GitLab

* Add HyperGeometric distribution to discrete.py; Add tests

* Add HyperGeo to distirbutions/__init__.py

* Fix minor linting issue

* Add ref_rand helper function. Clip lower in logp

* Fix bug. Now pymc3_matches_scipy runs without error but pymc3_random_discrete diverges from expected value

* passes match with scipy test in test_distributions.py but fails in test_distributions_random.py

* Modify HyperGeom.random; Random test still failing. match_with_scipy test passing

* rm stray print

* Fix failing random test by specifying domain

* Update pymc3/distributions/discrete.py

Remove stray newline
Co-authored-by: Tirth Patel <tirthasheshpatel@gmail.com>

* Add note in RELEASE-NOTES.md
Co-authored-by: Tirth Patel <tirthasheshpatel@gmail.com>

* Updated Mauna Loa notebook

* Add link to cell2location to README

* adding logcdf method to inverse gamma distribution

* formatting fixes

* more formatting fixes

* mark xfail for float32

* updating release notes to master

* adding PR to release notes
Co-authored-by: Eelke Spaak <eelke.spaak@gmail.com>

* update contributing

* rename

* put watermark back in (minimal impact anyway)

* keep graphviz in

* failfast

* add failing test

* fail fast

* don't fail a test on purpose

* indent

* indent

* remove -x

* remove arviz with conda

* dont remove patsy

* try using python version correctly

* move non-theano non-optional dependencies out of environment-dev.yml

* keep python in environment-dev.yml, that's a separate issue

* typo

* put patsy back; final commit?

* make separate env files

* show python version

* update env name in contributing.md

* redistribute tests

* add release manager name to release notes

As suggested by @twiecki adding @michaelosthege (3.9.0/3.9.1) and my name (3.10.0) to release notes.

* adding Alex for 3.9.2

* add Kyle for 3.9.3

* Fix a couple of typos

* Apply suggestions from code review
Co-authored-by: Thomas Wiecki <thomas.wiecki@gmail.com>

Co-authored-by: Robert P. Goldman <rpgoldman@sift.net>

* Fixed MvNormal.random method

* Added some comments

* Handled corner case when self.shape is not provided

* Fixed comments

* Considered the corner case of 'point' as well.

* Fixed MvNormal.random method

* Handled sample and batch dimensions in tau parametrization using numpy

Added batch dimensions to all parametrization

* Modified logic while inserting batch dimensions to parametrization

* Used shapes_utils.broadcast_dist_samples_to function for broadcasting

* Make pylint pass

* Make test passes :crossed_fingers: hopefully

* Modified logic and added tests

* Given a mention in release notes

* improve coverage

* redistribute tests

* use np.shape

* remove unreachable code

* clarify docs, invert guard

* clarify that steps is a sequence

* 📝 update docstring

* make test less strict

* no parallel testing

* dont parallelise arviz tests either

* Add JAX NUTS samplers from TFP and numpyro. With @junpenglao.

* Add missing import.

* Remove JAX as default linker.

* Add experimental warning and clean up imports.

* Add JAX nb.

* Add NB to toc.

* Black and isort.

* Change title.

* Remove comma

* Typo.

* nbqa NB

* Run pre-commit.

* Disable pylint.

* Add to release-notes.

* Re-create branch

* Fix merge conflict

* bug fix

* remove unneeded import

* Update pymc3/util.py as per twiecki
Co-authored-by: Thomas Wiecki <thomas.wiecki@gmail.com>

* fix test_examples.py

* fix test_step.py

* remove unnecessary import
Co-authored-by: Stephen Hogg <stephen.hogg@harrison.ai>
Co-authored-by: Thomas Wiecki <thomas.wiecki@gmail.com>

* add str/repr formatting options and change defaults accordingly
+ new options "latex_with_params" and "plain_with_params" replace the current default behavior of including input parameters (going back to default behaviour of the 3.9.3 release)
+ __latex__ and _repr_latex default to "latex_with_params"
+ __str__ and _str_repr default to "plain"
+ new formatting kwarg for model_to_graphviz enables switching between "plain" (default) and "plain_with_params"

* update if conditions, formatting defaults and tests
+ latex formatting should be detected by `if "latex" in formatting` to catch both format options
+ all latex reprs except for an entire model default to "latex_with_params"
+ tests now cover cases with and without params

* consolidate regression test into existing one
+ test_issue_4186 was merged into TestStrAndLatexRepr and now all 4 formatting options are covered.

* check graphviz results with all four formatting options

* put test_distributions in separate job

* 🎨 sort

* add arviz compat job

* fix ignored test

* document

* document

* noop

* put test_distributions in separate job

* 🎨 sort

* add arviz compat job

* update after_n_builds

* notify after 7 builds

* parallelise

* test

* remove comment, specify python>=3.6 in env.yml

* make prettier, upload codecov

Co-authored-by: Alexandre ANDORRA <andorra.alexandre@gmail.com>

* fallback __str__ to default Theano on error
* fix str repr for KroneckerNormal and MatrixNormal
* black formatting
* update release notes

* Add codecov.yml.

* Change correct config file.

* More experiments.

* Removed Text, SQLite and HDF5 backends
* Remove unnecessary line from .gitignnore
* Black formatting on sampling.py
* Restored save_trace and load_trace to global namespace
* Remove another backend test

* Adding new notebook for conjugate sampling

* Running precommit over notebook

* Fixing conflict for table of contents

* Updates to conjugate sampling notebook

* Fixing text and changing inline image format for conjugate sampling notebook

* Updating release notes to reference new conjugate samling notebook

* Updating release notes to reference new conjugate sampling notebook

* Updated Mauna Loa notebook

* Clarified parameterization of AR models in notebook

* remove print output and update text
+ the PyMC3 version is already shown in the watermark output at the end
+ a few small changes in the text of the cells that
Co-authored-by: Michael Osthege <m.osthege@fz-juelich.de>

Mention new string representations in release notes (as suggested by @michaelosthege).

* Changes for renamed scan_module in Theano

* Use direct import for OpFromGraph

* Remove direct use of Theano's old tests package

* Replace uses of old _get_test_value

* Update and pin theano-pymc version to 1.0.11

* Fix Theano type determination for generator observations

* Fix dtype in test_expand_packed_triangular

* Remove old test value exception compatibility code
Co-authored-by: Brandon T. Willard <brandonwillard@users.noreply.github.com>

* handling of parameters equal to None

* improved type juggling of bounds

* updating str repr for bounded variables

* adding bounded variable to str/_repr_latex test

* updating test

* black formatting

* removing unused import

* change string to r-string

* update from master

* black

* minor fix

* clean code

* blackify

* fix error residuals

* use a low number of max_stages for the first iteration, remove not necessary errors

* use Rockova prior, refactor prior leaf prob computaion

* clean code add docstring

* reduce code

* speed-up by fitting a subset of trees per step

* choose max

* improve docstrings

* refactor and clean code

* clean docstrings

* add tests and minor fixes.
Co-authored-by: aloctavodia <aloctavodia@gmail.com>
Co-authored-by: jmloyola <jmloyola@outlook.com>

* remove space.
Co-authored-by: aloctavodia <aloctavodia@gmail.com>
Co-authored-by: jmloyola <jmloyola@outlook.com>

* add variable importance report

* use ValueError

* wip return mean and std variable importance

* update variable importance report

* update release notes, remove vi hdi report

* test variable importance

* fix test
Co-authored-by: jmloyola <jmloyola@outlook.com>

closes #4219