Evgeni Burovski

Updated to allow `default_device()` to be None, according to the resolution in https://github.com/data-apis/array-api/issues/923

Cross-ref https://github.com/data-apis/array-api-strict/issues/134

Agreed, the 2024 spec clarification is pretty clear on arrays of different dtypes. There's also gh-339 which report issues for mixing float32 and float64. Here's a fix: gh-360

It seems that the issue is from using `cdef numpy.ndarray` typing. Switching over to typed memoryviews seems to work fine. Changing the OP sample into ``` # cython: language_level=3 #...

FWIW the RuntimeError of https://github.com/data-apis/array-api/issues/819#issuecomment-2229503135 seems to be a torch bug / artifact of incomplete uintN (N> 8) support. (until very recently, torch only had uint8). With uint8, pytorch behavior...

> JAX ... uses None to represent the concept of uncommitted placement on the default device. This is one example of thea library specific device placement rule, I'd say :-)....

IMO it's a no-brainer. No matter whether it's "main" or "master", the cognitive load of a single change is trivial; what gets annoying is having to switch between one and...

managed to catch it on numpy 2.2. A fix in gh-355

Reopening: as discussed in gh-355, that fix may fix a different issue. So let's keep this open until we can repro and fix this specific issue.

Looking at `test_std`, https://github.com/data-apis/array-api-tests/blob/master/array_api_tests/test_statistical_functions.py#L262, it does not seem to attempt any value testing. Then what is flaky, `assert_dtype` or `assert_keepdimable_shape`?