bfloat16 support and dimension order change

⚠️ BREAKING CHANGE (v0.3.0+): The output matrix dimensionality has changed from (n_tracks, batch_size, sequence_length) to (batch_size, sequence_length, n_tracks). This change was long overdue and eliminates the need for (potentially memory expensive) transpose operations downstream. If you're upgrading from an earlier version, please update your code accordingly (probaby you need to delete one transpose in your code).

✨ NEW FEATURE (v0.3.0+): Full bfloat16 support! You can now specify dtype="bfloat16" to get output tensors in bfloat16 format, reducing memory usage by 50%.

Update kvikio version dependency

We needed to add an upper bound to the kvikio version because of nvcomp deprecation. nvcomp did move to nvidia nvcomp but does not expose the same API. We need to find a way to still be able to use it.

v0.2.0

What's Changed

• Debug/default value error by @jorenretel in #18

Full Changelog: v0.1.5...v0.2.0

v0.1.5

What's Changed

• Adding feature to use custom "default_value" as filler for intervals … by @jorenretel in #17

Full Changelog: v0.1.4...v0.1.5

v0.1.4

Update readme

v0.1.3

fixing MANIFEST.in

v0.1.2

update changelog

v0.1.1

fixing pyproject.toml

v0.1.0

exclude c files in source dist

v0.0.3

Seeing whether I can build this with the dependencies.