GitLab

* Allow specification of dims instead of shape

* Add pm.TidyData

* Create coords for pm.Data(ndarray)

* empty commit to trigger CI

* Apply suggestions from code review
Co-authored-by: Alexandre ANDORRA <andorra.alexandre@gmail.com>

* apply black formatting

* address review comments & formatting

* Add demonstration of named coordinates/dims

* don't require dim names to be identifiers

* sort imports

* raise ShapeError instead of ValueError

* formatting

* robustify Dtype and ShapeError

* Removed TidyData and refined dims and coords implementation

* Changed name of kwarg export_dims and improved docstrings

* Add link to ArviZ in docstrings

* Removed TidyData from __all__

* Polished Data container NB

* Fixed line break in data.py

* Fix inference of coords for dataframes

* Refined Data container NB

* Updated getting started NB with new dims and coords features

* Reran getting started NB

* Blackified NBs

* rerun with ArviZ branch

* use np.shape to be compatible with tuples/lists

* add tests for named coordinate handling

* Extended tests for data container
Co-authored-by: Michael Osthege <m.osthege@fz-juelich.de>
Co-authored-by: Michael Osthege <michael.osthege@outlook.com>
Co-authored-by: Alexandre ANDORRA <andorra.alexandre@gmail.com>

* add DtypeError with informative error message

* use float64 in augment system because of scipy

* set dtypes explicitly in all places

* fix incorrect sensitivity matrix column order
closes #3676
+ order of sensitivity columns now beginning with y0
+ tests were aligned
+ order of y0, theta now everywhere

* more detailed docstrings

* re-run after the fix

* update & re-run after the fix

* addition of test for equality checking of ODE Ops
(not yet implemented)

* WIP: refactoring the DifferentialEquation Op
+ full support for test_values
+ explicit input/output types
+ 2D return shape
+ optional return of sensitivities
+ gradient without helper Op

* fully replace DifferentialEquation Op with the refactored implementation

* align tests with refactored API
+ whitespace & condensed formatting
+ test for equality of identical Ops

* use tt.stack as suggested by DeprecationWarning

* always cast y0 and theta to floatX

* allow some tests to fail on float32
(due to downcast exception)

* don't use f-strings to maintain 3.5 support

* link ODE refactor PR

* renamed ODE notebooks
+ add notebooks to examples index

* use (custom) errors instead of asserts

* move ShapeError to exceptions.py

The idea is that a programmer be able to ignore imputation warnings if they know that data is being imputed. It's easier to do this with a distinct class than with just UserWarning.

Check for errors from loading bad trace directory.
Add new exception class.
While doing this, added type annotations to ndarray.py

* Small NUTS refactoring

* Save position of divergence in nuts

* Implement review suggestions

* Add tests for nuts reports

* Improve nuts report error messages