rel-2.4.0

What's Changed

• Implement tril_indices, triu_indices, triu_indices_from and tril_indices_from by @zoj613 in #795
• Add NumPy RNG constructor Ops by @kc611 in #789
• Do not allow PatternSub to replace nodes with different number of outputs by @ricardoV94 in #803
• Require independent dimensions in size arguments to multivariate RandomVariables by @brandonwillard in #446
• Temporarily pin to scipy<1.8.0 by @michaelosthege in #811

Full Changelog: rel-2.3.8...rel-2.3.9

rel-2.3.8

What's Changed

• Support n_steps=0 (and empty sequence) case for Scan by @rlouf in #741
• Clean up Scan's Cython implementation by @brandonwillard in #774
• Remove broken logger warning by @brandonwillard in #781
• Update the high-level API documentation by @rlouf in #783
• Fixed behaviour of at.subtensor.take when axis is None by @kc611 in #787

Full Changelog: rel-2.3.7...rel-2.3.8

rel-2.3.7

What's Changed

• Add a Numba implementation for the BroadcastTo Op by @kc611 in #769
• Improve the performance of Numba-compiled array reduction Ops by @kc611 in #599

Full Changelog: rel-2.3.6...rel-2.3.7

rel-2.3.6

What's Changed

• Add Numba cache configuration by @kc611 in #639
• Fix bug in local_elemwise_alloc and add tests by @ricardoV94 in #767

Full Changelog: rel-2.3.5...rel-2.3.6

rel-2.3.5

What's Changed

• Make TensorConstants print their names when specified by @oscargus in #719
• Do not use old T alias in documentation. by @oscargus in #723
• Updated versioneer to enable Python 3.11 and removed some Python 2 and 3.6 support. by @oscargus in #721
• Convert lists to tuples by @oscargus in #722
• Tests for input checks in tensor/signal/pool.py by @aerubanov in #713
• Fix limit of log1mexp gradient at zero and improve numerical stablity by @ricardoV94 in #725
• Change FunctionGraph.change_input to change_node_input by @anooptp in #734
• Remove optimize_graph_with_cache by @MridulS in #735
• Replace aet aliases with at by @twiecki in #728
• Remove useless BroadcastTo Ops by @brandonwillard in #740
• Introduce aesara.as_symbolic by @brandonwillard in #743
• Add a sum method to _sparse_py_operators by @aerubanov in #745
• Enable complete static shape information at the type level by @brandonwillard in #711
• Add __pycache__, _build, and *.pkl to .gitignore by @brandonwillard in #756
• Remove PyPolyaGamma by @brandonwillard in #761
• Reintroduce deprecated broadcastable keyword arguments by @brandonwillard in #763
• Fix bad Windows free in dimshuffle.c by @brandonwillard in #762

New Contributors

• @aerubanov made their first contribution in #713
• @anooptp made their first contribution in #734
• @MridulS made their first contribution in #735

Full Changelog: rel-2.3.4...rel-2.3.5

rel-2.3.4

What's Changed

• Add missing Constant validation to OpFromGraph constructor by @brandonwillard in #704
• Fixed issue with NumPy 1.22 by @oscargus in #716

New Contributors

• @oscargus made their first contribution in #716

Full Changelog: rel-2.3.3...rel-2.3.4

rel-2.3.3

What's Changed

• Generalize and update the JAX Op conversion docs by @brandonwillard in #696
• Documentation refactoring and new unification and miniKanren sections by @brandonwillard in #697
• Update the articles for extending Types by @brandonwillard in #698
• Add axis argument to Softmax and related Ops by @ricardoV94 in #673
• Adding support for nightly builds by @dfm in #656
• Split solve_triangular functionality from the Solve class into separate SolveTriangular class by @fshart in #676
• Use NumPy C API to perform DimShuffle steps in its C implementation by @brandonwillard in #701

New Contributors

• @fshart made their first contribution in #676

Full Changelog: rel-2.3.2...rel-2.3.3

rel-2.3.2

What's Changed

• Remove TensorConstant.tag.unique_value by @Carlosbogo in #650
• Add JAX dispatch for Psi Op by @ricardoV94 in #683
• Propagate RandomVariable names in local_dimshuffle_rv_lift by @ricardoV94 in #503
• Add a pretty printer for Constant and keywords to FunctionPrinter by @brandonwillard in #681
• Update local_subtensor_make_vector and tests by @brandonwillard in #675
• Add an alternative, faster AdvancedIncSubtensor Op by @brandonwillard in #561
• Make local_optimizer automatically validate nodes using its tracks argument by @brandonwillard in #688
• Replace aesara.graph.unify with logical-unification and add a KanrenRelationSub local optimizer by @brandonwillard in #523
• Fix a RandomVariable DimShuffle lift case for empty sizes by @brandonwillard in #693
• Fix broadcastables inference for Alloc and an as_tensor_variable bug when all ndim=0 and all dimensions are broadcastable by @brandonwillard in #694

New Contributors

• @Carlosbogo made their first contribution in #650

Full Changelog: rel-2.3.1...rel-2.3.2

rel-2.3.1

What's Changed

• Do not add SpecifyShape to constant sizes in RandomVariable by @brandonwillard in #680

Full Changelog: rel-2.3.0...rel-2.3.1

rel-2.3.0

What's Changed

• Add shape assertion to local_useless_alloc by @brandonwillard in #647
• Update the developer start guide by @brandonwillard in #652
• Remove introductory sentences from dev_start_guide.rst by @brandonwillard in #653
• Fix a file extension typo in the dev guide by @brandonwillard in #654
• fixed typo on tutorial --extending aesara by @jguerra-astro in #663
• Refactor get_scalar_constant_value by @brandonwillard in #643
• Make input mismatch TypeError in make_node more readable by @ricardoV94 in #655
• Update Scan optimizations and fix Scan + RandomVariable shape inference issues by @brandonwillard in #635
• Fix issue with version logic when tags are not available by @brandonwillard in #670
• Made the RandomState conversion of Numba a valid RandomStateType by @kc611 in #667
• Add a CheckAndRaise COp by @brandonwillard in #672
• Make ScalarConstant a subclass of ScalarVariable by @brandonwillard in #678

New Contributors

• @jguerra-astro made their first contribution in #663

Full Changelog: rel-2.2.6...rel-2.3.0