Sebastian Berg
Sebastian Berg
Those exist now.
Hmmm, fun CI failures. The MacOS ones are spurious matmul runtime warnings, the windows ones looked like new warnings generated by cython code. Retriggered, but just out of curiosity...
Sorry, nvm. would need a re-base for the new allowed warnings. And that is just not worth it.
> the preference is to require eigenvalues to be always of a complex dtype. Grumpy old man hat, but: If the NumPy issue isn't resoelved (and it possibly cannot, at...
It was discussed as a class attribute and not a method, since there should never be a reason to pass any arguments. That said now that you mention it, a...
True, but since it is a static object (and has to be as per standard), you would be able to cache it anyway.
> Is that true? Classes can override the method per-instance. So we cannot actually cache per-type if it is a method. We specified that it *must* be cacheable and the...
@junrushao I think this moved quickly with (to *me*!) unclear (experimental) implementations or experience with it. And I think this request to change something quite central shows that a bit...
Maybe we can deprecate this? The `try/except` seems rather broader here than is nice for the sake of matrices that also seem like a rather niche case for being passed...
This needs new logic I would imagine. My understanding is that `...` is implemented as indicated _broadcasting_ dimensions and while this overloads it as an arbitrary number of reduce dimensions...