v0.5.3

What's Changed

• Fix issue with negative indexing in CatLinearOperator by @Balandat in #80
• Use absolute imports by @saitcakmak in #85
• Suppress linter warnings on import from non-existent file by @saitcakmak in #86
• Make pytorch a conda runtime dependency by @Balandat in #88
• Pin mpmath to <=1.3.0 by @Balandat in #94
• linear_operator cat_rows performance improvement by @naefjo in #93
• Fix errors in docs by @gpleiss in #96
• Typeguard 4.3.0 by @ChristianWirthContinental in #97
• Unify tolerance specification for test_diagonalization() by @Turakar in #91
• Fix dtype/device mismatch in _get_indices() by @Turakar in #90

New Contributors

• @naefjo made their first contribution in #93
• @ChristianWirthContinental made their first contribution in #97

Full Changelog: v0.5.2...v0.5.3

v0.5.2

What's Changed

• Fix type error when converting interpolated_linear_operator to double precision by @CY-Zhang in #74
• Update RTD configuration. by @gpleiss in #75

New Contributors

• @CY-Zhang made their first contribution in #74

Full Changelog: v0.5.1...v0.5.2

v0.5.1 (MaskedLinearOperator)

What's Changed

• Fix type of KroneckerProductLinearOperator.linear_ops (in some cases) by @Balandat in #66
• Add MaskedLinearOperator by @Turakar in #69
• Fix normalization of initial guess in linear_cg() by @Turakar in #71
• Add short path for LinearOperator.repeat() by @Turakar in #70
• Fix _t_matmul() in MaskedLinearOperator and add test by @Turakar in #72

New Contributors

• @Turakar made their first contribution in #69

Full Changelog: v0.5.0...v0.5.1

v0.5.0 (KernelLinearOperator)

What's Changed

• Add stricter type checking and expected matrix sizes by @corwinjoy in #42
• Re-remove @cached decorator for TriangularLinearOperator.to_dense by @Balandat in #59
• Make jaxtyping dep install requirement by @Balandat in #60
• Fix an issue for DiagLinearOperator if its _diag is a scalar tensor by @Balandat in #61
• Clean up deprecation warnings by @saitcakmak in #63
• Catch some warnings in linear_cg tests by @Balandat in #64
• Add KernelLinearOperator, deprecate KeOpsLinearOperator by @gpleiss in #62

New Contributors

• @corwinjoy made their first contribution in #42

Full Changelog: v0.4.0...v0.5.0

v0.4.0

What's Changed

• Allow .to() to operate on tensors by @Balandat in #45
• Change _dtype_value_context init args to *_value. by @Balandat in #47
• Fix missing dtype arg in _dtype_value_context by @saitcakmak in #49
• Use raw strings to avoid "DeprecationWarning: invalid escape sequence" by @saitcakmak in #52
• Remove @cached decorator for TriangularLinearOperator.to_dense by @Balandat in #51
• Format using ufmt, upgrade pre-commit config by @Balandat in #56

New Contributors

• @saitcakmak made their first contribution in #49

Full Changelog: V0.3.0...v0.4.0

v0.3.0

Merge pull request #39 from SebastianAment/permutation

Adding Permutation Linear Operators

v0.2.0 New torch api compatibility, bug fixes

What's Changed

Documentation

• Add basic example notebook by @Balandat in #17

New torch api compatibility

• Register torch.linalg.solve_triangular with LinearOperator by @Balandat in #29
• Register torch.isclose by @Balandat in #31
• Introduce reshape by @Balandat in #32

Performance

• Performant Scaling of BlockDiagLinearOperator by DiagLinearOperator by @SebastianAment in #14
• Edits to generic add, BlockDiagLinearOperator's matmul, and documentation by @SebastianAment in #10

Bug Fixes

• Fix linear operator mul torch.Tensor by @dannyfriar in #35
• Update LO.getitem to handle broadcasted tensor indices. by @gpleiss in #26
• adding unsqueeze to initial_guess in linear_cg when appropriate by @tvercaut in #22
• Removing deepcopy in add because of downstream failure in BoTorch by @SebastianAment in #12
• Patches for ConstantDiagLinearOperator._mul_constant and ConstantMulLinearOperator._getitem. by @j-wilson in #37

New Contributors

• @wbeardall made their first contribution in #9
• @tvercaut made their first contribution in #22
• @karstenBehrendt made their first contribution in #36
• @dannyfriar made their first contribution in #35
• @j-wilson made their first contribution in #37

Full Changelog: v0.1.1...v0.2.0

v0.1.1

Full Changelog: v0.1.0...v0.1.1

v0.1.0 (Initial Beta Release)

Welcome to LinearOperator!

This repo is in charge of what was previously known as LazyTensor in GPyTorch. It is currently a package that is predominantly used within GPyTorch, and assumes that the LinearOperators will be PSD kernel matrices. In the future, the goal is to make this a more generic implementation of LinearOperators that can be used in any number of applications.

What's new

For now, this implementation is very similar to what is currently in GPyTorch (v1.8.1). The only major difference is that *LazyTensor has been renamed to *LinearOperator.