Evgeni Burovski
Evgeni Burovski
FWIW I don't consider it a problem. We run doctests on the scipy documentation, and docstrings of private functions are not a part of the SciPy documentation. We have never...
¯\_(ツ)_/¯ I personally am -0 (== not going to work on this myself, not going to block it if there's a motivated individual). The first step IMO is to undo...
> @ev-br I have an editable install and can't use dev.py. Is there another way run smoke-docs? Sure! For a single .py file it's just (the path is the exact...
> But the whole scipy command doesn't work for me; I get failures during collection with or without --doctest-collect=api. You likely need a bunch of `--ignore=...` which `dev.py` fills in...
Would be great to report these to pytorch itself. Meanwhile we'll need a workaround in `-compat` indeed.
cross-ref https://github.com/data-apis/array-api-compat/pull/272 and https://github.com/data-apis/array-api-tests/pull/348 The bottom line seems to be that that in torch accepting scalars is more of an exception than a rule. Not sure who to ping, sadly,...
Testing shows 16 missing kernels, all xfailed for now: https://github.com/data-apis/array-api-compat/blob/1.11.2/torch-xfails.txt#L139
A gentle ping. While this change is of course totally not critical, it will be helpful for downstream (scipy, at least) to simplify the internals of delegation to `cupyx.scipy`, https://github.com/scipy/scipy/blob/main/scipy/ndimage/_support_alternative_backends.py#L41
Rebased on main to resolve the conflict.
Pushed an update with an attempt to fix the missing import, could you restrart the CI please?