v5.3.1

What's Changed

New Features 🎉

• Derive logprob of < and > operations by @shreyas3156 in #6662
• Derive logprob of >= and <= operations by @shreyas3156 in #6680

Bugfixes 🪲

• Fix WhiteNoise Covariance bug by @dehorsley in #6674

Documentation 📖

• Link to PyMC v5 port of DBDA book by @cluhmann in #6681

Maintenance 🔧

• CustomDist and Simulator no longer require class_name when creating a dist by @ricardoV94 in #6668
• Make metropolis elemwise updates independent of each other by @ricardoV94 in #6678
• Make logprob inference for binary ops independent of order of inputs by @shreyas3156 in #6682
• Rectify type-hints for set_data by @Dhruvanshu-Joshi in #6676

New Contributors

• @PlayingData made their first contribution in #6683
• @dehorsley made their first contribution in #6674

Full Changelog: v5.3.0...v5.3.1

v5.3.0

What's Changed

Major Changes 🛠

• Bump Pytensor dependency by @ricardoV94 in #6665
  • Latex and string representation of variables now uses long Distribution names x ~ N(0, 1) -> x ~ Normal(0, 1)

New Features 🎉

• Implement Model.debug() helper by @ricardoV94 in #6634
• Added ICDF for the exponential distribution by @gokuld in #6641
• Issue warning if RVs are present in derived probability graphs by @ricardoV94 in #6651

Bugfixes 🪲

• Fix bug in random function of HalfStudent by @Dhruvanshu-Joshi in #6658
• Fix dtype casting bug in icdf function by @ricardoV94 in #6669

Documentation 📖

• Fix docstrings formatting and add unnamed_distribution to glossary by @alporter08 in #6638
• Add Gitpod video to contributing page by @reshamas in #6646
• Add timestamps to list of possible contributions by @reshamas in #6623
• Improve docstrings in GP module by @daniel-saunders-phil in #6652
• Arrange distributions and sub-contents alphabetically by @Dhruvanshu-Joshi in #6653

Maintenance 🔧

• Update links to sponsor images by @reshamas in #6643
• Add type hints to distribution parameters by @iykat in #6635

New Contributors

• @alporter08 made their first contribution in #6638
• @iykat made their first contribution in #6635

Full Changelog: v5.2.0...v5.3.0

v5.2.0

What's Changed

Major Changes 🛠

• Rename logprob/joint_logp to logprob/basic and move logcdf and icdf functions there by @ricardoV94 in #6599

New Features 🎉

• Add HSGP Latent GP approximation by @bwengals in #6458
• Allow logcdf and icdf inference by @ricardoV94 in #6597
• Implement icdf for the uniform distribution by @michaelraczycki in #6528
• Implement icdf for the discrete uniform distribution. by @gokuld in #6617
• Infer logprob of scans with carried auxiliary states by @ricardoV94 in #6582
• Infer logprob of IfElse graphs by @ricardoV94 in #6529
• Infer logprob of elemwise transformations of multivariate variables by @ricardoV94 in #6607
• Infer logprob of SpecifyShape and CheckandRaise by @Dhruvanshu-Joshi in #6538

Documentation 📖

• Add definition of B in documentation for Beta distribution by @JoKeyser in #6604
• Fix a typo in the docstring for OrderedLogistic by @NathanielF in #6611
• Improved docstring for predictions argument in sample_posterior_predictive by @fonnesbeck in #6616

Maintenance 🔧

• Renamed internal at aliases to pt by @shreyas3156 in #6577
• Remove auto argument from pm.Deterministic docstring by @shreyas3156 in #6592
• Remove shape helper functions by @symeneses in #6556
• Give more readable error message when checking the starting values for MCMC by @Dhruvanshu-Joshi in #6632
• Improve collect_default_updates by @ricardoV94 in #6620
• Minor improvements to GitPod by @maresb in #6618

New Contributors

• @shreyas3156 made their first contribution in #6577
• @NathanielF made their first contribution in #6611
• @gokuld made their first contribution in #6617

Full Changelog: v5.1.2...v5.2.0

v5.1.2

What's Changed

New Features 🎉

• Allow passing dims to Potential and Deterministic by @Raj-Parekh24 in #6576
• Add nuts_sampler_kwargs and nuts_kwargs to sample by @fonnesbeck in #6581
• Implement check_icdf helper to test icdf implementations by @ricardoV94 in #6583

Bugfixes 🪲

• Fix warn_treedepth looking at the wrong stat by @michaelosthege in #6591

Documentation 📖

• Gitpod instructions by @reshamas in #6549

Maintenance 🔧

• Add explicit support for Python 3.11 by @ricardoV94 in #6584
• Fix typos in Potential docstring by @chriswmann in #6575
• Fix typo in ZeroInflatedNegBinomial by @aleicazatti in #6585

New Contributors

• @Raj-Parekh24 made their first contribution in #6576
• @chriswmann made their first contribution in #6575

Full Changelog: v5.1.1...v5.1.2

v5.1.1

While this release carries a minor-version number increase, it is actually a major release (5.1). 5.1.0 was skipped due to a packaging issue.

What's Changed

Major Changes 🛠

• Access to external NUTS samplers (nutpie, blackjax, numpyro) with new sample() kwarg called nuts_sampler which often provide huge speed-ups and improved convergence by @twiecki and @ricardoV94 in #6422
• Enforce dims elements to be strings by @ricardoV94 in #6470
• Align DEMetropolis defaults with literature recommendations by @gbrunkhorst in #6488
• Tests are no longer part of the package by @Armavica in #6540
• Replace rvs_to_total_sizes mapping by MinibatchRandomVariables by @ricardoV94 in #6523

New Features 🎉

• Use separate argument in CustomDist for functions that return symbolic representations by @ricardoV94 in #6462
• Add nu parametrization to beta distribution by @alekracicot in #6344
• Preregister shapes of sampler stats by @michaelosthege in #6517
• Optional McBackend support by @michaelosthege in #6510
• Make VI (posterior) mean and std accessible as a structured xarray by @fonnesbeck in #6387
• Add mutable coords to Model by @Dhruvanshu-Joshi in #6515
• Allow Scan logprob inference of non-pure RandomVariable outputs by @ricardoV94 in #6578
• Public test utilities can now be found in testing module by @ricardoV94 in #6571

Bugfixes 🪲

• Fix bug in transformed scan variables by @ricardoV94 in #6573

Maintenance 🔧

• Fix some UserWarnings by @ricardoV94 in #6407
• Decouple convergence checking from SamplerReport by @michaelosthege in #6453
• Updated LICENSE to include AePPL license info by @fonnesbeck in #6474
• Fix typo in model comparison notebook by @Mikhail-Naumov in #6480
• Add devcontainer Docker image by @maresb in #6482
• Update GitPod base image to use pymc-built image by @maresb in #6483
• Register Op as subclass of Distributions with rv_type defined by @ricardoV94 in #6493
• Refactoring towards IBaseTrace interfaces by @michaelosthege in #6475
• Fix some type issues related to retrieving stats from traces by @michaelosthege in #6504
• Move pymc/distributions/logprob.py to pymc/logprob/ by @Armavica in #6441
• Update GLM_linear.ipynb by @bsenst in #6519
• Fix plot in HyperGeometric docstring by @aleicazatti in #6513
• Build docs in simplified environment by @michaelosthege in #6537
• BetaBinomial variance is not correctly described by @soma2000-lang in #6516
• Rename zerosum_axes to n_zerosum_axes by @michaelraczycki in #6522
• issue #5791, dims & cords inference from xarray by @michaelraczycki in #6514
• Fix some code lints found with ruff by @Armavica in #6545
• Add check for variables in step samplers by @michaelraczycki in #6524
• Fix two typos in documentation by @JoKeyser in #6547
• Bump PyTensor dependency by @ricardoV94 in #6531
• Updated docstring for Potential function by @Dhruvanshu-Joshi in #6559
• TruncatedNormal only accepts mu and sigma as non keyword arguments by @michaelraczycki in #6568
• Exclude tests from package discovery by @Armavica in #6552

New Contributors

• @Mikhail-Naumov made their first contribution in #6480
• @gbrunkhorst made their first contribution in #6488
• @dependabot made their first contribution in #6490
• @bsenst made their first contribution in #6519
• @Dhruvanshu-Joshi made their first contribution in #6515
• @aleicazatti made their first contribution in #6513
• @michaelraczycki made their first contribution in #6522
• @JoKeyser made their first contribution in #6547

Full Changelog: v5.0.2...v5.1.1

v5.1.0

This release was skipped due to a packaging problem.

v5.0.2

What's Changed

Major Changes 🛠

• Bump PyTensor to 2.9.1 by @michaelosthege in #6431
• Remove add_values/remove_values to fix backend type issues by @michaelosthege in #6451

New Features 🎉

• Infer logprob of absolute operations and fix logprob of powers with negative values by @LukeLB in #6414

Bugfixes 🪲

• Revert "Add informative error if user tries to use Aesara function with PyTensor variables by @ricardoV94 in #6440
• Fix bug that does not correctly set the dtype of determinsitic variab… by @jessegrabowski in #6425

Maintenance 🔧

• Inconsistent kwarg progressbar or progress_bar by @Gimmesoup in #6389
• Default VonMises kappa to 1.0 and fix math notation in doc by @wd60622 in #6419
• ModelGraph only stops ancestry check at model defined named_vars by @lucianopaz in #6426
• Add Matern12 covariance function to docs by @fonnesbeck in #6429
• Reference template notebook in Jupyter style guide by @reshamas in #6433
• Refactor internal sampling functions by @michaelosthege in #6444
• Refactoring and addition of helpers to handle flat stats by @michaelosthege in #6443
• Add seealso note to install docs by @OriolAbril in #6448

New Contributors

• @Gimmesoup made their first contribution in #6389
• @jessegrabowski made their first contribution in #6425

Full Changelog: v5.0.1...v5.0.2

v5.0.1

What's Changed

New Features 🎉

• Implement logp derivation for division, subtraction and negation by @ricardoV94 in #6371
• Extend logprob inference to power transforms by @ricardoV94 in #6400

Bugfixes 🐛

• Update PyTensor dependency and fix bugs in inferred mixture logprob by @ricardoV94 in #6397

Maintenance 🔧

• Update Release Notes template by @ricardoV94 in #6392
• Remove global RandomStream by @ricardoV94 in #6396
• Fix error in docstring of Truncated by @ricardoV94 in #6395
• added postprocessing_chunks option to sample_blackjax_nuts and sample… by @wnorcbrown in #6388
• replaces numpy sqrt method with pytensor equivalent by @morganstrom in #6405

New Contributors

• @wnorcbrown made their first contribution in #6388

Full Changelog: v5.0.0...v5.0.1

v5.0.0

What's Changed

In this major release we are switching our graph computation backend from Aesara to PyTensor, which is a fork of Aesara under PyMC governance.
Read the full announcement here: PyMC is Forking Aesara to PyTensor.

The switch itself should be rather seamless and you can probably just update your imports:

import aesara.tensor as at       # old (pymc >=4,< 5)
import pytensor.tensor as pt     # new (pymc >=5)

If you encounter problems updating please check the latest Discussions and don't hesitate to get in touch.

Major Changes 🛠

• ⚠ Switched the graph backend from Aesara to PyTensor
• Merged AePPL into a new logprob submodule. Dispatch methods can be found in logprob.abstract
• ⚠ The log_likelihood, needed for arviz.compare is no longer computed by default. It can be added with idata = pm.compute_log_likelihood(idata) or using pm.sample(idata_kwargs=dict(log_likelihood=True)) by @ricardoV94 in #6374
• Changed Minibatch API by @ferrine in #6304
• Fix ordering transformation for batched dimensions, and deprecate in favor of univariate_ordered and multivariate_ordered by @TimOliverMaier in #6255 and @ricardoV94 in #6375

New Features & Bugfixes 🎉

• Support logp derivation in DensityDist when random function returns a PyTensor variable by @ricardoV94 in #6361
• Added alternative parametrization for AsymmetricLaplace by @aloctavodia in #6337

Docs & Maintenance 🔧

• Bugfixes to increase robustness against unnamed dims by @michaelosthege in #6339
• Updated GOVERNANCE.md by @canyon289 in #6358
• Fixed overriding user provided mp_ctx strings to pm.sample() on M1 MacOS by @digicosmos86 in #6363
• Simplify measurable transform rewrites by @ricardoV94 in #6370
• Fix measurable stack and join with interdependent inputs by @ricardoV94 in #6342
• Allow transforms to work with multiple-valued nodes by @ricardoV94 in #6341
• Fix transformed Scan values by @ricardoV94 in #6343
• Add issue templates by @ferrine in #6327
• Fail docs build on errors in core notebooks by @michaelosthege in #6324
• Curated ecosystem references by @michaelosthege in #6383
• Switched run_mypy.py from pass-listing to fail-listing by @michaelosthege in #6381
• Runing pydocstyle in pre-commit by @michaelosthege in #6382
• Removed NoDistribution from docs by @stestoll in #6316
• Fix transforms example by @ricardoV94 in #6333

New Contributors

• @digicosmos86 made their first contribution in #6363

Full Changelog: v4.4.0...v5.0.0

v4.4.0

What's Changed

Major Changes 🛠

• Removed support for selectively tracking variables via pm.sample(trace=[...]). by @michaelosthege in #6269
• Do not rely on tag information for rv and logp conversions by @ricardoV94 in #6281. This includes:
  • Deprecated accessing any of [value_variable|observations|transform|total_size] via var.tag in favor of model.rvs_to_[values|transforms|total_sizes]
  • Deprecated joint_logp in favor of model.logp
  • Deprecated aesaraf.rvs_to_value_vars in favor of model.replace_rvs_by_values
• Using keyword seed for initial point no longer supported by @wd60622 in #6291
• Sampling of transformed variables from prior_predictive is no longer allowed by @ricardoV94 in #6309
• Require all step methods to return stats by @michaelosthege in #6313. This includes
  • Require all step methods to return stats from their step/astep method.
  • The BlockedStep.generates_stats attribute was removed.

New Features & Bugfixes 🎉

• Fix shared variable latex by @ricardoV94 in #6260
• Fix bug when replacing random variables with nested value transforms by @ricardoV94 in #6281
• Do not infer graph_model node types based on variable Op class by @ricardoV94 in #6259
• Do not propagate dims to observed component of imputed variable by @ricardoV94 in #6263
• Fix Categorical and Multinomial bugs by @ricardoV94 in #6265
• Sample stats for blackjax nuts by @TimOliverMaier in #6264
• Add warning if observed in DensityDist is dict by @symeneses in #6292
• Fix versioneer config to match version tags by @michaelosthege in #6314

Docs & Maintenance 🔧

• Split sampling.py into sampling.py and sampling_forward.py by @michaelosthege in #6257
• Improve join_nonshared_inputs documentation by @wd60622 in #6216
• Move sampling code to a submodule by @michaelosthege in #6268
• Fixed a typo in the overview notebook by @grtyvr in #6274
• Check that sampler vars correspond to value variables in the model by @ricardoV94 in #6235
• Updated pymc.DiscreteWeibull docstring by @hyosubkim in #6283
• Improve random seed processing for SMC sampling by @juanitorduz in #6298
• Remove wrong type-hints and stale docstrings from distributions by @ricardoV94 in #6280
• update theme by @OriolAbril in #6296
• Fix some typos and lints by @Armavica in #6300
• Update docstrings for sample_smc and smc.py by @rowangayleschaefer in #6114
• Fix Flaky Euler-Maruyama Tests by @wd60622 in #6287

New Contributors

• @grtyvr made their first contribution in #6274
• @hyosubkim made their first contribution in #6283

Full Changelog: v4.3.0...v4.4.0