Merge branch 'branch-25.04' into drop_deprecated_functors

.github/workflows/pr.yaml

@@ -56,6 +56,9 @@ jobs:
         uses: rapidsai/shared-actions/check_nightly_success/dispatch@main
         with:
           repo: cuml
+          # TODO: remove this once upstream issues in RAFT are resolved on 11.4.
+          # The limit was temporarily increased to unblock work.
+          max_days_without_success: 30
   changed-files:
     secrets: inherit
     needs: telemetry-setup

ci/test_python_dask.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2022-2024, NVIDIA CORPORATION.
+# Copyright (c) 2022-2025, NVIDIA CORPORATION.
 
 # Support invoking test_python_dask.sh outside the script directory
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
@@ -16,7 +16,6 @@ test_args=(
   --cov-config=../../../.coveragerc
   --cov=cuml_dask
   --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-dask-coverage.xml"
-  --cov-report=term
 )
 
 # Run tests

ci/test_python_integration.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2024, NVIDIA CORPORATION.
+# Copyright (c) 2024-2025, NVIDIA CORPORATION.
 
 # Support invoking test_python_singlegpu.sh outside the script directory
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
@@ -18,8 +18,7 @@ rapids-logger "pytest cuml integration"
   --junitxml="${RAPIDS_TESTS_DIR}/junit-cuml.xml" \
   --cov-config=../../.coveragerc \
   --cov=cuml \
-  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-coverage.xml" \
-  --cov-report=term
+  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-coverage.xml"
 
 rapids-logger "Test script exiting with value: $EXITCODE"
 exit ${EXITCODE}

ci/test_python_singlegpu.sh

@@ -18,8 +18,7 @@ rapids-logger "pytest cuml single GPU"
   --junitxml="${RAPIDS_TESTS_DIR}/junit-cuml.xml" \
   --cov-config=../../.coveragerc \
   --cov=cuml \
-  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-coverage.xml" \
-  --cov-report=term
+  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-coverage.xml"
 
   rapids-logger "pytest cuml accelerator"
 ./ci/run_cuml_singlegpu_accel_pytests.sh \
@@ -28,8 +27,7 @@ rapids-logger "pytest cuml single GPU"
   --junitxml="${RAPIDS_TESTS_DIR}/junit-cuml-accel.xml" \
   --cov-config=../../../../.coveragerc \
   --cov=cuml \
-  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-accel-coverage.xml" \
-  --cov-report=term
+  --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-accel-coverage.xml"
 
 
 if [ "${RAPIDS_BUILD_TYPE}" == "nightly" ]; then
@@ -41,7 +39,6 @@ if [ "${RAPIDS_BUILD_TYPE}" == "nightly" ]; then
     --cov-config=../../.coveragerc \
     --cov=cuml \
     --cov-report=xml:"${RAPIDS_COVERAGE_DIR}/cuml-memleak-coverage.xml" \
-    --cov-report=term \
     -m "memleak"
 fi
 

conda/environments/all_cuda-118_arch-aarch64.yaml

@@ -0,0 +1,81 @@
+# This file is generated by `rapids-dependency-file-generator`.
+# To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
+channels:
+- rapidsai
+- rapidsai-nightly
+- dask/label/dev
+- conda-forge
+- nvidia
+dependencies:
+- c-compiler
+- certifi
+- cmake>=3.30.4
+- cuda-python>=11.8.5,<12.0a0
+- cuda-version=11.8
+- cudatoolkit
+- cudf==25.4.*,>=0.0.0a0
+- cupy>=12.0.0
+- cuvs==25.4.*,>=0.0.0a0
+- cxx-compiler
+- cython>=3.0.0
+- dask-cuda==25.4.*,>=0.0.0a0
+- dask-cudf==25.4.*,>=0.0.0a0
+- dask-ml
+- doxygen=1.9.1
+- gcc_linux-64=11.*
+- graphviz
+- hdbscan>=0.8.39,<0.8.40
+- hypothesis>=6.0,<7
+- ipykernel
+- ipython
+- joblib>=0.11
+- libcublas-dev=11.11.3.6
+- libcublas=11.11.3.6
+- libcufft-dev=10.9.0.58
+- libcufft=10.9.0.58
+- libcumlprims==25.4.*,>=0.0.0a0
+- libcurand-dev=10.3.0.86
+- libcurand=10.3.0.86
+- libcusolver-dev=11.4.1.48
+- libcusolver=11.4.1.48
+- libcusparse-dev=11.7.5.86
+- libcusparse=11.7.5.86
+- libcuvs==25.4.*,>=0.0.0a0
+- libraft==25.4.*,>=0.0.0a0
+- librmm==25.4.*,>=0.0.0a0
+- nbsphinx
+- ninja
+- nltk
+- numba>=0.59.1,<0.61.0a0
+- numpy>=1.23,<3.0a0
+- numpydoc
+- nvcc_linux-aarch64=11.8
+- packaging
+- pydata-sphinx-theme!=0.14.2
+- pylibraft==25.4.*,>=0.0.0a0
+- pynndescent
+- pytest-benchmark
+- pytest-cases
+- pytest-cov
+- pytest-xdist
+- pytest==7.*
+- python>=3.10,<3.13
+- raft-dask==25.4.*,>=0.0.0a0
+- rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==25.4.*,>=0.0.0a0
+- rapids-logger==0.1.*,>=0.0.0a0
+- recommonmark
+- rmm==25.4.*,>=0.0.0a0
+- scikit-build-core>=0.10.0
+- scikit-learn==1.5.*
+- scipy>=1.8.0
+- seaborn
+- sphinx-copybutton
+- sphinx-markdown-tables
+- sphinx<8.2.0
+- statsmodels
+- sysroot_linux-aarch64==2.28
+- treelite==4.4.1
+- umap-learn==0.5.6
+- xgboost>=2.1.0
+name: all_cuda-118_arch-aarch64

conda/environments/all_cuda-128_arch-aarch64.yaml

@@ -0,0 +1,77 @@
+# This file is generated by `rapids-dependency-file-generator`.
+# To make changes, edit ../../dependencies.yaml and run `rapids-dependency-file-generator`.
+channels:
+- rapidsai
+- rapidsai-nightly
+- dask/label/dev
+- conda-forge
+- nvidia
+dependencies:
+- c-compiler
+- certifi
+- cmake>=3.30.4
+- cuda-cudart-dev
+- cuda-nvcc
+- cuda-profiler-api
+- cuda-python>=12.6.2,<13.0a0
+- cuda-version=12.8
+- cudf==25.4.*,>=0.0.0a0
+- cupy>=12.0.0
+- cuvs==25.4.*,>=0.0.0a0
+- cxx-compiler
+- cython>=3.0.0
+- dask-cuda==25.4.*,>=0.0.0a0
+- dask-cudf==25.4.*,>=0.0.0a0
+- dask-ml
+- doxygen=1.9.1
+- gcc_linux-aarch64=13.*
+- graphviz
+- hdbscan>=0.8.39,<0.8.40
+- hypothesis>=6.0,<7
+- ipykernel
+- ipython
+- joblib>=0.11
+- libcublas-dev
+- libcufft-dev
+- libcumlprims==25.4.*,>=0.0.0a0
+- libcurand-dev
+- libcusolver-dev
+- libcusparse-dev
+- libcuvs==25.4.*,>=0.0.0a0
+- libraft==25.4.*,>=0.0.0a0
+- librmm==25.4.*,>=0.0.0a0
+- nbsphinx
+- ninja
+- nltk
+- numba>=0.59.1,<0.61.0a0
+- numpy>=1.23,<3.0a0
+- numpydoc
+- packaging
+- pydata-sphinx-theme!=0.14.2
+- pylibraft==25.4.*,>=0.0.0a0
+- pynndescent
+- pytest-benchmark
+- pytest-cases
+- pytest-cov
+- pytest-xdist
+- pytest==7.*
+- python>=3.10,<3.13
+- raft-dask==25.4.*,>=0.0.0a0
+- rapids-build-backend>=0.3.0,<0.4.0.dev0
+- rapids-dask-dependency==25.4.*,>=0.0.0a0
+- rapids-logger==0.1.*,>=0.0.0a0
+- recommonmark
+- rmm==25.4.*,>=0.0.0a0
+- scikit-build-core>=0.10.0
+- scikit-learn==1.5.*
+- scipy>=1.8.0
+- seaborn
+- sphinx-copybutton
+- sphinx-markdown-tables
+- sphinx<8.2.0
+- statsmodels
+- sysroot_linux-aarch64==2.28
+- treelite==4.4.1
+- umap-learn==0.5.6
+- xgboost>=2.1.0
+name: all_cuda-128_arch-aarch64

cpp/src/umap/umap.cuh

@@ -34,7 +34,10 @@ static const int TPB_X = 256;
 
 inline bool dispatch_to_uint64_t(int n_rows, int n_neighbors, int n_components)
 {
-  uint64_t nnz1 = static_cast<uint64_t>(n_rows) * n_neighbors;
+  // The fuzzy simplicial set graph can have at most 2 * n * n_neighbors elements after
+  // symmetrization and removal of zeroes
+  uint64_t nnz1 = 2 * static_cast<uint64_t>(n_rows) * n_neighbors;
+  // The embeddings have n * n_neighbors elements
   uint64_t nnz2 = static_cast<uint64_t>(n_rows) * n_components;
   return nnz1 > std::numeric_limits<int32_t>::max() || nnz2 > std::numeric_limits<int32_t>::max();
 }

dependencies.yaml

@@ -4,7 +4,7 @@ files:
     output: conda
     matrix:
       cuda: ["11.8", "12.8"]
-      arch: [x86_64]
+      arch: [x86_64, aarch64]
     includes:
       - common_build
       - cuda

python/cuml/cuml/cluster/dbscan.pyx

@@ -320,8 +320,8 @@ class DBSCAN(UniversalBase,
 
             # metric
             metric_parsing = {
-                "L2": DistanceType.L2SqrtUnexpanded,
-                "euclidean": DistanceType.L2SqrtUnexpanded,
+                "L2": DistanceType.L2SqrtExpanded,
+                "euclidean": DistanceType.L2SqrtExpanded,
                 "cosine": DistanceType.CosineExpanded,
                 "precomputed": DistanceType.Precomputed
             }

python/cuml/cuml/common/array_descriptor.py

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2020-2023, NVIDIA CORPORATION.
+# Copyright (c) 2020-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -66,7 +66,6 @@ def __init__(self, order="K"):
 
     def __set_name__(self, owner, name):
         self.name = name
-        setattr(owner, name + "_order", self.order)
 
     def _get_meta(
         self, instance, throw_on_missing=False

python/cuml/cuml/experimental/fil/fil.pyx

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2023-2024, NVIDIA CORPORATION.
+# Copyright (c) 2023-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -297,7 +297,8 @@ cdef class ForestInference_impl():
                 chunk_specification
             )
 
-        self.raft_proto_handle.synchronize()
+        if GlobalSettings().device_type == DeviceType.device:
+            self.raft_proto_handle.synchronize()
 
         return preds