rel-2.11.0

What's Changed

Major Changes 🛠

• Refactored RV _print_name to long names by @shreyas3156 in #244
• Rename sgn to sign by @sudarsan2k5 in #228

New Features 🎉

• Fix Scan JAX dispatcher by @ricardoV94 in #232
• Optimize while scans when only last state is needed by @ricardoV94 in #216
• Allow string keys in eval utility by @Raj-Parekh24 in #242

Bugfixes 🐛

• Fix scan save memory rewrite bug by @ricardoV94 in #236
• Fix non-contiguous reshapes in numba backend by @aseyboldt in #255
• get_scalar_constant_value now raises for non-scalar inputs by @shreyas3156 in #248
• Raise NotImplementedError when trying to convert MaskedArrays by @ricardoV94 in #260
• Fix bug in gradient of set_subtensor by @ricardoV94 in #267

Documentation 📖

• Fix a typo in Graph Rewriting Doc by @shreyas3156 in #261

Maintenance 🔧

• Fix numpy DeprecationWarning when converting integers to PyTensor Constants by @ricardoV94 in #238
• Add explicit support for Python 3.11 by @ricardoV94 in #198
• 🔄 From Aesara: 1337: Add basic overloads for Numba sparse types by @twiecki in #102
• Bump pypa/gh-action-pypi-publish from 1.8.1 to 1.8.3 by @dependabot in #257

New Contributors

• @shreyas3156 made their first contribution in #244
• @Raj-Parekh24 made their first contribution in #242

Full Changelog: rel-2.10.1...rel-2.11.0

rel-2.10.1

What's Changed

Bugfixes 🐛

• Fix spurious warning from FusionOptimizer by @ricardoV94 in #225

Full Changelog: rel-2.10.0...rel-2.10.1

rel-2.10.0

What's Changed

Major Changes 🛠

• Drop support for Python 3.7 by @ricardoV94 in #210

New Features 🎉

• Extend support of RandomVariables in JAX backend by @ricardoV94 in #200
• Rewrite products of exponents as exponent of sum by @tamastokes in #186
• Propagate static shape in MaxAndArgmax by @covertg in #206
• Implement JAX dispatch for Split by @ricardoV94 in #209
• Fuse Elemwise graphs that have multiple outputs and clients by @ricardoV94 in #121

Bugfixes 🐛

• 🔄 From Aesara: 1345: Fix the JAX Subtensor and IncSubtensor dispatcher by @ricardoV94 in #211
• Fix numba implementation of empty DimShuffle by @aseyboldt in #218

Maintenance 🔧

• Fix math typo in sigmoid by @SangamSwadiK in #202
• Bump pypa/cibuildwheel from 2.11.4 to 2.12.0 by @dependabot in #197
• Fix stack type hints by @covertg in #205
• Don't cancel tests in CI early by @aseyboldt in #147
• Add pyupgrade to pre-commit and apply it by @Armavica in #213
• Fix DeprecationWarning from scipy.sparse.base by @covertg in #214
• Upgrade pre-commit tools and apply black format by @Armavica in #217

New Contributors

• @SangamSwadiK made their first contribution in #202
• @dependabot made their first contribution in #197
• @covertg made their first contribution in #205
• @tamastokes made their first contribution in #186

Full Changelog: rel-2.9.1...rel-2.10.0

rel-2.9.1

ℹ️ Versions 2.8.12 and 2.9.0 did not get released on PyPI or conda-forge due to hiccups in our CI pipeline. The corresponding release notes were merged into the seconds below.

What's Changed

Major Changes 🛠

• [from 2.8.12] Allow passing static shape to tensor creation helpers by @ricardoV94 in #118
  • The first and only positional argument to tensor is now name. dtype and shape are now keyword-only arguments in all tensor constructor helpers (tensor, matrix, ...)

New Features 🎉

• Add slogdet for Numba and JAX by @mtsokol in #172
• [from 2.9.0] Compute static shape types in outputs of Join by @michaelosthege in #164
• [from 2.9.0] Add Ops for Gaussian Hypergeometric Function, Pochhammer Symbol, and Factorials by @ColtAllen in #90
• [from 2.8.12] 🔄 From Aesara: 1362: "Add HalfNormalRV JAX implementation" by @theorashid in #129

Bugfixes 🐛

• Handle invalid BroadcastTo shape in C backend by @ricardoV94 in #175
• [from 2.8.12] Fix bug in truncated_graph_inputs by @ricardoV94 in #113
• [from 2.8.12] Fix JAX dispatch for multi-output Composite by @ricardoV94 in #123

Maintenance 🔧

• Improve CI pinelines by @danhphan in #176
• Seed logsumexp benchmark tests by @ricardoV94 in #185
• Create README.md by @canyon289 in #190
• Fix release pipeline by @cerrussell in #183
• [from 2.9.0] Fix PyPI description and update GitHub Actions by @michaelosthege in #171
• [from 2.9.0] A more low level implementation of vectorize in numba by @aseyboldt in #92
• [from 2.8.12] Update Release Notes template by @ricardoV94 in #117
• [from 2.8.12] 🔄 From Aesara: 1345: "Splitting before using JAX key" by @AdrienCorenflos in #130
• [from 2.8.12] Make tests compatible with newer version of JAX by @ricardoV94 in #133
• [from 2.8.12] Make scan to be a cython extention by @ferrine in #77
• [from 2.8.12] Remove deprecated modules by @ferrine in #111
• [from 2.8.12] 🔄 From Aesara: #1347 and # 1365: " Add CI support for benchmarking" by @brandonwillard in #139
• [from 2.8.12] Added PyTensor logo SVG by @fonnesbeck in #142
• [from 2.8.12] Remove some inline arguments and make some njit functions global by @aseyboldt in #144
• [from 2.8.12] Fix logit sigmoid function typo by @guin0x in #157

New Contributors

• @danhphan made their first contribution in #176
• @mtsokol made their first contribution in #172
• @canyon289 made their first contribution in #190
• @cerrussell made their first contribution in #183
• @fonnesbeck made their first contribution in #142
• @guin0x made their first contribution in #157
• @ColtAllen made their first contribution in #90

Full Changelog: rel-2.8.11...rel-2.8.12
Full Changelog: rel-2.8.12...rel-2.9.0
Full Changelog: rel-2.9.0...rel-2.9.1

rel-2.9.0

⚠️ This version was not released on PyPI or conda-forge.

What's Changed

New Features 🎉

• Compute static shape types in outputs of Join by @michaelosthege in #164
• Add Ops for Gaussian Hypergeometric Function, Pochhammer Symbol, and Factorials by @ColtAllen in #90

Maintenance 🔧

• Fix PyPI description and update GitHub Actions by @michaelosthege in #171
• A more low level implementation of vectorize in numba by @aseyboldt in #92

New Contributors

• @ColtAllen made their first contribution in #90

Full Changelog: rel-2.8.12...rel-2.9.0

rel-2.8.12

⚠️ This version was not released on PyPI or conda-forge.

What's Changed

Major Changes 🛠

• Allow passing static shape to tensor creation helpers by @ricardoV94 in #118
  • The first and only positional argument to tensor is now name. dtype and shape are now keyword-only arguments in all tensor constructor helpers (tensor, matrix, ...)

New Features 🎉

• 🔄 From Aesara: 1362: "Add HalfNormalRV JAX implementation" by @theorashid in #129

Bugfixes 🐛

• Fix bug in truncated_graph_inputs by @ricardoV94 in #113
• Fix JAX dispatch for multi-output Composite by @ricardoV94 in #123

Maintenance 🔧

• Update Release Notes template by @ricardoV94 in #117
• 🔄 From Aesara: 1345: "Splitting before using JAX key" by @AdrienCorenflos in #130
• Make tests compatible with newer version of JAX by @ricardoV94 in #133
• Make scan to be a cython extention by @ferrine in #77
• Remove deprecated modules by @ferrine in #111
• 🔄 From Aesara: #1347 and # 1365: " Add CI support for benchmarking" by @brandonwillard in #139
• Added PyTensor logo SVG by @fonnesbeck in #142
• Remove some inline arguments and make some njit functions global by @aseyboldt in #144
• Fix logit sigmoid function typo by @guin0x in #157

New Contributors

• @fonnesbeck made their first contribution in #142
• @guin0x made their first contribution in #157

Full Changelog: rel-2.8.11...rel-2.8.12

rel-2.8.11

Changes in PyTensor

• Optimize Scan inner graph when compiling to Numba by @ricardoV94 in #28
• Some numba backend fixes by @aseyboldt in #46
• Misc RandomVariable improvements by @ricardoV94 in #79
• Fix bug in AdvancedSubtensor infer_shape by @ricardoV94 in #101
• Add script to downstream PRs. by @twiecki in #39
• Add plain issue template by @tjburch in #42
• Fix contributing broken link by @madt2709 in #45
• Some pylint and pyupgrade cleanups by @Armavica in #29
• Fix argument name in is_in_ancestors by @sudarsan2k5 in #67
• Stop using deprecated generated_jit by @aseyboldt in #69
• Add dict type to check_deterministic by @ferrine in #75
• Add script to upstream PRs. by @twiecki in #68
• Add boundschecks in numba backend by @aseyboldt in #94
• Implement graph replace utility by @ferrine in #66
• Typing fixes by @michaelosthege in #104
• Remove unused files, update contributing guide and PR template by @michaelosthege in #105
• Improve numba DimShuffle compile time by @aseyboldt in #95

Changes in Aesara that were incorporated in PyTensor

• Add solve_discrete_lyapunov and solve_continuous_lyapunov by @jessegrabowski in #33
• Remove CholeskyGrad Op 1302 by @sudarsan2k5 in #36
• Replace time.time with time.perf_counter by @redbopo in #34
• Harmonize softplus implementations by @ricardoV94 in #35
• Refactor SharedVariable and its interfaces by @brandonwillard in #37
• Remove redundant graph_inputs usage in OpFromGraph by @brandonwillard in #38
• Add a fusion rewrite for CAReduces with Elemwise inputs by @brandonwillard in #40
• Add type hint None to specify_shape by @sudarsan2k5 in #82
• Add basic support for SciPy CSC and CSR sparse types in Numba by @brandonwillard in #96
• Fix the JAX RandomVariable dispatcher by @rlouf in #103
• Add JAX implementation for LogNormalRV by @juanitorduz in #106

New Contributors

• @larryshamalama made their first contribution in #27
• @ricardoV94 made their first contribution in #28
• @twiecki made their first contribution in #33
• @tjburch made their first contribution in #42
• @madt2709 made their first contribution in #45
• @Armavica made their first contribution in #29
• @sudarsan2k5 made their first contribution in #67
• @aseyboldt made their first contribution in #69

Full Changelog: rel-2.8.10...rel-2.8.11

rel-2.8.10

What changed since forking from aesara-devs/aesara@5aa17801

• The project was renamed to PyTensor by @pymc-devs.
• Updates to the license file, governance structure, code of conduct and readme by @pymc-devs, @ricardoV94, @twiecki, @michaelosthege, @OriolAbril.
• Addition of new issue templates by @ferrine in #8
• Switched to PyMC documentation theme by @OriolAbril in #18
• Updated the pre-commit to mypy v0.991 by @michaelosthege in d88f3f9

Changes since Aesara 2.8.9

The last Aesara release before we forked was 2.8.9 and these were the contributions to aesara-devs/aesara that happened before we forked:

• Updated time.clock() to time.perf_counter() in documentation under tutorial/conditions by @ntmkhanh in aesara-devs/aesara#1293
• Add kwargs to aesara.graph.basic.clone_get_equiv by @brandonwillard in aesara-devs/aesara#1296
• Replace some uses of TensorType.broadcastable with TensorType.shape by @brandonwillard in aesara-devs/aesara#1297
• Make as_symbolic work with sparse matrices by @brandonwillard in aesara-devs/aesara#1298
• Update codecov-action version by @brandonwillard in aesara-devs/aesara#1300
• Add name argument to clone methods via kwargs by @wd60622 in aesara-devs/aesara#1282