Scikit-Learn 2.0.0

pts/scikit-learn-2.0.0 - 10 May 2023 - Update against SciKit-Learn 1.2.2 upstream, enable more tests.

downloads.xml

<?xml version="1.0"?>
<!--Phoronix Test Suite v10.8.4-->
<PhoronixTestSuite>
  <Downloads>
    <Package>
      <URL>https://github.com/scikit-learn/scikit-learn/archive/refs/tags/1.2.2.tar.gz</URL>
      <MD5>553a395c6a1d0bac55fa16e9dec32138</MD5>
      <SHA256>992694e21ce4285aab71b09939d3ed7e5ddb41b8803eb98e10aaba927b74bdf1</SHA256>
      <FileName>scikit-learn-1.2.2.tar.gz</FileName>
      <FileSize>7398844</FileSize>
    </Package>
    <Package>
      <URL>https://github.com/xianyi/OpenBLAS/archive/437c0bf2b4697339d96c7bd0bb0bcdac09eccba1.zip</URL>
      <MD5>622ddcb999fe99aec08f350eac316ffd</MD5>
      <SHA256>48261b648e3c726e1689e89fc8f2440a00624f35b07b4a0134f1a97fd4eddc30</SHA256>
      <FileName>OpenBLAS-437c0bf2b4697339d96c7bd0bb0bcdac09eccba1.zip</FileName>
      <FileSize>42119973</FileSize>
    </Package>
    <Package>
      <URL>https://github.com/numpy/numpy/releases/download/v1.24.3/numpy-1.24.3.tar.gz</URL>
      <MD5>89e5e2e78407032290ae6acf6dcaea46</MD5>
      <SHA256>ab344f1bf21f140adab8e47fdbc7c35a477dc01408791f8ba00d018dd0bc5155</SHA256>
      <FileName>numpy-1.24.3.tar.gz</FileName>
      <FileSize>10909904</FileSize>
    </Package>
    <Package>
      <URL>http://www.phoronix-test-suite.com/benchmark-files/scipy-20230505.tar.xz</URL>
      <MD5>1a360d852c0e9601ddd3ceef5c993dd0</MD5>
      <SHA256>0ec404694ce71db90bb2f421e12f98ce8d04e8d1e4771161793485360f8d5676</SHA256>
      <FileName>scipy-20230505.tar.xz</FileName>
      <FileSize>45026736</FileSize>
    </Package>
    <Package>
      <URL>https://github.com/joblib/joblib/archive/refs/tags/1.2.0.tar.gz</URL>
      <MD5>d51f8e86fe75ed94b88be83604c3f43d</MD5>
      <SHA256>574eca5aaeb0a06c4fe0a8e8e67b24715218fb0f9d0fdc9d4f30519100885e53</SHA256>
      <FileName>joblib-1.2.0.tar.gz</FileName>
      <FileSize>345345</FileSize>
    </Package>
    <Package>
      <URL>https://github.com/joblib/threadpoolctl/archive/refs/tags/3.1.0.tar.gz</URL>
      <MD5>5fa423ab40183a76b5b05e6ffbc6dcc5</MD5>
      <SHA256>449b806f9147a3cf28ed7ec6389dbff9c450331d695873ab75dea79fe661f1e7</SHA256>
      <FileName>threadpoolctl-3.1.0.tar.gz</FileName>
      <FileSize>29899</FileSize>
    </Package>
    <Package>
      <URL>http://www.phoronix-test-suite.com/benchmark-files/scikit-learn-benchmark-improvements-1.tar.xz</URL>
      <MD5>6746fd86739fec5f8b4e1c46a9f1c0c9</MD5>
      <SHA256>19c65c3c10d4a279a17d028e5de4235ebb53e8f1e46d05f85362e3a781c505a6</SHA256>
      <FileName>scikit-learn-benchmark-improvements-1.tar.xz</FileName>
      <FileSize>14480</FileSize>
    </Package>
  </Downloads>
</PhoronixTestSuite>

install.sh

#!/bin/bash
# Hack as scipy otherwise seems to have issues...
TEST_HOME=$HOME
export HOME=$DEBUG_REAL_HOME
#pip3 install --user numpy
# Build our own BLAS to deal with NUM_THREADS limits with default Ubuntu packages on high core count systems....
mkdir $TEST_HOME/blas-install
unzip -o OpenBLAS-437c0bf2b4697339d96c7bd0bb0bcdac09eccba1.zip
cd OpenBLAS-437c0bf2b4697339d96c7bd0bb0bcdac09eccba1
make NUM_THREADS=512 USE_OPENMP=1 USE_THREADS=1 DYNAMIC_ARCH=0 NO_AFFINITY=1 NO_WARMUP=1 BUILD_RELAPACK=0 COMMON_OPT="-O3" CFLAGS="-O3" FCOMMON_OPT="-O3 -fopenmp" FCFLAGS="-O3 -fopenmp"
make PREFIX=$HOME/blas-install/ NUM_THREADS=512 USE_OPENMP=1 USE_THREADS=1 DYNAMIC_ARCH=0 NO_AFFINITY=1 NO_WARMUP=1 BUILD_RELAPACK=0 COMMON_OPT="-O3" CFLAGS="-O3" FCOMMON_OPT="-O3 -fopenmp" FCFLAGS="-O3 -fopenmp" install
cd $TEST_HOME

export PATH=$HOME/blas-install/bin:$HOME.local/bin:$HOME/.local/bin:$HOME:$PATH
export LD_LIBRARY_PATH=$HOME/blas-install/lib:$HOME/blas-install/include:$LD_LIBRARY_PATH

tar -xf numpy-1.24.3.tar.gz
cd numpy-1.24.3
rm -f site.cfg
cp site.cfg.example site.cfg
echo "[openblas]
libraries = openblas
library_dirs = $HOME/blas-install/lib
include_dirs = $HOME/blas-install/include
runtime_library_dirs = $HOME/blas-install/lib" >> site.cfg
pip3 install --user . 2>&1
echo $? > $TEST_HOME/install-exit-status
pip3 install -U --user setuptools cython doit pydevtool pybind11 pythran meson mesonpy
pip3 install --use-pep517 -U --user pandas==2.0.0 matplotlib memory_profiler rich_click

cd $TEST_HOME
echo "import numpy as np
try:
	incdir = os.path.relpath(np.get_include())
except Exception:
	incdir = np.get_include()
print(\"NUMPYDIR: \" + incdir)" > test.py
echo "TEST PY"
python3 test.py 2>&1
NUMPY_DIR=`python3 test.py | cut -d':' -f2 | xargs`
rm -rf scipy
tar -xf scipy-20230505.tar.xz
cd scipy
rm -f site.cfg
cp site.cfg.example site.cfg
echo "[openblas]
libraries = openblas
library_dirs = $HOME/blas-install/lib
include_dirs = $HOME/blas-install/include
runtime_library_dirs = $HOME/blas-install/lib" >> site.cfg
sed -i 's/1.1.0/1.0.0/g' meson.build
sed -i "47 i incdir_numpy = '$NUMPYDIR'" scipy/meson.build
pip3 install --user .

cd $TEST_HOME
tar -xf joblib-1.2.0.tar.gz
pip3 install --user joblib-1.2.0/

# INSTALL THREADPOOLCTL
tar -xf threadpoolctl-3.1.0.tar.gz
pip3 install --user threadpoolctl-3.1.0/

tar -xf scikit-learn-1.2.2.tar.gz
pip3 install --user scikit-learn-1.2.2/
tar -xf scikit-learn-benchmark-improvements-1.tar.xz
cd scikit-learn-1.2.2
patch -p0 < ../benchmarks-better.patch
cd benchmarks/
rm -f bench_multilabel_metrics.py
rm -f bench_plot_randomized_svd.py
sed -i 's/plt.show()/ /g' bench_feature_expansions.py 
sed -i 's/plt.show()/ /g' bench_glm.py
sed -i 's/plt.show()/ /g' bench_glmnet.py
sed -i 's/plt.show()/ /g' bench_hist_gradient_boosting.py
sed -i 's/plt.show()/ /g' bench_isolation_forest.py
sed -i 's/plt.show()/ /g' bench_isotonic.py
sed -i 's/plt.show()/ /g' bench_kernel_pca_solvers_time_vs_n_components.py
sed -i 's/plt.show()/ /g' bench_kernel_pca_solvers_time_vs_n_samples.py
sed -i 's/plt.show()/ /g' bench_lasso.py
sed -i 's/plt.show()/ /g' bench_lof.py
sed -i 's/plt.show()/ /g' bench_online_ocsvm.py
sed -i 's/plt.show()/ /g' bench_plot_fastkmeans.py
sed -i 's/plt.show()/ /g' bench_plot_hierarchical.py
sed -i 's/plt.show()/ /g' bench_plot_incremental_pca.py
sed -i 's/plt.show()/ /g' bench_plot_lasso_path.py
sed -i 's/plt.show()/ /g' bench_plot_neighbors.py
sed -i 's/plt.show()/ /g' bench_plot_nmf.py
sed -i 's/plt.show()/ /g' bench_plot_omp_lars.py
sed -i 's/plt.show()/ /g' bench_plot_parallel_pairwise.py
sed -i 's/plt.show()/ /g' bench_plot_polynomial_kernel_approximation.py
sed -i 's/plt.show()/ /g' bench_plot_svd.py
sed -i 's/plt.show()/ /g' bench_plot_ward.py
sed -i 's/plt.show()/ /g' bench_rcv1_logreg_convergence.py
sed -i 's/plt.show()/ /g' bench_sample_without_replacement.py
sed -i 's/plt.show()/ /g' bench_sgd_regression.py
sed -i 's/plt.show()/ /g' bench_tree.py

pip3 install --use-pep517 --user glmnet glm

# Cache the datasets or anything else that may get downloaded...
#for f in "bench_l*.py"
#do
#  python3 $f
#done
cd $TEST_HOME
echo "#!/bin/sh
TEST_HOME=\$HOME
export HOME=\$DEBUG_REAL_HOME
cd scikit-learn-1.2.2/benchmarks/
python3 bench_\$@
echo \$? > \$TEST_HOME/test-exit-status" > scikit-learn
chmod +x scikit-learn

results-definition.xml

<?xml version="1.0"?>
<!--Phoronix Test Suite v10.8.4-->
<PhoronixTestSuite>
  <ResultsParser>
    <OutputTemplate>logistic_regression  #_RESULT_#   0.0136s     0.7266</OutputTemplate>
    <LineHint>Train Time, Seconds</LineHint>
  </ResultsParser>
  <ResultsParser>
    <OutputTemplate>logistic_regression  22.5031s   #_RESULT_#     0.7266  </OutputTemplate>
    <LineHint>Test Time, Seconds</LineHint>
  </ResultsParser>
  <SystemMonitor>
    <Sensor>sys.time</Sensor>
  </SystemMonitor>
</PhoronixTestSuite>

test-definition.xml

<?xml version="1.0"?>
<!--Phoronix Test Suite v10.8.4-->
<PhoronixTestSuite>
  <TestInformation>
    <Title>Scikit-Learn</Title>
    <AppVersion>1.2.2</AppVersion>
    <Description>Scikit-learn is a Python module for machine learning built on NumPy, SciPy, and is BSD-licensed.</Description>
    <ResultScale>Seconds</ResultScale>
    <Proportion>LIB</Proportion>
    <TimesToRun>3</TimesToRun>
    <IgnoreRuns>1</IgnoreRuns>
  </TestInformation>
  <TestProfile>
    <Version>2.0.0</Version>
    <SupportedPlatforms>Linux, BSD</SupportedPlatforms>
    <SoftwareType>Utility</SoftwareType>
    <TestType>System</TestType>
    <License>Free</License>
    <Status>Verified</Status>
    <ExternalDependencies>python, build-utilities, cmake, fortran-compiler, meson</ExternalDependencies>
    <ProjectURL>https://scikit-learn.org/</ProjectURL>
    <RepositoryURL>https://github.com/scikit-learn/scikit-learn</RepositoryURL>
    <Maintainer>Michael Larabel</Maintainer>
    <SystemDependencies>pkg-config, python3-config</SystemDependencies>
  </TestProfile>
  <TestSettings>
    <Option>
      <DisplayName>Benchmark</DisplayName>
      <Identifier>benchmark</Identifier>
      <Menu>
        <Entry>
          <Name>20 Newsgroups / Logistic Regression</Name>
          <Value>20newsgroups.py -e logistic_regression</Value>
        </Entry>
        <Entry>
          <Name>Covertype Dataset Benchmark</Name>
          <Value>covertype.py</Value>
        </Entry>
        <Entry>
          <Name>Feature Expansions</Name>
          <Value>feature_expansions.py</Value>
        </Entry>
        <Entry>
          <Name>GLM</Name>
          <Value>glm.py</Value>
        </Entry>
        <Entry>
          <Name>Glmnet</Name>
          <Value>glmnet.py</Value>
        </Entry>
        <Entry>
          <Name>Hist Gradient Boosting</Name>
          <Value>hist_gradient_boosting.py</Value>
        </Entry>
        <Entry>
          <Name>Hist Gradient Boosting Adult</Name>
          <Value>hist_gradient_boosting_adult.py</Value>
        </Entry>
        <Entry>
          <Name>Hist Gradient Boosting Categorical Only</Name>
          <Value>hist_gradient_boosting_categorical_only.py</Value>
        </Entry>
        <Entry>
          <Name>Hist Gradient Boosting Higgs Boson</Name>
          <Value>hist_gradient_boosting_higgsboson.py</Value>
        </Entry>
        <Entry>
          <Name>Hist Gradient Boosting Threading</Name>
          <Value>hist_gradient_boosting_threading.py</Value>
        </Entry>
        <Entry>
          <Name>Isolation Forest</Name>
          <Value>isolation_forest.py</Value>
        </Entry>
        <Entry>
          <Name>Isotonic / Perturbed Logarithm</Name>
          <Value>isotonic.py --iterations 100 --log_min_problem_size 1 --log_max_problem_size 10 --dataset perturbed_logarithm</Value>
        </Entry>
        <Entry>
          <Name>Isotonic / Logistic</Name>
          <Value>isotonic.py --iterations 100 --log_min_problem_size 1 --log_max_problem_size 10 --dataset logistic</Value>
        </Entry>
        <Entry>
          <Name>Isotonic / Pathological</Name>
          <Value>isotonic.py --iterations 100 --log_min_problem_size 1 --log_max_problem_size 10 --dataset pathological</Value>
        </Entry>
        <Entry>
          <Name>Kernel PCA Solvers / Time vs. N Components</Name>
          <Value>kernel_pca_solvers_time_vs_n_components.py</Value>
        </Entry>
        <Entry>
          <Name>Kernel PCA Solvers / Time vs. N Samples</Name>
          <Value>kernel_pca_solvers_time_vs_n_samples.py</Value>
        </Entry>
        <Entry>
          <Name>Lasso</Name>
          <Value>lasso.py</Value>
        </Entry>
        <Entry>
          <Name>LocalOutlierFactor</Name>
          <Value>lof.py</Value>
        </Entry>
        <Entry>
          <Name>SGDOneClassSVM</Name>
          <Value>online_ocsvm.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Fast KMeans</Name>
          <Value>plot_fastkmeans.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Hierarchical</Name>
          <Value>plot_hierarchical.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Incremental PCA</Name>
          <Value>plot_incremental_pca.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Lasso Path</Name>
          <Value>plot_lasso_path.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Neighbors</Name>
          <Value>plot_neighbors.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Non-Negative Matrix Factorization</Name>
          <Value>plot_nmf.py</Value>
        </Entry>
        <Entry>
          <Name>Plot OMP vs. LARS</Name>
          <Value>plot_omp_lars.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Parallel Pairwise</Name>
          <Value>plot_parallel_pairwise.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Polynomial Kernel Approximation</Name>
          <Value>plot_polynomial_kernel_approximation.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Singular Value Decomposition</Name>
          <Value>plot_svd.py</Value>
        </Entry>
        <Entry>
          <Name>Plot Ward</Name>
          <Value>plot_ward.py</Value>
        </Entry>
        <Entry>
          <Name>Sparse Random Projections / 100 Iterations</Name>
          <Value>random_projections.py --n-times 100</Value>
        </Entry>
        <Entry>
          <Name>RCV1 Logreg Convergencet</Name>
          <Value>rcv1_logreg_convergence.py</Value>
        </Entry>
        <Entry>
          <Name>SAGA</Name>
          <Value>saga.py</Value>
        </Entry>
        <Entry>
          <Name>Sample Without Replacement</Name>
          <Value>sample_without_replacement.py</Value>
        </Entry>
        <Entry>
          <Name>SGD Regression</Name>
          <Value>sgd_regression.py</Value>
        </Entry>
        <Entry>
          <Name>Sparsify</Name>
          <Value>sparsify.py</Value>
        </Entry>
        <Entry>
          <Name>Text Vectorizers</Name>
          <Value>text_vectorizers.py</Value>
        </Entry>
        <Entry>
          <Name>Tree</Name>
          <Value>tree.py</Value>
        </Entry>
        <Entry>
          <Name>MNIST Dataset</Name>
          <Value>mnist.py</Value>
        </Entry>
        <Entry>
          <Name>TSNE MNIST Dataset</Name>
          <Value>tsne_mnist.py</Value>
        </Entry>
      </Menu>
    </Option>
  </TestSettings>
</PhoronixTestSuite>