Mario Lezcano Casado

@ngimel the PR that fixed the warnings was already merged, this one is just concerned about avoiding copies. One of the cases where it elides a copy is when you...

yup, I think https://github.com/pytorch/pytorch/pull/76828 would fix the regression

Re. OOM. Wow, that's certainly unexpected. I'm not sure what's the best way to follow up on that. Probably @ngimel has a better idea how to proceed. Regardless, landing that...

I think the one that fixed the out version was https://github.com/pytorch/pytorch/pull/75197 and a previous one in that stack, but it may be the case that 75195 also fixed an out...

We haven't deprecated it. It's left there as an alias: ```c++ // Alias std::tuple slogdet(const Tensor& A) { return at::linalg_slogdet(A); } std::tuple slogdet_out(const Tensor& A, Tensor& sign, Tensor& logabsdet) {...

fwiw, these two tests should pass once https://github.com/pytorch/pytorch/pull/80217 is merged

or at least the second one. I'm not sure what's making the first one fail.

I'm on PTO, I'll look at this when I'm back. Feel free to activate the tests again, and I can skip them or fix them in the pr

It looks like that PR is about to land. Sorry for the very long wait, but I've been a fair amount of time out on PTO

It landed, it was reverted, then it landed again. Let's see if it sticks this time.