- Add ``bs.utils.apply_torch_normalisation`` wrapper function - Add ``bs.utils.apply_torch_dropout`` wrapper function - Add ``bias`` parameter to ``apply_torch_linear`` - Implement target layout specification for ``adapt_layout``
Fixed
- Fix ``scatter`` not passing ``dim``
1.9
Added
- Add ``dim`` parameter for ``gather``, ``scatter_reduce``, and ``build_distribution_layout`` functions
Fixed
- Add masks for reverse block index lookup, fixing potential memory leaks
1.8.3
Added
- Expose validation functions - Add ``bs.utils.apply_torch_linear`` function
Changed
- Move operations to ``bs.ops`` module - Move miscellaneous operations to ``bs.ops.misc`` module - Move experimental operations to ``bs.ops.experimental`` module - Rename ``bs.layout`` to ``bs.layouting`` - Rename ``bs.utils`` to ``bs.utils``
1.8.2
Added
- Add ``BlksprsTensor`` wrapper class to indicate block-sparse tensors
1.8.1
Fixed
- Fix ``build_sparsity_layout_fast`` not being exposed
1.8
Added
- Add validation for ``to_sparse`` for input dense tensors - Add alias for ``to_sparse`` and ``to_dense`` functions - Add documentation for ``repeat`` - Add gradient calculation for ``repeat_interleave``
Changed
- Change ``repeat_interleave`` to use same flow kernels as ``repeat`` does