inputs adding a dimension of size 1 in both inputs (see https://github.com/pytorch/pytorch/issues/18027) but this seems fairly clunky. There is no easy...

I think that we will be adding this behaviour as we add this function to `torch.linalg`. Note that the change in Torch would not be BC-breaking. We tend to be...

As discussed internally: > I suggested that `linalg.cross` supports broadcasting because in most (all?) linalg functions in PyTorch we allow broadcasting (e.g. in the solvers). Now, I don't feel strongly...

For what is worth, now PyTorch implements this behaviour. See https://github.com/pytorch/pytorch/pull/83798

I think that rather than blindly reverting, we should get to the root of the problem, as it is very weird to get such a regression when dispatching from a...

fwiw, I think this may be related to the open PR I have to avoid copies in matmul. Could you check whether https://github.com/pytorch/pytorch/pull/76828 fixes this? In any case, I'm fine...