Merge remote-tracking branch 'origin/branch-25.02' into wip_rf_interop

rapidsai · Jan 22, 2025 · 4b2bcb4 · 4b2bcb4
2 parents cc9e1e2 + 01e19bb
commit 4b2bcb4
Show file tree

Hide file tree

Showing 35 changed files with 501 additions and 108 deletions.
diff --git a/.github/CODEOWNERS b/.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

diff --git a/.github/workflows/build.yaml b/.github/workflows/build.yaml
@@ -91,3 +91,4 @@ jobs:
       sha: ${{ inputs.sha }}
       date: ${{ inputs.date }}
       package-name: cuml
+      package-type: python
diff --git a/.gitignore b/.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
diff --git a/ci/test_wheel.sh b/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}"

diff --git a/conda/environments/all_cuda-118_arch-x86_64.yaml b/conda/environments/all_cuda-118_arch-x86_64.yaml
@@ -75,7 +75,7 @@ dependencies:
 - sphinx-copybutton
 - sphinx-markdown-tables
 - statsmodels
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - treelite==4.4.1
 - umap-learn==0.5.6
 - xgboost>=2.1.0

diff --git a/conda/environments/all_cuda-125_arch-x86_64.yaml b/conda/environments/all_cuda-125_arch-x86_64.yaml
@@ -25,7 +25,7 @@ dependencies:
 - dask-ml
 - doxygen=1.9.1
 - fmt>=11.0.2,<12
-- gcc_linux-64=11.*
+- gcc_linux-64=13.*
 - graphviz
 - hdbscan>=0.8.39,<0.8.40
 - hypothesis>=6.0,<7
@@ -71,7 +71,7 @@ dependencies:
 - sphinx-copybutton
 - sphinx-markdown-tables
 - statsmodels
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - treelite==4.4.1
 - umap-learn==0.5.6
 - xgboost>=2.1.0

diff --git a/conda/environments/clang_tidy_cuda-118_arch-x86_64.yaml b/conda/environments/clang_tidy_cuda-118_arch-x86_64.yaml
@@ -33,6 +33,6 @@ dependencies:
 - ninja
 - nvcc_linux-64=11.8
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 - tomli
 name: clang_tidy_cuda-118_arch-x86_64
diff --git a/conda/environments/cpp_all_cuda-118_arch-x86_64.yaml b/conda/environments/cpp_all_cuda-118_arch-x86_64.yaml
@@ -31,5 +31,5 @@ dependencies:
 - ninja
 - nvcc_linux-64=11.8
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 name: cpp_all_cuda-118_arch-x86_64
diff --git a/conda/environments/cpp_all_cuda-125_arch-x86_64.yaml b/conda/environments/cpp_all_cuda-125_arch-x86_64.yaml
@@ -15,7 +15,7 @@ dependencies:
 - cuda-version=12.5
 - cxx-compiler
 - fmt>=11.0.2,<12
-- gcc_linux-64=11.*
+- gcc_linux-64=13.*
 - libcublas-dev
 - libcufft-dev
 - libcumlprims==25.2.*,>=0.0.0a0
@@ -27,5 +27,5 @@ dependencies:
 - librmm==25.2.*,>=0.0.0a0
 - ninja
 - spdlog>=1.14.1,<1.15
-- sysroot_linux-64==2.17
+- sysroot_linux-64==2.28
 name: cpp_all_cuda-125_arch-x86_64
diff --git a/conda/recipes/cuml-cpu/conda_build_config.yaml b/conda/recipes/cuml-cpu/conda_build_config.yaml
@@ -1,8 +1,10 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"
@@ -11,4 +13,4 @@ c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
diff --git a/conda/recipes/cuml/conda_build_config.yaml b/conda/recipes/cuml/conda_build_config.yaml
@@ -1,14 +1,14 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cuda_compiler:
-  - cuda-nvcc
-
-cuda11_compiler:
-  - nvcc
+  - cuda-nvcc  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - nvcc  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"
@@ -17,7 +17,7 @@ c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
 
 treelite_version:
   - "=4.3.0"
diff --git a/conda/recipes/cuml/meta.yaml b/conda/recipes/cuml/meta.yaml
@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -33,10 +33,8 @@ build:
     - SCCACHE_S3_KEY_PREFIX=cuml-linux64 # [linux64]
     - SCCACHE_S3_USE_SSL
   ignore_run_exports_from:
-    {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }}
-    {% else %}
     - {{ compiler('cuda') }}
+    {% if cuda_major != "11" %}
     - cuda-cudart-dev
     {% endif %}
     - cuda-python
@@ -46,7 +44,7 @@ requirements:
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
     {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }} ={{ cuda_version }}
+    - {{ compiler('cuda') }} ={{ cuda_version }}
     {% else %}
     - {{ compiler('cuda') }}
     {% endif %}

diff --git a/conda/recipes/libcuml/conda_build_config.yaml b/conda/recipes/libcuml/conda_build_config.yaml
@@ -1,20 +1,20 @@
 c_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cxx_compiler_version:
-  - 11
+  - 13  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - 11  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 cuda_compiler:
-  - cuda-nvcc
-
-cuda11_compiler:
-  - nvcc
+  - cuda-nvcc  # [not os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
+  - nvcc  # [os.environ.get("RAPIDS_CUDA_VERSION", "").startswith("11")]
 
 c_stdlib:
   - sysroot
 
 c_stdlib_version:
-  - "=2.17"
+  - "=2.28"
 
 cmake_version:
   - ">=3.26.4,!=3.30.0"

diff --git a/conda/recipes/libcuml/meta.yaml b/conda/recipes/libcuml/meta.yaml
@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -14,9 +14,7 @@ source:
 
 build:
   ignore_run_exports_from:
-    {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }}
-    {% endif %}
+    - {{ compiler('cuda') }}
   script_env:
     - AWS_ACCESS_KEY_ID
     - AWS_SECRET_ACCESS_KEY
@@ -38,7 +36,7 @@ requirements:
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
     {% if cuda_major == "11" %}
-    - {{ compiler('cuda11') }} ={{ cuda_version }}
+    - {{ compiler('cuda') }} ={{ cuda_version }}
     {% else %}
     - {{ compiler('cuda') }}
     {% endif %}
@@ -84,10 +82,8 @@ outputs:
       number: {{ GIT_DESCRIBE_NUMBER }}
       string: cuda{{ cuda_major }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
       ignore_run_exports_from:
-        {% if cuda_major == "11" %}
-        - {{ compiler('cuda11') }}
-        {% else %}
         - {{ compiler('cuda') }}
+        {% if cuda_major != "11" %}
         - cuda-cudart-dev
         {% endif %}
     requirements:
@@ -131,10 +127,8 @@ outputs:
       number: {{ GIT_DESCRIBE_NUMBER }}
       string: cuda{{ cuda_major }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
       ignore_run_exports_from:
-        {% if cuda_major == "11" %}
-        - {{ compiler('cuda11') }}
-        {% else %}
         - {{ compiler('cuda') }}
+        {% if cuda_major != "11" %}
         - cuda-cudart-dev
         {% endif %}
     requirements:

diff --git a/cpp/CMakeLists.txt b/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.")
@@ -420,6 +422,27 @@ if(BUILD_CUML_CPP_LIBRARY)
         src/hdbscan/hdbscan.cu
         src/hdbscan/condensed_hierarchy.cu
         src/hdbscan/prediction_data.cu)
+
+    # When using GCC 13, some maybe-uninitialized warnings appear from CCCL and are treated as errors.
+    # See this issue: https://github.com/rapidsai/cuml/issues/6225
+    set_property(
+      SOURCE src/hdbscan/condensed_hierarchy.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
+    set_property(
+      SOURCE src/hdbscan/hdbscan.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
+    set_property(
+      SOURCE src/hdbscan/prediction_data.cu
+      APPEND_STRING
+      PROPERTY COMPILE_FLAGS
+      " -Xcompiler=-Wno-maybe-uninitialized"
+    )
   endif()
 
   if(all_algo OR holtwinters_algo)

diff --git a/cpp/cmake/thirdparty/get_cuvs.cmake b/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}"
 
     )

diff --git a/cpp/cmake/thirdparty/get_raft.cmake b/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}

diff --git a/cpp/src/hdbscan/condensed_hierarchy.cu b/cpp/src/hdbscan/condensed_hierarchy.cu
@@ -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.
@@ -27,6 +27,7 @@
 
 #include <cub/cub.cuh>
 #include <cuda/functional>
+#include <cuda/std/functional>
 #include <thrust/copy.h>
 #include <thrust/device_ptr.h>
 #include <thrust/execution_policy.h>
@@ -157,8 +158,9 @@ void CondensedHierarchy<value_idx, value_t>::condense(value_idx* full_parents,
     thrust::cuda::par.on(stream),
     full_sizes,
     full_sizes + size,
-    cuda::proclaim_return_type<bool>([=] __device__(value_idx a) -> bool { return a != -1; }),
-    0,
+    cuda::proclaim_return_type<value_idx>(
+      [=] __device__(value_idx a) -> value_idx { return static_cast<value_idx>(a != -1); }),
+    static_cast<value_idx>(0),
     thrust::plus<value_idx>());
 
   parents.resize(n_edges, stream);

diff --git a/cpp/src/hdbscan/detail/utils.h b/cpp/src/hdbscan/detail/utils.h
@@ -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.
@@ -34,6 +34,7 @@
 #include <rmm/exec_policy.hpp>
 
 #include <cub/cub.cuh>
+#include <cuda/functional>
 #include <thrust/copy.h>
 #include <thrust/execution_policy.h>
 #include <thrust/for_each.h>
@@ -114,8 +115,9 @@ Common::CondensedHierarchy<value_idx, value_t> make_cluster_tree(
     thrust_policy,
     sizes,
     sizes + condensed_tree.get_n_edges(),
-    cuda::proclaim_return_type<bool>([=] __device__(value_idx a) -> bool { return a > 1; }),
-    0,
+    cuda::proclaim_return_type<value_idx>(
+      [=] __device__(value_idx a) -> value_idx { return static_cast<value_idx>(a > 1); }),
+    static_cast<value_idx>(0),
     thrust::plus<value_idx>());
 
   // remove leaves from condensed tree