rel-2.36.1

What's Changed

New Features 🎉

• Added PyTorch clip dispatch by @fonnesbeck in #1797
• Implement pack/unpack helpers by @jessegrabowski in #1578

Bugfixes 🐛

• Allow numba cache path to create parent directory by @ricardoV94 in #1798
• Fix output ordering in new Scan API by @ricardoV94 in #1799

Maintenance 🔧

• Revert accidental structured sparse changes by @ricardoV94 in #1800

Full Changelog: rel-2.36.0...rel-2.36.1

rel-2.36.0

What's Changed

Major Changes 🛠

• Support Python 3.14 by @Armavica in #1661
• Numba full backend support and required dependency by @ricardoV94 in #811
• Start deprecating shared updates functionality in Scan by @ricardoV94 in #1704
• Reorganize the sparse module by @jessegrabowski in #1674
• Remove tensor/io.py by @ricardoV94 in #1766
• Remove InRange scalar Op by @Copilot in #1699

New Features 🎉

• Add more MLX dispatches by @jessegrabowski in #1684
• Add linalg Ops to MLX backend by @cetagostini in #1700
• Reliable caching of Graphs and individual Ops in numba backend by @ricardoV94 in #1637
• Support TypedList on Numba backend by @ricardoV94 in #1745
• Implement numba dispatch for ScalarLoop by @ricardoV94 in #1445
• Implement Convolve2D Op by @jessegrabowski in #1397
• Numba AdvancedIndexing: Complete support for integer advanced indexing by @ricardoV94 in #1778
• Implement Scan based filter helper by @emekaokoli19 in #1717
• Support Scan with MIT-MOT in JAX backend by @ricardoV94 in #1651

Bugfixes 🐛

• Ignore axis argument in numba CumOp when input is 1d by @jessegrabowski in #1691
• Fix pytensor_cache clear by @ricardoV94 in #1694
• Fix 0-sized CGemV with unitialized memory by @ricardoV94 in #1719
• Handle non-constant NoneTypeT variables by @ricardoV94 in #1728
• Fix numba cache bugs and suppress noisy warnings by @ricardoV94 in #1731
• Fix shape error formatting in C impl of Elemwise by @ricardoV94 in #1749
• Numba Pow: Fix failure with discrete integer exponents by @ricardoV94 in #1758
• Handle upcasting of scalar to vector arrays by scipy vector optimizers by @ricardoV94 in #1768
• Validate compatible linker in Scan Python/C backend by @ricardoV94 in #1757
• Numba and JAX Scan fixes by @ricardoV94 in #1754
• Handle F-order and arbitrary index ndim in numba UnravelIndex and RavelMultiIndex by @ricardoV94 in #1770
• Misc numba fixes by @ricardoV94 in #1771
• Numba Dot: Handle complex inputs by @ricardoV94 in #1776
• Numba RavelMultiIndex: Fix scalars with clip mode by @ricardoV94 in #1775
• Fix Elemwise inside OpFromGraph in numba mode by @ricardoV94 in #1774
• Avoid runtime broadcast error due to dot_to_mul rewrite by @ricardoV94 in #1783
• Optimize: Handle gradient wrt scalar inputs and guard against unsupported types by @ricardoV94 in #1784
• Numba CAReduce: respect acc_dtype by @ricardoV94 in #1773
• Fix wrap_jax when there is a mix of statically known and unknown shapes by @ricardoV94 in #1722

Documentation 📖

• Fix latex formatting in VJP autodiff example by @williambdean in #1726

Maintenance 🔧

• Rename sparse functions to match numpy array API by @jessegrabowski in #1663
• Remove scalar_ prefix from several Ops by @Copilot in #1683
• Add __rtruediv__ and __rfloordiv__ to Scalar variables by @Copilot in #1701
• Fix Blockwise vmap dispatch for no batch dimensions by @cetagostini in #1705
• JITLinker improvements by @ricardoV94 in #1746
• Numba fallback to object mode in more cases by @ricardoV94 in #1747
• Remove transfer_type and default inplace Elemwise Ops by @ricardoV94 in #1752
• Speedup RNG copy in non-mutable RVs by @emekaokoli19 in #1721
• Sort and Argsort: Check axis are integers by @ricardoV94 in #1769
• Numba do not cache inplace list ops by @ricardoV94 in #1767
• Handle mixed input dtypes and empty size in numba lapack functions by @ricardoV94 in #1764
• Rewrite concatenate([x, x]) as tile by @tchan102 in #1714

New Contributors

• @cetagostini made their first contribution in #1700
• @tchan102 made their first contribution in #1714
• @emekaokoli19 made their first contribution in #1717
• @Mr-Imperium made their first contribution in #1748

Full Changelog: rel-2.35.1...rel-2.36.0

rel-2.35.1

What's Changed

Major Changes 🛠

• Handle axis=None symbolically instead of within CumOp by @Copilot in #1574

New Features 🎉

• Refactor Expm and add jax dispatch by @jessegrabowski in #1668

Bugfixes 🐛

• Fix C-cache bug related to input order of nominal variables by @ricardoV94 in #1673
• Don't return tuple from jax.scipy.linalg.qr when mode = 'r' by @jessegrabowski in #1682

Maintenance 🔧

• Move shared HStack and VStack methods to Stack superclass by @jessegrabowski in #1662
• Fix MKL BLAS detection with renamed libiomp5md.lib by @maresb in #1667
• Cleanup numba dispatches by @ricardoV94 in #1648
• Use incsubtesor in gradient of repeat by @ricardoV94 in #1621
• Suppress numpy warnings in python implementation of Elemwise by @ricardoV94 in #1681

Full Changelog: rel-2.35.0...rel-2.35.1

rel-2.35.0

What's Changed

Major Changes 🛠

• Drop Python 3.10 and NumPy < 2 by @Armavica in #1253
• Remove sparse.sandbox by @jessegrabowski in #1664

New Features 🎉

• Add MLX backend by @williambdean in #1365

Bugfixes 🐛

• Fix rewrite weakref leak by @ricardoV94 in #1660

Documentation 📖

• Fix issue #1554: Preserve custom thumbnails in gallery script by @Copilot in #1649

Maintenance 🔧

• Use LAPACK functions for cho_solve, lu_factor, solve_triangular by @Fyrebright in #1605

Full Changelog: rel-2.34.0...rel-2.35.0

rel-2.34.0

What's Changed

Major Changes 🛠

• Implement wrap_jax and rename as_op to wrap_py by @aseyboldt in #1614

New Features 🎉

• Add logsumexp to xtensor by @zaxtax in #1629

Bugfixes 🐛

• Fix flip to handle negative axis correctly by @Copilot in #1628
• Fix FusionOptimizer bug by @ricardoV94 in #1630

New Contributors

• @Copilot made their first contribution in #1628

Full Changelog: rel-2.33.0...rel-2.34.0

rel-2.33.0

What's Changed

Major Changes 🛠

• Remove unused graph rewriting functionality by @ricardoV94 in #1595
• Speedup graph traversal functions by @ricardoV94 in #1596

New Features 🎉

• Implement OpPattern for more flexible tracks and PatternNodeRewriter by @ricardoV94 in #1594

Bugfixes 🐛

• Fix failure with narrowing of ssize_t to int in macos by @ricardoV94 in #1602

Documentation 📖

• Add gallery notebook on vector jacobian products by @ricardoV94 in #1544
• Normalizing Flows in PyTensor Notebook by @juanitorduz in #1588
• Add intro notebook from workshop by @juanitorduz in #1585

Maintenance 🔧

• Add rewrite for log(sqrt(x)) by @williambdean in #1555
• Tweak rewrites in infer_shape dataset by @ricardoV94 in #1563
• Add rewrite for softplus(log(x)) -> log1p(x) by @lciti in #1452
• Implement static shape inference for AdvancedSubtensor by @ricardoV94 in #1566
• Rewrite more cases of Blockwise IncSubtensor by @ricardoV94 in #1560
• Skip scan rewrites if there is no Scan Op in the graph by @ricardoV94 in #1600
• Add compiler_verbose config flag by @ricardoV94 in #1520
• Use vectorized jacobian in Minimize Op by @Michal-Novomestsky in #1582
• Speedup FusionOptimizer by @ricardoV94 in #1615

New Contributors

• @jorenham made their first contribution in #1597
• @Michal-Novomestsky made their first contribution in #1582

Full Changelog: rel-2.32.0...rel-2.33.0

rel-2.32.0

What's Changed

Major Changes 🛠

• Improve dot lift rewrites by @ricardoV94 in #1471
• Change Dot Op to only accept matrix inputs by @ricardoV94 in #1538

New Features 🎉

• Refactor and update QR Op by @jessegrabowski in #1518
• Optimize matmuls involving block diagonal matrices by @jessegrabowski in #1493

Bugfixes 🐛

• Fix CheckAndRaise Op C implementation by @ricardoV94 in #1521
• Fix mean, var and std for XTensorVariables by @ricardoV94 in #1533
• Fix AdvancedIncSubtensor1 C-compilation with empty indices by @ricardoV94 in #1551

Documentation 📖

• Fix missing xtensor.type cross-reference targets by @OriolAbril in #1530

Maintenance 🔧

• Converts negative constant axis to positive if present in Join(COp) by @PabloRoque in #1527
• Preserve static shape information in block_diag by @jessegrabowski in #1529
• Get rid of pydot-ng because it's deprecated since 7 years by @maresb in #1534
• Canonicalize subtensor negative integer indices by @ricardoV94 in #1541
• Use stricter numerical tolerance in rewrites and allow casting in PatternNodeRewriter by @lciti in #1526

New Contributors

• @PabloRoque made their first contribution in #1527

Full Changelog: rel-2.31.7...rel-2.32.0

rel-2.31.7

What's Changed

New Features 🎉

• Implement full_like, ones_like, and zeros_like for XTensorVariables by @ricardoV94 in #1519

Documentation 📖

• Add xtensor docs by @ricardoV94 in #1504

Maintenance 🔧

• Extend C implementation of psi to support negative inputs by @Abhinav-Khot in #1523
• Make convolve mode symbolic to avoid unnecessary large convolution in gradient graph by @ricardoV94 in #1522

Full Changelog: rel-2.31.6...rel-2.31.7

rel-2.31.6

What's Changed

New Features 🎉

• Implement broadcast for XtensorVariables by @ricardoV94 in #1513

Bugfixes 🐛

• Do not skip validation between consecutive Elemwise inplace replacements by @ricardoV94 in #1494
• Fix shape errors in scalar_solve_to_division rewrite by @ricardoV94 in #1516

Full Changelog: rel-2.31.5...rel-2.31.6

rel-2.31.5

What's Changed

New Features 🎉

• Implement vectorized jacobian and allow arbitrary expression shapes by @ricardoV94 in #1228
• Rewrite scalar solve to division by @jessegrabowski in #1453

Bugfixes 🐛

• Fix xtensor (...,) Transpose by @ricardoV94 in #1500

Maintenance 🔧

• Small fixups to xtensor type and XRV by @ricardoV94 in #1503

Full Changelog: rel-2.31.4...rel-2.31.5