v0.17.0

This release adds scikit-learn 1.9 support (fixing an import causalml.dataset failure present on the 0.16.0 wheel, #926), native Polars support across all meta-learners, a JAX/flax.nnx backend for DragonNet, and makes the meta-learners scikit-learn compliant (BaseEstimator, so clone() and get_params() work). It also adds the RATE evaluation metric and post-fit confidence intervals for BaseTLearner, plus numerous bug fixes.

New Features

• Add native Polars DataFrame, Series, and LazyFrame support for all meta-learners by @aman-coder03 in #901
• Polish Polars support and add bootstrap CI test coverage by @aman-coder03 in #921
• Add JAX/flax.nnx backend for DragonNet by @xrhd in #918
• docs: document JAX backend for DragonNet by @xrhd in #919
• Make meta-learners scikit-learn compliant via BaseEstimator by @aman-coder03 in #912
• Add Rank-weighted Average Treatment Effect (RATE) metric by @aman-coder03 in #887
• Add rate_score() with bootstrap confidence intervals and p-values by @aman-coder03 in #890
• Add AIPW docstring warning to get_toc() and rate_score() by @jeongyoonlee in #891
• Add post-fit confidence intervals to BaseTLearner via store_bootstraps and return_ci by @aman-coder03 in #886
• Support NaN values in UpliftTree and UpliftRandomForest by @aman-coder03 in #860

scikit-learn 1.9 Support

• Fix for compatibility with sklearn v1.9.0 by @jakevdp in #903
• Fix CausalRandomForestRegressor.fit() on scikit-learn 1.9 by @jeongyoonlee in #907
• Support CausalRandomForestRegressor.calculate_error() on scikit-learn >= 1.9 by @jeongyoonlee in #909

Bug Fixes

• Fix #904: Prevent deepcopy of fitted templates in bootstrap and correct predict validation ordering by @Saurav-Gupta-9741 in #910
• Optimised training, inference and memory for metalearners in multitreatment settings by @Ic3fr0g in #896
• Fix UpliftRandomForest predict shape mismatch with multiple treatments by @jeongyoonlee in #884
• Fix uplift tree p-value NaN from division by zero by @jeongyoonlee in #882
• Fix CausalRandomForestRegressor predicting inf from division by zero by @jeongyoonlee in #883
• Fix SensitivityPlaceboTreatment ignoring actual treatment groups by @jeongyoonlee in #880
• Fix seed parameter TypeError in BaseDRLearner bootstrap CI by @mohsinm-dev in #879
• Fix ValueError on read-only arrays in BaseSLearner.predict() by @mohsinm-dev in #878
• Add input validation to auuc_score for missing model columns by @jeongyoonlee in #881
• Make xgboost optional in synthetic dataset generation by @Si-ra-kri in #872
• Bug Fix: use iloc to index pd.Series by @bekojuniranjan in #877

Build / CI

• Make Cython line tracing opt-in to keep release wheels fast by @HSJung93 in #914
• Remove the PyPI token from GitHub Actions in favor of the Trusted Publishing by @jeongyoonlee in #871
• Upgrade GitHub Actions for Node 24 compatibility by @salmanmkc in #874
• Upgrade GitHub Actions to latest versions by @salmanmkc in #875
• ci: declare workflow-level contents: read on 4 workflows by @arpitjain099 in #900

⚠️ Breaking Changes

• Meta-learner __init__ signatures (#912): to become scikit-learn BaseEstimators, each learner now stores its constructor arguments verbatim and builds models in fit(). Most visibly, XGBRRegressor no longer accepts arbitrary **kwargs — pass XGBoost parameters via the explicit xgb_kwargs=<dict> argument.

Full Changelog: v0.16.0...v0.17.0