Merge branch 'branch-25.02' into new-umap-defaults

.github/CODEOWNERS

@@ -5,11 +5,9 @@ cpp/               @rapidsai/cuml-cpp-codeowners
 python/            @rapidsai/cuml-python-codeowners
 
 #cmake code owners
-**/CMakeLists.txt          @rapidsai/cuml-cmake-codeowners
-**/cmake/                  @rapidsai/cuml-cmake-codeowners
-python/cuml/pyproject.toml @rapidsai/cuml-cmake-codeowners
-build.sh                   @rapidsai/cuml-cmake-codeowners
-**/build.sh                @rapidsai/cuml-cmake-codeowners
+CMakeLists.txt @rapidsai/cuml-cmake-codeowners
+*.cmake        @rapidsai/cuml-cmake-codeowners
+**/cmake/      @rapidsai/cuml-cmake-codeowners
 
 #CI code owners
 /.github/                @rapidsai/ci-codeowners

.github/workflows/build.yaml

@@ -91,3 +91,4 @@ jobs:
       sha: ${{ inputs.sha }}
       date: ${{ inputs.date }}
       package-name: cuml
+      package-type: python

.gitignore

@@ -68,3 +68,8 @@ cpp/Doxyfile
 # clang tooling
 compile_commands.json
 .clangd/
+
+# generally prefer 'pyproject.toml' to 'pytest.ini' for pytest options
+# ref: https://github.com/rapidsai/cuml/pull/6201
+pytest.ini
+!python/cuml/cuml/benchmark/automated/pytest.ini

ci/test_wheel.sh

@@ -1,14 +1,15 @@
 #!/bin/bash
-# Copyright (c) 2023-2024, NVIDIA CORPORATION.
+# Copyright (c) 2023-2025, NVIDIA CORPORATION.
 
 set -euo pipefail
 
 mkdir -p ./dist
 RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
-RAPIDS_PY_WHEEL_NAME="cuml_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./dist
+RAPIDS_PY_WHEEL_NAME="cuml_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 python ./dist
 
 # echo to expand wildcard before adding `[extra]` requires for pip
-python -m pip install $(echo ./dist/cuml*.whl)[test]
+python -m pip install \
+  "$(echo ./dist/cuml*.whl)[test]"
 
 RAPIDS_TESTS_DIR=${RAPIDS_TESTS_DIR:-"${PWD}/test-results"}
 mkdir -p "${RAPIDS_TESTS_DIR}"

cpp/CMakeLists.txt

@@ -65,6 +65,7 @@ option(USE_CCACHE "Cache build artifacts with ccache" OFF)
 option(CUDA_STATIC_RUNTIME "Statically link the CUDA runtime" OFF)
 option(CUDA_STATIC_MATH_LIBRARIES "Statically link the CUDA math libraries" OFF)
 option(CUML_USE_CUVS_STATIC "Build and statically link the CUVS library" OFF)
+option(CUML_USE_RAFT_STATIC "Build and statically link the RAFT library" OFF)
 option(CUML_USE_TREELITE_STATIC "Build and statically link the treelite library" OFF)
 option(CUML_EXPORT_TREELITE_LINKAGE "Whether to publicly or privately link treelite to libcuml++" OFF)
 option(CUML_USE_CUMLPRIMS_MG_STATIC "Build and statically link the cumlprims_mg library" OFF)
@@ -99,6 +100,7 @@ message(VERBOSE "CUML_CPP: Cache build artifacts with ccache: ${USE_CCACHE}")
 message(VERBOSE "CUML_CPP: Statically link the CUDA runtime: ${CUDA_STATIC_RUNTIME}")
 message(VERBOSE "CUML_CPP: Statically link the CUDA math libraries: ${CUDA_STATIC_MATH_LIBRARIES}")
 message(VERBOSE "CUML_CPP: Build and statically link CUVS libraries: ${CUML_USE_CUVS_STATIC}")
+message(VERBOSE "CUML_CPP: Build and statically link RAFT library: ${CUML_USE_RAFT_STATIC}")
 message(VERBOSE "CUML_CPP: Build and statically link Treelite library: ${CUML_USE_TREELITE_STATIC}")
 
 set(CUML_ALGORITHMS "ALL" CACHE STRING "Experimental: Choose which algorithms are built into libcuml++.so. Can specify individual algorithms or groups in a semicolon-separated list.")

cpp/cmake/thirdparty/get_cuvs.cmake

@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright (c) 2024, NVIDIA CORPORATION.
+# Copyright (c) 2024-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -55,6 +55,7 @@ function(find_and_configure_cuvs)
         OPTIONS
           "BUILD_TESTS OFF"
           "BUILD_BENCH OFF"
+          "BUILD_CAGRA_HNSWLIB OFF"
           "BUILD_MG_ALGOS ${CUVS_BUILD_MG_ALGOS}"
 
     )

cpp/cmake/thirdparty/get_raft.cmake

@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright (c) 2021-2024, NVIDIA CORPORATION.
+# Copyright (c) 2021-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -43,7 +43,7 @@ function(find_and_configure_raft)
     message(VERBOSE "CUML: raft FIND_PACKAGE_ARGUMENTS COMPONENTS ${RAFT_COMPONENTS}")
 
     rapids_cpm_find(raft ${PKG_VERSION}
-      GLOBAL_TARGETS      raft::raft
+      GLOBAL_TARGETS      raft::raft raft::raft_logger raft::raft_logger_impl
       BUILD_EXPORT_SET    cuml-exports
       INSTALL_EXPORT_SET  cuml-exports
       COMPONENTS          ${RAFT_COMPONENTS}

dependencies.yaml

@@ -10,6 +10,7 @@ files:
       - cuda
       - cuda_version
       - depends_on_cudf
+      - depends_on_cuda_python
       - depends_on_cupy
       - depends_on_cuvs
       - depends_on_dask_cudf
@@ -21,8 +22,8 @@ files:
       - depends_on_raft_dask
       - depends_on_rmm
       - docs
-      - py_build
-      - py_run
+      - py_build_cuml
+      - py_run_cuml
       - py_version
       - rapids_build_backend
       - test_python
@@ -90,7 +91,7 @@ files:
       - depends_on_pylibraft
       - depends_on_raft_dask
       - depends_on_rmm
-      - py_run
+      - py_run_cuml
       - py_version
       - test_notebooks
   py_build_cuml:
@@ -108,27 +109,29 @@ files:
       key: requires
     includes:
       - common_build
+      - depends_on_cuda_python
       - depends_on_cuvs
       - depends_on_libcumlprims
       - depends_on_libraft_headers
       - depends_on_pylibraft
       - depends_on_rmm
-      - py_build
+      - py_build_cuml
   py_run_cuml:
     output: pyproject
     pyproject_dir: python/cuml
     extras:
       table: project
     includes:
       - cuda_wheels
+      - depends_on_cuda_python
       - depends_on_cudf
       - depends_on_cupy
       - depends_on_cuvs
       - depends_on_dask_cudf
       - depends_on_pylibraft
       - depends_on_raft_dask
       - depends_on_rmm
-      - py_run
+      - py_run_cuml
   py_test_cuml:
     output: pyproject
     pyproject_dir: python/cuml
@@ -210,28 +213,14 @@ dependencies:
               cuda: "12.*"
             packages:
               - cuda-nvcc
-  py_build:
+  py_build_cuml:
     common:
       - output_types: [conda, requirements, pyproject]
         packages:
           - &cython cython>=3.0.0
           - &treelite treelite==4.3.0
-    specific:
-      - output_types: [conda, requirements, pyproject]
-        matrices:
-          - matrix:
-              cuda: "12.*"
-            packages:
-              - cuda-python>=12.6.2,<13.0a0
-          - matrix:
-              cuda: "11.*"
-            packages:
-              - cuda-python>=11.8.5,<12.0a0
-          - matrix:
-            packages:
-              - cuda-python
 
-  py_run:
+  py_run_cuml:
     common:
       - output_types: [conda, requirements, pyproject]
         packages:
@@ -469,6 +458,21 @@ dependencies:
           - *scikit_learn
           - seaborn
           - *xgboost
+  depends_on_cuda_python:
+    specific:
+      - output_types: [conda, requirements, pyproject]
+        matrices:
+          - matrix:
+              cuda: "12.*"
+            packages:
+              - cuda-python>=12.6.2,<13.0a0
+          - matrix:
+              cuda: "11.*"
+            packages:
+              - cuda-python>=11.8.5,<12.0a0
+          - matrix:
+            packages:
+              - cuda-python
   depends_on_cudf:
     common:
       - output_types: conda

docs/source/sphinxext/github_link.py

@@ -1,5 +1,20 @@
 # This contains code with copyright by the scikit-learn project, subject to the
 # license in /thirdparty/LICENSES/LICENSE.scikit_learn
+#
+# Copyright (c) 2024-2025, NVIDIA CORPORATION.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
 
 import inspect
 import os
@@ -103,7 +118,7 @@ def _linkcode_resolve(domain, info, package, url_fmt, revision):
     else:
         if fn.endswith(".pyx"):
             sp_path = next(x for x in sys.path if re.match(".*site-packages$", x))
-            fn = fn.replace("/opt/conda/conda-bld/work/python", sp_path)
+            fn = fn.replace("/opt/conda/conda-bld/work/python/cuml", sp_path)
 
         # Convert to relative from module root
         fn = os.path.relpath(fn,

python/cuml/CMakeLists.txt

@@ -1,5 +1,5 @@
 # =============================================================================
-# Copyright (c) 2022-2024, NVIDIA CORPORATION.
+# Copyright (c) 2022-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except
 # in compliance with the License. You may obtain a copy of the License at
@@ -174,7 +174,7 @@ endif()
 include("${CUML_CPP_SRC}/cmake/modules/ConfigureAlgorithms.cmake")
 include(cmake/ConfigureCythonAlgorithms.cmake)
 
-if(${CUML_CPU})
+if(CUML_CPU)
   # libcuml requires metrics built if HDSCAN is built, which is not the case
   # for cuml-cpu
   unset(metrics_algo)

python/cuml/cuml/internals/import_utils.py

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2019-2024, NVIDIA CORPORATION.
+# Copyright (c) 2019-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -13,7 +13,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 #
-
 import platform
 
 from packaging.version import Version
@@ -145,11 +144,15 @@ def check_min_cupy_version(version):
         return False
 
 
-def has_scipy(raise_if_unavailable=False):
+def has_scipy(raise_if_unavailable=False, min_version=None):
     try:
         import scipy  # NOQA
 
-        return True
+        if min_version is None:
+            return True
+        else:
+            return Version(str(scipy.__version__)) >= Version(min_version)
+
     except ImportError:
         if not raise_if_unavailable:
             return False

python/cuml/cuml/tests/test_arima.py

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2019-2023, NVIDIA CORPORATION.
+# Copyright (c) 2019-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -36,7 +36,10 @@
 from cuml.internals.safe_imports import gpu_only_import
 import statsmodels.api as sm
 from sklearn.model_selection import train_test_split
-from cuml.internals.safe_imports import cpu_only_import_from
+from cuml.internals.safe_imports import (
+    cpu_only_import_from,
+    gpu_only_import_from,
+)
 import warnings
 import os
 import pytest
@@ -49,6 +52,7 @@
 approx_fprime = cpu_only_import_from("scipy.optimize", "approx_fprime")
 
 cudf = gpu_only_import("cudf")
+cudf_pandas_active = gpu_only_import_from("cudf.pandas", "LOADED")
 
 
 ###############################################################################
@@ -143,7 +147,7 @@ def __init__(
     n_obs=137,
     n_test=10,
     dataset="population_estimate",
-    tolerance_integration=0.01,
+    tolerance_integration=0.06,
 )
 
 # ARIMA(1,1,1) with intercept (missing observations)
@@ -255,7 +259,8 @@ def __init__(
     ((1, 1, 1, 0, 0, 0, 0, 1), test_111c_missing),
     ((1, 0, 1, 1, 1, 1, 4, 0), test_101_111_4),
     ((5, 1, 0, 0, 0, 0, 0, 0), test_510),
-    ((1, 1, 1, 2, 0, 0, 4, 1), test_111_200_4c),
+    # Skip due to update to Scipy 1.15
+    # ((1, 1, 1, 2, 0, 0, 4, 1), test_111_200_4c),
     ((1, 1, 1, 2, 0, 0, 4, 1), test_111_200_4c_missing),
     ((1, 1, 1, 2, 0, 0, 4, 1), test_111_200_4c_missing_exog),
     ((1, 1, 2, 0, 1, 2, 4, 0), test_112_012_4),
@@ -409,6 +414,11 @@ def fill_interpolation(df_in):
 @pytest.mark.parametrize("dtype", [np.float64])
 def test_integration(key, data, dtype):
     """Full integration test: estimate, fit, forecast"""
+    if (
+        data.dataset == "endog_hourly_earnings_by_industry_missing_exog"
+        and cudf_pandas_active
+    ):
+        pytest.skip(reason="https://github.com/rapidsai/cuml/issues/6209")
     order, seasonal_order, intercept = extract_order(key)
     s = max(1, seasonal_order[3])
 

python/cuml/cuml/tests/test_batched_lbfgs.py

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2019-2023, NVIDIA CORPORATION.
+# Copyright (c) 2019-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -14,6 +14,9 @@
 # limitations under the License.
 #
 
+import pytest
+
+from cuml.common import has_scipy
 from cuml.tsa.batched_lbfgs import batched_fmin_lbfgs_b
 from cuml.internals.safe_imports import cpu_only_import
 
@@ -64,6 +67,10 @@ def g_batched_rosenbrock(
     return gall
 
 
+@pytest.mark.xfail(
+    condition=has_scipy(min_version="1.15"),
+    reason="https://github.com/rapidsai/cuml/issues/6210",
+)
 def test_batched_lbfgs_rosenbrock():
     """Test batched rosenbrock using batched lbfgs implemtnation"""
 
@@ -107,7 +114,6 @@ def gf(x, n=None):
     res_xk, _, _ = batched_fmin_lbfgs_b(
         f, x0, num_batches, gf, iprint=-1, factr=100
     )
-
     np.testing.assert_allclose(res_xk, res_true, rtol=1e-5)