Merged latest changes from sklearn upstream 07/20/24

Signed-off-by: Adam Li <adam2392@gmail.com>

Author: adam2392

.codecov.yml

@@ -30,5 +30,4 @@ ignore:
 - "sklearn/_build_utils"
 - "sklearn/__check_build"
 - "sklearn/_min_dependencies.py"
-- "**/setup.py"
 - "**/conftest.py"

.coveragerc

@@ -6,4 +6,3 @@ omit =
     */sklearn/externals/*
     */sklearn/_build_utils/*
     */benchmarks/*
-    **/setup.py

.github/workflows/cuda-gpu-ci.yml renamed to .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

.github/workflows/wheels.yml

@@ -164,7 +164,6 @@ jobs:
           CIBW_PRERELEASE_PYTHONS: ${{ matrix.prerelease_pythons }}
           CIBW_FREE_THREADED_SUPPORT: ${{ matrix.free_threaded_support }}
           CIBW_ENVIRONMENT: SKLEARN_SKIP_NETWORK_TESTS=1
-            SKLEARN_BUILD_PARALLEL=3
           CIBW_BUILD: cp${{ matrix.python }}-${{ matrix.platform_id }}
           CIBW_ARCHS: all
           CIBW_MANYLINUX_X86_64_IMAGE: ${{ matrix.manylinux_image }}
@@ -216,8 +215,6 @@ jobs:
 
       - name: Build source distribution
         run: bash build_tools/github/build_source.sh
-        env:
-          SKLEARN_BUILD_PARALLEL: 3
 
       - name: Test source distribution
         run: bash build_tools/github/test_source.sh

MANIFEST.in

@@ -7,30 +7,21 @@ all:
 	@echo "Please use 'make <target>' where <target> is one of"
 	@echo "  dev                  build scikit-learn with Meson"
 	@echo "  clean                clean scikit-learn Meson build. Very rarely needed,"
-	@echo "                       one use case is when switching back to setuptools"
-	@echo "  dev-setuptools       build scikit-learn with setuptools (deprecated)"
-	@echo "  clean-setuptools     clean scikit-learn setuptools build (deprecated)"
+	@echo "                       since meson-python recompiles on import."
 
 .PHONY: all
 
 dev: dev-meson
 
 dev-meson:
-	pip install --verbose --no-build-isolation --editable . --check-build-dependencies --config-settings editable-verbose=true
+	pip install --verbose --no-build-isolation --editable . --config-settings editable-verbose=true
 
 clean: clean-meson
 
 clean-meson:
 	pip uninstall -y scikit-learn
 	# It seems in some cases removing the folder avoids weird compilation
-	# errors (e.g. when switching from numpy>=2 to numpy<2). For some 
+	# errors (e.g. when switching from numpy>=2 to numpy<2). For some
 	# reason ninja clean -C $(DEFAULT_MESON_BUILD_DIR) is not
-	# enough
+	# enough.
 	rm -rf $(DEFAULT_MESON_BUILD_DIR)
-
-dev-setuptools:
-	$(PYTHON) setup.py build_ext -i
-
-clean-setuptools:
-	$(PYTHON) setup.py clean
-	rm -rf dist

Makefile

@@ -20,10 +20,10 @@
 
     // Customizable commands for building, installing, and
     // uninstalling the project. See asv.conf.json documentation.
-    //
-    // "install_command": ["python -mpip install {wheel_file}"],
-    // "uninstall_command": ["return-code=any python -mpip uninstall -y {project}"],
-    // "build_command": ["python -m build --wheel -o {build_cache_dir} {build_dir}"],
+    "install_command": ["python -mpip install {wheel_file}"],
+    "uninstall_command": ["return-code=any python -mpip uninstall -y {project}"],
+    "build_command": ["python -m build --wheel -o {build_cache_dir} {build_dir}"],
+
     // List of branches to benchmark. If not provided, defaults to "master
     // (for git) or "default" (for mercurial).
     "branches": ["main"],
@@ -72,12 +72,12 @@
     // those due to dependency changes.
     //
     "matrix": {
-        "numpy": ["1.25.2"],
-        "scipy": ["1.11.2"],
+        "numpy": ["2.0.0"],
+        "scipy": ["1.14.0"],
         "cython": ["3.0.10"],
         "joblib": ["1.3.2"],
         "threadpoolctl": ["3.2.0"],
-        "pandas": ["2.1.0"]
+        "pandas": ["2.2.2"]
     },
 
     // Combinations of libraries/python versions can be excluded/included

asv_benchmarks/asv.conf.json

@@ -208,7 +208,6 @@ jobs:
         SKLEARN_ENABLE_DEBUG_CYTHON_DIRECTIVES: '1'
         SKLEARN_RUN_FLOAT32_TESTS: '1'
         SKLEARN_TESTS_GLOBAL_RANDOM_SEED: '2'  # non-default seed
-        BUILD_WITH_SETUPTOOLS: 'true'
       # Linux environment to test the latest available dependencies.
       # It runs tests requiring lightgbm, pandas and PyAMG.
       pylatest_pip_openblas_pandas:
@@ -236,7 +235,7 @@ jobs:
       )
     matrix:
       debian_atlas_32bit:
-        DOCKER_CONTAINER: 'i386/debian:11.2'
+        DOCKER_CONTAINER: 'i386/debian:12'
         DISTRIB: 'debian-32'
         COVERAGE: "true"
         LOCK_FILE: './build_tools/azure/debian_atlas_32bit_lock.txt'

azure-pipelines.yml

@@ -107,10 +107,6 @@ scikit_learn_install() {
     setup_ccache
     show_installed_libraries
 
-    # Set parallelism to 3 to overlap IO bound tasks with CPU bound tasks on CI
-    # workers with 2 cores when building the compiled extensions of scikit-learn.
-    export SKLEARN_BUILD_PARALLEL=3
-
     if [[ "$UNAMESTR" == "Darwin" && "$SKLEARN_TEST_NO_OPENMP" == "true" ]]; then
         # Without openmp, we use the system clang. Here we use /usr/bin/ar
         # instead because llvm-ar errors
@@ -129,9 +125,7 @@ scikit_learn_install() {
         export LDFLAGS="$LDFLAGS -Wl,--sysroot=/"
     fi
 
-    if [[ "$BUILD_WITH_SETUPTOOLS" == "true" ]]; then
-        python setup.py develop
-    elif [[ "$PIP_BUILD_ISOLATION" == "true" ]]; then
+    if [[ "$PIP_BUILD_ISOLATION" == "true" ]]; then
         # Check that pip can automatically build scikit-learn with the build
         # dependencies specified in pyproject.toml using an isolated build
         # environment:
@@ -143,12 +137,6 @@ scikit_learn_install() {
            # toolchain
            ADDITIONAL_PIP_OPTIONS='-Csetup-args=--vsenv'
         fi
-        # TODO Always add --check-build-dependencies when all CI builds have
-        # pip >= 22.1.1. At the time of writing, two CI builds (debian32_atlas and
-        # ubuntu_atlas) have an older pip
-        if pip install --help | grep check-build-dependencies; then
-            ADDITIONAL_PIP_OPTIONS="$ADDITIONAL_PIP_OPTIONS --check-build-dependencies"
-        fi
         # Use the pre-installed build dependencies and build directly in the
         # current environment.
         pip install --verbose --no-build-isolation --editable . $ADDITIONAL_PIP_OPTIONS

build_tools/azure/install.sh

@@ -4,7 +4,7 @@
 channels:
   - defaults
 dependencies:
-  - python=3.9
+  - python=3.11
   - ccache
   - pip
   - pip: