Support for Numpy 2.0 - Githubissues

nvoxland-al commented 2 weeks ago

🚀 🚀 Pull Request

Impact

[X] Bug fix (non-breaking change which fixes expected existing functionality)
[ ] Enhancement/New feature (adds functionality without impacting existing logic)
[X] Breaking change (fix or feature that would cause existing functionality to change)

Description

Numpy 2.0 changes some of it's behavior in ways that are not compatible with the existing deeplake code.

This PR updates deeplake to be compatible with both 1.x and 2.0

Things to be aware of

Numpy removed support for can_cast() between python types and numpy types as part of https://numpy.org/neps/nep-0050-scalar-promotion.html . This PR makes deeplake's cast checks still support a comparison between a python sample and a dtype, but follows the new numpy standard of not taking the value into account.

Because python ints/floats are np int64/float64s, this means that you can no longer add a python int to an int32 tensor -- you must explicitly create it as an np.int32.

This better matches other codepaths deeplake already has, so is bringing more consistency. But, there may be cases where doesn't work with existing scripts.