Merge branch 'main' into gmIC

.github/workflows/cuda-gpu-ci.yml → .github/workflows/cuda-ci.yml

@@ -7,10 +7,6 @@ on:
     types:
       - labeled
 
-# In order to remove the "CUDA CI" label we need to have write permissions for PRs
-permissions:
-  pull-requests: write
-
 jobs:
   tests:
     if: contains(github.event.pull_request.labels.*.name, 'CUDA CI')
@@ -21,9 +17,6 @@ jobs:
     timeout-minutes: 20
     name: Run Array API unit tests
     steps:
-      - uses: actions-ecosystem/action-remove-labels@v1
-        with:
-          labels: CUDA CI
       - uses: actions/setup-python@v5
         with:
           # XXX: The 3.12.4 release of Python on GitHub Actions is corrupted:

.github/workflows/cuda-label-remover.yml

@@ -0,0 +1,23 @@
+name: Remove "CUDA CI" Label
+
+# This workflow removes the "CUDA CI" label that triggers the actual
+# CUDA CI. It is separate so that we can use the `pull_request_target`
+# trigger which has a API token with write access.
+on:
+  pull_request_target:
+    types:
+      - labeled
+
+# In order to remove the "CUDA CI" label we need to have write permissions for PRs
+permissions:
+  pull-requests: write
+
+jobs:
+  label-remover:
+    if: contains(github.event.pull_request.labels.*.name, 'CUDA CI')
+    name: Remove "CUDA CI" Label
+    runs-on: ubuntu-20.04
+    steps:
+      - uses: actions-ecosystem/action-remove-labels@v1
+        with:
+          labels: CUDA CI

doc/developers/maintainer.rst

@@ -105,12 +105,12 @@ This PR will be used to push commits related to the release as explained in
 :ref:`making_a_release`.
 
 You can also create a second PR from main and targeting main to increment the
-``__version__`` variable in `sklearn/__init__.py` and in `pyproject.toml` to increment
-the dev version. This means while we're in the release candidate period, the latest
-stable is two versions behind the main branch, instead of one. In this PR targeting
-main you should also include a new file for the matching version under the
-``doc/whats_new/`` folder so PRs that target the next version can contribute their
-changelog entries to this file in parallel to the release process.
+``__version__`` variable in `sklearn/__init__.py` to increment the dev version.
+This means while we're in the release candidate period, the latest stable is
+two versions behind the main branch, instead of one. In this PR targeting main
+you should also include a new file for the matching version under the
+``doc/whats_new/`` folder so PRs that target the next version can contribute
+their changelog entries to this file in parallel to the release process.
 
 Minor version release (also known as bug-fix release)
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -211,7 +211,7 @@ Making a release
      enough) and to update the on-going development entry.
 
 2. On the branch for releasing, update the version number in ``sklearn/__init__.py``,
-   the ``__version__`` variable, and in `pyproject.toml`.
+   the ``__version__`` variable.
 
    For major releases, please add a 0 at the end: `0.99.0` instead of `0.99`.
 

doc/modules/array_api.rst

@@ -117,12 +117,14 @@ Metrics
 - :func:`sklearn.metrics.d2_tweedie_score`
 - :func:`sklearn.metrics.max_error`
 - :func:`sklearn.metrics.mean_absolute_error`
+- :func:`sklearn.metrics.mean_absolute_percentage_error`
 - :func:`sklearn.metrics.mean_gamma_deviance`
 - :func:`sklearn.metrics.mean_squared_error`
 - :func:`sklearn.metrics.mean_tweedie_deviance`
 - :func:`sklearn.metrics.pairwise.additive_chi2_kernel`
 - :func:`sklearn.metrics.pairwise.chi2_kernel`
 - :func:`sklearn.metrics.pairwise.cosine_similarity`
+- :func:`sklearn.metrics.pairwise.cosine_distances`
 - :func:`sklearn.metrics.pairwise.euclidean_distances` (see :ref:`device_support_for_float64`)
 - :func:`sklearn.metrics.pairwise.paired_cosine_distances`
 - :func:`sklearn.metrics.pairwise.rbf_kernel` (see :ref:`device_support_for_float64`)

doc/whats_new/v1.6.rst

@@ -37,13 +37,17 @@ See :ref:`array_api` for more details.
 - :func:`sklearn.metrics.max_error` :pr:`29212` by :user:`Edoardo Abati <EdAbati>`;
 - :func:`sklearn.metrics.mean_absolute_error` :pr:`27736` by :user:`Edoardo Abati <EdAbati>`
   and :pr:`29143` by :user:`Tialo <Tialo>` and :user:`Loïc Estève <lesteve>`;
-- :func:`sklearn.metrics.mean_gamma_deviance` :pr:`29239` by :usser:`Emily Chen <EmilyXinyi>`;
+- :func:`sklearn.metrics.mean_absolute_percentage_error` :pr:`29300` by :user:`Emily Chen <EmilyXinyi>`;
+- :func:`sklearn.metrics.mean_gamma_deviance` :pr:`29239` by :user:`Emily Chen <EmilyXinyi>`;
 - :func:`sklearn.metrics.mean_squared_error` :pr:`29142` by :user:`Yaroslav Korobko <Tialo>`;
 - :func:`sklearn.metrics.mean_tweedie_deviance` :pr:`28106` by :user:`Thomas Li <lithomas1>`;
 - :func:`sklearn.metrics.pairwise.additive_chi2_kernel` :pr:`29144` by :user:`Yaroslav Korobko <Tialo>`;
 - :func:`sklearn.metrics.pairwise.chi2_kernel` :pr:`29267` by :user:`Yaroslav Korobko <Tialo>`;
 - :func:`sklearn.metrics.pairwise.cosine_similarity` :pr:`29014` by :user:`Edoardo Abati <EdAbati>`;
-- :func:`sklearn.metrics.pairwise.paired_cosine_distances` :pr:`29112` by :user:`Edoardo Abati <EdAbati>`.
+- :func:`sklearn.metrics.pairwise.cosine_distances` :pr:`29265` by :user:`Emily Chen <EmilyXinyi>`;
+- :func:`sklearn.metrics.pairwise.euclidean_distances` :pr:`29433` by :user:`Omar Salman <OmarManzoor>`;
+- :func:`sklearn.metrics.pairwise.paired_cosine_distances` :pr:`29112` by :user:`Edoardo Abati <EdAbati>`;
+- :func:`sklearn.metrics.pairwise.rbf_kernel` :pr:`29433` by :user:`Omar Salman <OmarManzoor>`.
 
 **Classes:**
 

examples/inspection/plot_permutation_importance_multicollinear.py

@@ -66,7 +66,6 @@ def plot_permutation_importance(clf, X, y, ax):
 
 mdi_importances = pd.Series(clf.feature_importances_, index=X_train.columns)
 tree_importance_sorted_idx = np.argsort(clf.feature_importances_)
-tree_indices = np.arange(0, len(clf.feature_importances_)) + 0.5
 
 fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 8))
 mdi_importances.sort_values().plot.barh(ax=ax1)

examples/linear_model/plot_quantile_regression.py

@@ -109,11 +109,6 @@
 #
 # We will use the quantiles at 5% and 95% to find the outliers in the training
 # sample beyond the central 90% interval.
-from sklearn.utils.fixes import parse_version, sp_version
-
-# This is line is to avoid incompatibility if older SciPy version.
-# You should use `solver="highs"` with recent version of SciPy.
-solver = "highs" if sp_version >= parse_version("1.6.0") else "interior-point"
 
 # %%
 from sklearn.linear_model import QuantileRegressor
@@ -122,7 +117,7 @@
 predictions = {}
 out_bounds_predictions = np.zeros_like(y_true_mean, dtype=np.bool_)
 for quantile in quantiles:
-    qr = QuantileRegressor(quantile=quantile, alpha=0, solver=solver)
+    qr = QuantileRegressor(quantile=quantile, alpha=0)
     y_pred = qr.fit(X, y_normal).predict(X)
     predictions[quantile] = y_pred
 
@@ -184,7 +179,7 @@
 predictions = {}
 out_bounds_predictions = np.zeros_like(y_true_mean, dtype=np.bool_)
 for quantile in quantiles:
-    qr = QuantileRegressor(quantile=quantile, alpha=0, solver=solver)
+    qr = QuantileRegressor(quantile=quantile, alpha=0)
     y_pred = qr.fit(X, y_pareto).predict(X)
     predictions[quantile] = y_pred
 
@@ -254,7 +249,7 @@
 from sklearn.metrics import mean_absolute_error, mean_squared_error
 
 linear_regression = LinearRegression()
-quantile_regression = QuantileRegressor(quantile=0.5, alpha=0, solver=solver)
+quantile_regression = QuantileRegressor(quantile=0.5, alpha=0)
 
 y_pred_lr = linear_regression.fit(X, y_pareto).predict(X)
 y_pred_qr = quantile_regression.fit(X, y_pareto).predict(X)

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "scikit-learn"
-version = "1.6.dev0"
+dynamic = ["version"]
 description = "A set of python modules for machine learning and data mining"
 readme = "README.rst"
 maintainers = [

sklearn/decomposition/_lda.py

@@ -194,15 +194,14 @@ class LatentDirichletAllocation(
         In general, if the data size is large, the online update will be much
         faster than the batch update.
 
-        Valid options::
-
-            'batch': Batch variational Bayes method. Use all training data in
-                each EM update.
-                Old `components_` will be overwritten in each iteration.
-            'online': Online variational Bayes method. In each EM update, use
-                mini-batch of training data to update the ``components_``
-                variable incrementally. The learning rate is controlled by the
-                ``learning_decay`` and the ``learning_offset`` parameters.
+        Valid options:
+
+        - 'batch': Batch variational Bayes method. Use all training data in each EM
+          update. Old `components_` will be overwritten in each iteration.
+        - 'online': Online variational Bayes method. In each EM update, use mini-batch
+          of training data to update the ``components_`` variable incrementally. The
+          learning rate is controlled by the ``learning_decay`` and the
+          ``learning_offset`` parameters.
 
         .. versionchanged:: 0.20
             The default learning method is now ``"batch"``.

sklearn/kernel_approximation.py

@@ -8,13 +8,9 @@
 
 import numpy as np
 import scipy.sparse as sp
+from scipy.fft import fft, ifft
 from scipy.linalg import svd
 
-try:
-    from scipy.fft import fft, ifft
-except ImportError:  # scipy < 1.4
-    from scipy.fftpack import fft, ifft
-
 from .base import (
     BaseEstimator,
     ClassNamePrefixFeaturesOutMixin,

sklearn/linear_model/_quantile.py

@@ -47,7 +47,7 @@ class QuantileRegressor(LinearModel, RegressorMixin, BaseEstimator):
         Method used by :func:`scipy.optimize.linprog` to solve the linear
         programming formulation.
 
-        From `scipy>=1.6.0`, it is recommended to use the highs methods because
+        It is recommended to use the highs methods because
         they are the fastest ones. Solvers "highs-ds", "highs-ipm" and "highs"
         support sparse input data and, in fact, always convert to sparse csc.
 
@@ -100,8 +100,7 @@ class QuantileRegressor(LinearModel, RegressorMixin, BaseEstimator):
     >>> X = rng.randn(n_samples, n_features)
     >>> # the two following lines are optional in practice
     >>> from sklearn.utils.fixes import sp_version, parse_version
-    >>> solver = "highs" if sp_version >= parse_version("1.6.0") else "interior-point"
-    >>> reg = QuantileRegressor(quantile=0.8, solver=solver).fit(X, y)
+    >>> reg = QuantileRegressor(quantile=0.8).fit(X, y)
     >>> np.mean(y <= reg.predict(X))
     0.8
     """
@@ -180,30 +179,18 @@ def fit(self, X, y, sample_weight=None):
         # So we rescale the penalty term, which is equivalent.
         alpha = np.sum(sample_weight) * self.alpha
 
-        if self.solver in (
-            "highs-ds",
-            "highs-ipm",
-            "highs",
-        ) and sp_version < parse_version("1.6.0"):
+        if self.solver == "interior-point" and sp_version >= parse_version("1.11.0"):
             raise ValueError(
-                f"Solver {self.solver} is only available "
-                f"with scipy>=1.6.0, got {sp_version}"
-            )
-        else:
-            solver = self.solver
-
-        if solver == "interior-point" and sp_version >= parse_version("1.11.0"):
-            raise ValueError(
-                f"Solver {solver} is not anymore available in SciPy >= 1.11.0."
+                f"Solver {self.solver} is not anymore available in SciPy >= 1.11.0."
             )
 
-        if sparse.issparse(X) and solver not in ["highs", "highs-ds", "highs-ipm"]:
+        if sparse.issparse(X) and self.solver not in ["highs", "highs-ds", "highs-ipm"]:
             raise ValueError(
                 f"Solver {self.solver} does not support sparse X. "
                 "Use solver 'highs' for example."
             )
         # make default solver more stable
-        if self.solver_options is None and solver == "interior-point":
+        if self.solver_options is None and self.solver == "interior-point":
             solver_options = {"lstsq": True}
         else:
             solver_options = self.solver_options
@@ -246,7 +233,7 @@ def fit(self, X, y, sample_weight=None):
             c[0] = 0
             c[n_params] = 0
 
-        if solver in ["highs", "highs-ds", "highs-ipm"]:
+        if self.solver in ["highs", "highs-ds", "highs-ipm"]:
             # Note that highs methods always use a sparse CSC memory layout internally,
             # even for optimization problems parametrized using dense numpy arrays.
             # Therefore, we work with CSC matrices as early as possible to limit
@@ -271,7 +258,7 @@ def fit(self, X, y, sample_weight=None):
             c=c,
             A_eq=A_eq,
             b_eq=b_eq,
-            method=solver,
+            method=self.solver,
             options=solver_options,
         )
         solution = result.x