Mario Lezcano Casado

oof, it even has its own index dispatcher. What an absolute mess of a function.

Can't you test ti through `nll_loss_nd`, as it calls into it?

So, at the moment, the test calls `nll_loss` once, so just one of the implementations must be covered by the test, right? We would need a second call in the...

> Well, so it has the index dispatcher for different dtypes of the class input, which does not seem unreasonable (the alternative being to always cast the index tensor before...

You are right. There are two variables here`C` and the indexing into `input` (assuming `input` of size `[N, C]`). `C` controls the values that may appear within `target`. The precondition...

Yes it will. Now, regardless of `torch.lstsq`, we should investigate why is `linalg.lstsq` nondeterministic. We can track that issue in the other issue you opened.

@pytorchbot merge -l

@pytorchbot merge

@ngimel could you approve this one?

That's a good question, I have no idea. I'll have a look at that tomorrow.