✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: 145ba36. Link to the linter CI: here}

Tests fail on torch because it does not support negative indexing (data-apis/array-api-compat#144, pytorch/pytorch#59786) - it seems we could use flip but that creates a copy...?

Seems unlikely to change?

And from previous interactions with pytorch devs, my impression was that the assumption that slice steps are non-negative is baked too deep into the torch core to really hope it'll get fixed in torch itself.

Not sure if there is a standard workaround for this? Or do we need to re-write to avoid negative indexing? Or skip for torch...?

Edit: scipy uses flip, even though it creates a copy, but shouldn't be too bad in this case, and does not seem like there is anything else we can do so 🤷