Iminuit

Latest version: v2.30.1

Safety actively analyzes 682416 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 13 of 13

1.3.3

-----------------------

- fix for broken table layout in print_param() and print_matrix()
- fix for missing error report when error is raised in user function
- fix of printout when ipython is used as a shell
- fix of slow convergence when analytical gradient is provided
- improved user guide with more detail information and improved structure

1.3.2

----------------------

- allow fixing parameter by setting limits (x, x) with some value x
- better defaults for maxcall arguments of hesse() and minos()
- nicer output for print_matrix()
- bug-fix: covariance matrix reported by iminuit was broken when some parameters were fixed
- bug-fix: segfault when something in PythonCaller raised an exception

1.3.1

---------------------

- fixed failing tests when only you installed iminuit with pip and don't
have Cython installed

1.3

------------------

- iminuit 1.3 is a big release, there are many improvements. All users are encouraged to update.
- Python 2.7 as well as Python 3.5 or later are supported, on Linux, MacOS and Windows.
- Source packages are available for PyPI/pip and we maintain binary package for conda (see :ref:`install`).
- The bundled Minuit C++ library has been updated to the latest version (takend from ROOT 6.12.06).
- The documentation has been mostly re-written. To learn about iminuit and all the new features,
read the :ref:`tutorials`.
- Numpy is now a core dependency, required to compile iminuit.
- For Numpy users, a second callback function interface and a ``Minuit.from_array_func`` constructor
was added, where the parameters are passed as an array.
- Results are now also available as Numpy arrays, e.g. ``np_values``, ``np_errors`` and ``np_covariance``.
- A wrapper function ``minimize`` for the MIGRAD optimiser was added,
that has the same arguments and return value format as ``scipy.optimize.minimize``.
- Support for analytical gradients has been added, users can pass a ``grad`` callback function.
This works, but for unknown reasons doesn't lead to performance improvements yet.
If you can help debug or fix this issue, please comment `here <https://github.com/scikit-hep/iminuit/issues/252>`__.
- Several issues have been fixed.
A complete list of issues and pull requests that went into the 1.3 release is
`here <https://github.com/scikit-hep/iminuit/milestone/4?closed=1>`__.

Previously
----------

- For iminuit releases before v1.3, we did not fill a change log.
- To summarise: the first iminuit release was v1.0 in Dec 2012.
In 2013 there were several releases, and in Jan 2014 the v1.1.1 release
was made. After that development was mostly inactive, except for the
v1.2 release in Nov 2015.
- The release history is available here: https://pypi.org/project/iminuit/#history
- The git history and pull requests are here: https://github.com/scikit-hep/iminuit


How to make a release

- Sync local `main` and `develop` with Github

- `for x in main develop; git checkout $x; git pull`

- On `develop` branch

- Update version in `pyproject.toml`
- For a beta release, add `.betaN`, where N is a number >= 0
- For a release candidate, add `.rcN`
- Run `python3 doc/update_changelog.py` or update `doc/changelog.rst` manually
- Check the output if you used the script
- If necessary, update ROOT version in `doc/conf.py`, see instructions there

- Merge `develop` into `main`

- Every push to `main` triggers building wheels, uploading to PyPI, and tagging/publishing on GitHub
- If there are problems with the wheels, commit fixes to `develop`, then merge again into `main`
- Note: Upload to PyPI uses API tokens configured in PyPI and Github "Secrets"

- conda-forge should pick up our release automatically and generate conda packages

Page 13 of 13

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.