Sebastian Berg
Sebastian Berg
gh-28080 will remove any limitation from the iterator. Once that is done, one can look at removing the limitation from ufuncs as well (I am not sure if this issue...
I don't mind it, although I don't love that the `concate` signature would get very long, which does make a `broadcast_arrays` approach also appealing (but less obvious, but I am...
Updating branch to trigger CI. I _think_ RAPIDS dependencies should now be resolved sufficiently to actually get NumPy 2 (not sure about all runs, as torch may get in the...
Sorry, this comment was meant for the wholegraph PR. (Although it still means wholegraph doesn't matter for this one...
Just to note, the test failures seem all due to an out of memory. ``` terminate called after throwing an instance of 'rmm::out_of_memory' what(): std::bad_alloc: out_of_memory: RMM failure at:/opt/conda/envs/test/include/rmm/mr/device/pool_memory_resource.hpp:257: Maximum...
Seems those failures did not clear up, although it isn't clear to me why (or if) they are related to these changes.
Close/re-open to trigger CI. Were still related issues here or is it in principle good to go?
I had experienced similar failures that may have been due to gh-763. Of course gh-763 only makes sense if you have multiple projects using kvikio there and that seems slightly...
I think it would be completely fine to just backport the string promotion changes, they are trivial and fix the issue. The rest just fixes some even more niche things...
The PR probably didn't do all the changes, you liked. So feel free to open a new issue or PR about it.