Merge branch 'main' of github.com:Quantco/metalearners into mkdocs

.pre-commit-config.yaml

@@ -59,13 +59,13 @@ repos:
         language: system
         entry: pixi run -e lint trailing-whitespace-fixer
         types: [text]
-        stages: [commit, push, manual]
+        stages: [pre-commit, pre-push, manual]
       - id: end-of-file-fixer
         name: fix end of files
         language: system
         entry: pixi run -e lint end-of-file-fixer
         types: [text]
-        stages: [commit, push, manual]
+        stages: [pre-commit, pre-push, manual]
       - id: check-merge-conflict
         name: check for merge conflicts
         language: system

CHANGELOG.rst

@@ -7,6 +7,14 @@
 Changelog
 =========
 
+0.12.0 (2025-01-29)
+-------------------
+
+**Other changes**
+
+* Comply with ``scikit-learn`` versions 1.6 and higher.
+
+
 0.11.0 (2024-09-05)
 -------------------
 

metalearners/_utils.py

@@ -10,11 +10,12 @@
 import numpy as np
 import pandas as pd
 import scipy
-from sklearn.base import check_array, check_X_y, is_classifier, is_regressor
+from sklearn.base import is_classifier, is_regressor
 from sklearn.ensemble import (
     HistGradientBoostingClassifier,
     HistGradientBoostingRegressor,
 )
+from sklearn.utils import check_array, check_X_y
 
 from metalearners._typing import Matrix, PredictMethod, Vector, _ScikitModel
 
@@ -131,6 +132,7 @@ def check_propensity_score(
         )
 
     if features is not None:
+
         check_X_y(features, propensity_scores, multi_output=True, **check_kwargs)
     else:
         check_array(propensity_scores, **check_kwargs)

metalearners/grid_search.py

@@ -79,7 +79,7 @@ def _fit_and_score(job: _FitAndScoreJob) -> GSResult:
 
 
 def _format_results(
-    results: list[GSResult] | Generator[GSResult, None, None]
+    results: list[GSResult] | Generator[GSResult, None, None],
 ) -> pd.DataFrame:
     rows = []
     for result in results:

metalearners/rlearner.py

@@ -479,7 +479,7 @@ def _pseudo_outcome_and_weights(
     ) -> tuple[np.ndarray, np.ndarray]:
         """Compute the R-Learner pseudo outcome and corresponding weights.
 
-        If `mask` is provided, the retuned pseudo outcomes and weights are only
+        If `mask` is provided, the returned pseudo outcomes and weights are only
         with respect the observations that the mask selects.
 
         Since the pseudo outcome is a fraction of residuals, we add a small

pixi.toml

@@ -27,7 +27,9 @@ mypy = "*"
 pytest = ">=6"
 pytest-cov = "*"
 pytest-xdist = "*"
-scikit-learn = ">=1.4"
+# Remove the upper bound once this xgboost problem
+# has been fixed: https://github.com/dmlc/xgboost/issues/11093
+scikit-learn = ">=1.4,<1.6"
 numpy = "*"
 pandas = "*"
 typing-extensions = "*"
@@ -84,7 +86,7 @@ docs-build = "mkdocs build --strict"
 readthedocs = "rm -rf $READTHEDOCS_OUTPUT/html && cp -r site/ $READTHEDOCS_OUTPUT/html"
 
 [feature.build.dependencies]
-build = "*"
+python-build = "*"
 
 [feature.build.tasks]
 build = "python -m build"
@@ -93,7 +95,7 @@ build = "python -m build"
 platforms = ["linux-64", "osx-arm64", "osx-64"]
 [feature.benchmark.dependencies]
 git_root = "*"
-python = "3.10"
+python = "3.10.*"
 scikit-learn = ">=1.3"
 numpy = "*"
 pandas = "*"

tests/test_rlearner.py

@@ -1,4 +1,4 @@
-# Copyright (c) QuantCo 2024-2024
+# Copyright (c) QuantCo 2024-2025
 # SPDX-License-Identifier: BSD-3-Clause
 
 from itertools import repeat
@@ -137,4 +137,4 @@ def test_rlearner_onnx(
         ["tau"],
         {"X": onnx_X},
     )
-    np.testing.assert_allclose(ml.predict(X, True, "overall"), pred_onnx, atol=5e-4)
+    np.testing.assert_allclose(ml.predict(X, True, "overall"), pred_onnx, atol=5e-3)