Release Notes#

Version 1.5.0 (released Nov 06, 2026)#

This version adds support for Python version 3.15 and for missing values, and adds wheels for arm64 Linux. It drops support for Python 3.9 and 3.10 and now requires scikit-learn 1.9 or later. The leaf samples are now weighted by their sample weights, and the default interpolation is now "averaged_inverted_cdf", which matches scikit-learn’s quantile definition, so predictions can change. The training samples of each leaf are now stored without padding, which can greatly reduce the memory taken by a model and its pickle. The sparse_pickle parameter of fit is removed. Models pickled by earlier versions still load, but earlier versions cannot load models pickled by this version.

Breaking changes

  • Weight leaf samples by their sample weights, and default to scikit-learn’s quantile definition by @reidjohnson in #190

  • Store the training samples of each leaf without padding by @reidjohnson in #201

  • Prepare 1.5.0: release notes, nightly scikit-learn tests, cleanups by @reidjohnson in #202

New features

Performance

  • Make weighted predictions and proximity counts independent of training size by @reidjohnson in #179

  • Speed up out-of-bag sample lookups by @reidjohnson in #178

  • Store training targets as an array, so predict doesn’t copy them by @reidjohnson in #184

  • Compute OOB quantiles without a Python loop over rows by @reidjohnson in #185

  • Compute monotonic bounds one tree level at a time by @reidjohnson in #186

Bug fixes

  • Avoid changing global random state during leaf subsampling by @Ra1nhy in #167

  • Fix sample weight indexing when collecting leaf samples by @AryanRG in #168

  • Fix float max_samples_leaf in predict and leaf padding by @reidjohnson in #170

  • Stop quantile_ranks from resetting fitted feature attributes by @reidjohnson in #172

  • Validate sample weights before storing them in fit by @reidjohnson in #174

  • Fix proximity_counts argument checks and return types, and docstrings by @reidjohnson in #177

  • Fix models loaded from sparse pickles or read-only memory maps by @reidjohnson in #180

  • Fix sparse inputs in predict by @reidjohnson in #173

  • Validate quantiles as array-likes and reject NaN or empty quantiles by @reidjohnson in #175

  • Declare that missing values are not supported, and reject them up front by @reidjohnson in #176

  • Fix weighted quantiles that use zero-weight inputs or miss exact ranks by @reidjohnson in #171

  • Group leaf samples with a stable sort, so results don’t depend on the CPU by @reidjohnson in #181

  • Validate inputs once for OOB samples, without a warning per tree by @reidjohnson in #182

  • Use the fitted trees, not n_estimators, in predict by @reidjohnson in #183

  • Rank among the clipped leaf values in quantile_ranks with monotonic constraints by @reidjohnson in #188

  • Fix small issues found in a review of main by @reidjohnson in #199

  • Redraw OOB samples as the forest drew them, for older models by @reidjohnson in #200

Documentation

Maintenance

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.4.2…v1.5.0

Version 1.4.2 (released Jun 21, 2026)#

This version adds support for Python version 3.14 and scikit-learn 1.9.

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.4.1…v1.4.2

Version 1.4.1 (released Sep 10, 2025)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.4.0…v1.4.1

Version 1.4.0 (released Jan 21, 2025)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.11…v1.4.0

Version 1.3.11 (released Sep 10, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.10…v1.3.11

Version 1.3.10 (released Aug 31, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.9…v1.3.10

Version 1.3.9 (released Aug 23, 2024)#

  • Fix reproduction of scikit-learn trees with MAE criterion by @reidjohnson in #77

  • Update documentation and examples

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.8…v1.3.9

Version 1.3.8 (released Aug 15, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.7…v1.3.8

Version 1.3.7 (released Jun 19, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.6…v1.3.7

Version 1.3.6 (released May 22, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.5…v1.3.6

Version 1.3.5 (released Apr 15, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.4…v1.3.5

Version 1.3.4 (released Feb 21, 2024)#

  • Reorder multi-target outputs by @reidjohnson in #35

  • Add tests for model serialization by @reidjohnson in #36

  • Update and fix documentation and examples

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.3…v1.3.4

Version 1.3.3 (released Feb 16, 2024)#

  • Set default value of weighted_leaves at prediction time to False by @reidjohnson in #34

  • Update and fix documentation and examples

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.2…v1.3.3

Version 1.3.2 (released Feb 15, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.1…v1.3.2

Version 1.3.1 (released Feb 12, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.3.0…v1.3.1

Version 1.3.0 (released Feb 11, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.5…v1.3.0

Version 1.2.5 (released Feb 10, 2024)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.4…v1.2.5

Version 1.2.4 (released Jan 16, 2024)#

  • Use base model parameter validation when available

  • Resolve Cython 3 deprecation warnings

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.3…v1.2.4

Version 1.2.3 (released Oct 09, 2023)#

  • Fix bug that could prevent interpolation from being correctly applied by @reidjohnson in #15

  • Update documentation and docstrings

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.2…v1.2.3

Version 1.2.2 (released Oct 08, 2023)#

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.1…v1.2.2

Version 1.2.1 (released Oct 04, 2023)#

  • More efficient calculation of weighted quantiles by @reidjohnson in #11

  • Add support for Python version 3.12

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.2.0…v1.2.1

Version 1.2.0 (released Aug 01, 2023)#

  • Add optional default_quantiles parameter to the model initialization

  • Update documentation

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.1.3…v1.2.0

Version 1.1.3 (released Jul 08, 2023)#

  • Fix building from the source distribution

  • Minor update to documentation

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.1.2…v1.1.3

Version 1.1.2 (released Mar 22, 2023)#

  • Fix for compatibility with development version of scikit-learn

  • Update documentation and examples

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.1.1…v1.1.2

Version 1.1.1 (released Dec 19, 2022)#

  • Fix for compatibility with scikit-learn 1.2.0

  • Fix to documentation

  • Update version requirements

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.1.0…v1.1.1

Version 1.1.0 (released Nov 07, 2022)#

  • Update default max_samples_leaf to 1 (previously None)

  • Update documentation and unit tests

  • Miscellaneous update for compatibility with scikit-learn >= 1.1.0

This version supports Python versions 3.8 to 3.11. Note that support for 32-bit Python on Windows has been dropped in this release.

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.0.2…v1.1.0

Version 1.0.2 (released Mar 28, 2022)#

  • Add sample weighting by leaf size

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.0.1…v1.0.2

Version 1.0.1 (released Mar 23, 2022)#

  • Suppress UserWarning

Full Changelog: https://github.com/zillow/quantile-forest/compare/v1.0.0…v1.0.1

Version 1.0.0 (released Mar 23, 2022)#

Initial release.