Forward Merge branch-24.06 into branch-24.08

.devcontainer/Dockerfile

@@ -7,6 +7,11 @@ FROM ${BASE} as pip-base
 
 ENV DEFAULT_VIRTUAL_ENV=rapids
 
+RUN apt update -y \
+ && DEBIAN_FRONTEND=noninteractive apt install -y \
+    libblas-dev liblapack-dev \
+ && rm -rf /tmp/* /var/tmp/* /var/cache/apt/* /var/lib/apt/lists/*;
+
 FROM ${BASE} as conda-base
 
 ENV DEFAULT_CONDA_ENV=rapids

.devcontainer/cuda11.8-conda/devcontainer.json

@@ -8,6 +8,11 @@
       "BASE": "rapidsai/devcontainers:24.08-cpp-cuda11.8-mambaforge-ubuntu22.04"
     }
   },
+  "runArgs": [
+    "--rm",
+    "--name",
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-conda"
+  ],
   "hostRequirements": {"gpu": "optional"},
   "features": {
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.8": {}

.devcontainer/cuda11.8-pip/devcontainer.json

@@ -5,14 +5,16 @@
     "args": {
       "CUDA": "11.8",
       "PYTHON_PACKAGE_MANAGER": "pip",
-      "BASE": "rapidsai/devcontainers:24.08-cpp-cuda11.8-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.08-cpp-cuda11.8-ucx1.15.0-openmpi-ubuntu22.04"
     }
   },
+  "runArgs": [
+    "--rm",
+    "--name",
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda11.8-pip"
+  ],
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/ucx:24.8": {
-      "version": "1.14.1"
-    },
     "ghcr.io/rapidsai/devcontainers/features/cuda:24.8": {
       "version": "11.8",
       "installcuBLAS": true,
@@ -23,7 +25,6 @@
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.8": {}
   },
   "overrideFeatureInstallOrder": [
-    "ghcr.io/rapidsai/devcontainers/features/ucx",
     "ghcr.io/rapidsai/devcontainers/features/cuda",
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"
   ],

.devcontainer/cuda12.2-conda/devcontainer.json

@@ -8,6 +8,11 @@
       "BASE": "rapidsai/devcontainers:24.08-cpp-mambaforge-ubuntu22.04"
     }
   },
+  "runArgs": [
+    "--rm",
+    "--name",
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-conda"
+  ],
   "hostRequirements": {"gpu": "optional"},
   "features": {
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.8": {}

.devcontainer/cuda12.2-pip/devcontainer.json

@@ -5,14 +5,16 @@
     "args": {
       "CUDA": "12.2",
       "PYTHON_PACKAGE_MANAGER": "pip",
-      "BASE": "rapidsai/devcontainers:24.08-cpp-cuda12.2-ubuntu22.04"
+      "BASE": "rapidsai/devcontainers:24.08-cpp-cuda12.2-ucx1.15.0-openmpi-ubuntu22.04"
     }
   },
+  "runArgs": [
+    "--rm",
+    "--name",
+    "${localEnv:USER:anon}-rapids-${localWorkspaceFolderBasename}-24.08-cuda12.2-pip"
+  ],
   "hostRequirements": {"gpu": "optional"},
   "features": {
-    "ghcr.io/rapidsai/devcontainers/features/ucx:24.8": {
-      "version": "1.14.1"
-    },
     "ghcr.io/rapidsai/devcontainers/features/cuda:24.8": {
       "version": "12.2",
       "installcuBLAS": true,
@@ -23,7 +25,6 @@
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils:24.8": {}
   },
   "overrideFeatureInstallOrder": [
-    "ghcr.io/rapidsai/devcontainers/features/ucx",
     "ghcr.io/rapidsai/devcontainers/features/cuda",
     "ghcr.io/rapidsai/devcontainers/features/rapids-build-utils"
   ],

.github/workflows/pr.yaml

@@ -196,5 +196,5 @@ jobs:
       extra-repo-deploy-key: CUGRAPH_OPS_SSH_PRIVATE_DEPLOY_KEY
       build_command: |
         sccache -z;
-        build-all --verbose -j$(nproc --ignore=1);
+        build-all --verbose -j$(nproc --ignore=1) -DBUILD_CUGRAPH_MG_TESTS=ON;
         sccache -s;

benchmarks/cugraph/standalone/bulk_sampling/cugraph_bulk_sampling.py

@@ -344,7 +344,7 @@ def generate_rmat_dataset(
     del label_df
     gc.collect()
 
-    dask_label_df = dask_cudf.from_dask_dataframe(dask_label_df)
+    dask_label_df = dask_label_df.to_backend("cudf")
 
     node_offsets = {"paper": 0}
     edge_offsets = {("paper", "cites", "paper"): 0}

benchmarks/nx-cugraph/pytest-based/bench_algos.py

@@ -848,6 +848,23 @@ def bench_weakly_connected_components(benchmark, graph_obj, backend_wrapper):
     assert type(result) is list
 
 
+def bench_ego_graph(benchmark, graph_obj, backend_wrapper):
+    G = get_graph_obj_for_benchmark(graph_obj, backend_wrapper)
+    node = get_highest_degree_node(graph_obj)
+    result = benchmark.pedantic(
+        target=backend_wrapper(nx.ego_graph),
+        args=(G,),
+        kwargs=dict(
+            n=node,
+            radius=100,
+        ),
+        rounds=rounds,
+        iterations=iterations,
+        warmup_rounds=warmup_rounds,
+    )
+    assert isinstance(result, (nx.Graph, nxcg.Graph))
+
+
 @pytest.mark.skip(reason="benchmark not implemented")
 def bench_complete_bipartite_graph(benchmark, graph_obj, backend_wrapper):
     pass

ci/build_wheel.sh

@@ -40,7 +40,7 @@ if ! rapids-is-release-build; then
     alpha_spec=',>=0.0.0a0'
 fi
 
-for dep in rmm cudf cugraph raft-dask pylibcugraph pylibcugraphops pylibraft ucx-py; do
+for dep in rmm cudf cugraph raft-dask pylibcugraph pylibcugraphops pylibwholegraph pylibraft ucx-py; do
     sed -r -i "s/${dep}==(.*)\"/${dep}${PACKAGE_CUDA_SUFFIX}==\1${alpha_spec}\"/g" ${pyproject_file}
 done
 

ci/release/update-version.sh

@@ -105,6 +105,7 @@ find .devcontainer/ -type f -name devcontainer.json -print0 | while IFS= read -r
     sed_runner "s@rapidsai/devcontainers/features/ucx:[0-9.]*@rapidsai/devcontainers/features/ucx:${NEXT_SHORT_TAG_PEP440}@" "${filename}"
     sed_runner "s@rapidsai/devcontainers/features/cuda:[0-9.]*@rapidsai/devcontainers/features/cuda:${NEXT_SHORT_TAG_PEP440}@" "${filename}"
     sed_runner "s@rapidsai/devcontainers/features/rapids-build-utils:[0-9.]*@rapidsai/devcontainers/features/rapids-build-utils:${NEXT_SHORT_TAG_PEP440}@" "${filename}"
+    sed_runner "s@rapids-\${localWorkspaceFolderBasename}-[0-9.]*@rapids-\${localWorkspaceFolderBasename}-${NEXT_SHORT_TAG}@g" "${filename}"
 done
 
 sed_runner "s/:[0-9][0-9]\.[0-9][0-9]/:${NEXT_SHORT_TAG}/" ./notebooks/README.md

ci/run_cugraph_pyg_pytests.sh

@@ -6,7 +6,10 @@ set -euo pipefail
 # Support invoking run_cugraph_pyg_pytests.sh outside the script directory
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../python/cugraph-pyg/cugraph_pyg
 
-pytest --cache-clear --ignore=tests/mg "$@" .
+pytest --cache-clear --benchmark-disable "$@" .
+
+# Used to skip certain examples in CI due to memory limitations
+export CI_RUN=1
 
 # Test examples
 for e in "$(pwd)"/examples/*.py; do

ci/test.sh

@@ -103,7 +103,7 @@ if hasArg "--run-python-tests"; then
     conda list
     cd ${CUGRAPH_ROOT}/python/cugraph-pyg/cugraph_pyg
     # rmat is not tested because of MG testing
-    pytest --cache-clear --junitxml=${CUGRAPH_ROOT}/junit-cugraph-pytests.xml -v --cov-config=.coveragerc --cov=cugraph_pyg --cov-report=xml:${WORKSPACE}/python/cugraph_pyg/cugraph-coverage.xml --cov-report term --ignore=raft --ignore=tests/mg --ignore=tests/int --ignore=tests/generators --benchmark-disable
+    pytest -sv -m sg --cache-clear --junitxml=${CUGRAPH_ROOT}/junit-cugraph-pytests.xml -v --cov-config=.coveragerc --cov=cugraph_pyg --cov-report=xml:${WORKSPACE}/python/cugraph_pyg/cugraph-coverage.xml --cov-report term --ignore=raft --benchmark-disable
     echo "Ran Python pytest for cugraph_pyg : return code was: $?, test script exit code is now: $EXITCODE"
 
     echo "Python pytest for cugraph-service (single-GPU only)..."

ci/test_python.sh

@@ -3,10 +3,6 @@
 
 set -euo pipefail
 
-# TODO: Enable dask query planning (by default) once some bugs are fixed.
-# xref: https://github.com/rapidsai/cudf/issues/15027
-export DASK_DATAFRAME__QUERY_PLANNING=False
-
 # Support invoking test_python.sh outside the script directory
 cd "$(dirname "$(realpath "${BASH_SOURCE[0]}")")"/../
 
@@ -217,13 +213,14 @@ if [[ "${RAPIDS_CUDA_VERSION}" == "11.8.0" ]]; then
 
     # Install pyg dependencies (which requires pip)
 
-    pip install ogb
+    pip install \
+      ogb \
+      tensordict
+
     pip install \
         pyg_lib \
         torch_scatter \
         torch_sparse \
-        torch_cluster \
-        torch_spline_conv \
       -f ${PYG_URL}
 
     rapids-print-env

ci/test_wheel.sh

@@ -3,10 +3,6 @@
 
 set -eoxu pipefail
 
-# TODO: Enable dask query planning (by default) once some bugs are fixed.
-# xref: https://github.com/rapidsai/cudf/issues/15027
-export DASK_DATAFRAME__QUERY_PLANNING=False
-
 package_name=$1
 package_dir=$2
 

ci/test_wheel_cugraph-dgl.sh

@@ -32,8 +32,18 @@ fi
 PYTORCH_URL="https://download.pytorch.org/whl/cu${PYTORCH_CUDA_VER}"
 DGL_URL="https://data.dgl.ai/wheels/cu${PYTORCH_CUDA_VER}/repo.html"
 
+# Starting from 2.2, PyTorch wheels depend on nvidia-nccl-cuxx>=2.19 wheel and
+# dynamically link to NCCL. RAPIDS CUDA 11 CI images have an older NCCL version that
+# might shadow the newer NCCL required by PyTorch during import (when importing
+# `cupy` before `torch`).
+if [[ "${NCCL_VERSION}" < "2.19" ]]; then
+  PYTORCH_VER="2.1.0"
+else
+  PYTORCH_VER="2.3.0"
+fi
+
 rapids-logger "Installing PyTorch and DGL"
-rapids-retry python -m pip install torch --index-url ${PYTORCH_URL}
+rapids-retry python -m pip install "torch==${PYTORCH_VER}" --index-url ${PYTORCH_URL}
 rapids-retry python -m pip install dgl==2.0.0 --find-links ${DGL_URL}
 
 python -m pytest python/cugraph-dgl/tests

ci/test_wheel_cugraph-pyg.sh

@@ -24,6 +24,9 @@ python -m pip install $(ls ./dist/${python_package_name}*.whl)[test]
 # RAPIDS_DATASET_ROOT_DIR is used by test scripts
 export RAPIDS_DATASET_ROOT_DIR="$(realpath datasets)"
 
+# Used to skip certain examples in CI due to memory limitations
+export CI_RUN=1
+
 if [[ "${CUDA_VERSION}" == "11.8.0" ]]; then
   PYTORCH_URL="https://download.pytorch.org/whl/cu118"
   PYG_URL="https://data.pyg.org/whl/torch-2.1.0+cu118.html"
@@ -39,15 +42,14 @@ rapids-retry python -m pip install \
   pyg_lib \
   torch_scatter \
   torch_sparse \
-  torch_cluster \
-  torch_spline_conv \
+  tensordict \
   -f ${PYG_URL}
 
 rapids-logger "pytest cugraph-pyg (single GPU)"
 pushd python/cugraph-pyg/cugraph_pyg
 python -m pytest \
   --cache-clear \
-  --ignore=tests/mg \
+  --benchmark-disable \
   tests
 # Test examples
 for e in "$(pwd)"/examples/*.py; do

conda/recipes/cugraph-pyg/meta.yaml

@@ -34,6 +34,7 @@ requirements:
     - cupy >=12.0.0
     - cugraph ={{ version }}
     - pylibcugraphops ={{ minor_version }}
+    - tensordict >=0.1.2
     - pyg >=2.5,<2.6
 
 tests:

cpp/CMakeLists.txt

@@ -92,14 +92,14 @@ set(CUGRAPH_CXX_FLAGS "")
 set(CUGRAPH_CUDA_FLAGS "")
 
 if(CMAKE_COMPILER_IS_GNUCXX)
-    list(APPEND CUGRAPH_CXX_FLAGS -Werror -Wno-error=deprecated-declarations)
+    list(APPEND CUGRAPH_CXX_FLAGS -Werror -Wno-error=deprecated-declarations -Wno-deprecated-declarations -DRAFT_HIDE_DEPRECATION_WARNINGS)
 endif(CMAKE_COMPILER_IS_GNUCXX)
 
 
 message("-- Building for GPU_ARCHS = ${CMAKE_CUDA_ARCHITECTURES}")
 
 list(APPEND CUGRAPH_CUDA_FLAGS --expt-extended-lambda --expt-relaxed-constexpr)
-list(APPEND CUGRAPH_CUDA_FLAGS -Werror=cross-execution-space-call -Wno-deprecated-declarations -Xptxas=--disable-warnings)
+list(APPEND CUGRAPH_CUDA_FLAGS -Werror=cross-execution-space-call -Wno-deprecated-declarations -DRAFT_HIDE_DEPRECATION_WARNINGS -Xptxas=--disable-warnings)
 list(APPEND CUGRAPH_CUDA_FLAGS -Xcompiler=-Wall,-Wno-error=sign-compare,-Wno-error=unused-but-set-variable)
 list(APPEND CUGRAPH_CUDA_FLAGS -Xfatbin=-compress-all)
 
@@ -180,6 +180,7 @@ set(CUGRAPH_SOURCES
     src/community/detail/refine_sg.cu
     src/community/detail/refine_mg.cu
     src/community/edge_triangle_count_sg.cu
+    src/community/edge_triangle_count_mg.cu
     src/community/detail/maximal_independent_moves_sg.cu
     src/community/detail/maximal_independent_moves_mg.cu
     src/detail/utility_wrappers.cu
@@ -288,6 +289,8 @@ set(CUGRAPH_SOURCES
     src/structure/symmetrize_edgelist_mg.cu
     src/community/triangle_count_sg.cu
     src/community/triangle_count_mg.cu
+    src/community/approx_weighted_matching_sg.cu
+    src/community/approx_weighted_matching_mg.cu
     src/traversal/k_hop_nbrs_sg.cu
     src/traversal/k_hop_nbrs_mg.cu
     src/mtmg/vertex_result.cu

cpp/include/cugraph/algorithms.hpp

@@ -2007,6 +2007,24 @@ void triangle_count(raft::handle_t const& handle,
                     raft::device_span<edge_t> counts,
                     bool do_expensive_check = false);
 
+/*
+ * @brief Compute edge triangle counts.
+ *
+ * Compute edge triangle counts for the entire set of edges.
+ *
+ * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
+ * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
+ * @tparam multi_gpu Flag indicating whether template instantiation should target single-GPU (false)
+ * @param handle RAFT handle object to encapsulate resources (e.g. CUDA stream, communicator, and
+ * handles to various CUDA libraries) to run graph algorithms.
+ * @param graph_view Graph view object.
+ *
+ * @return edge_property_t containing the edge triangle count
+ */
+template <typename vertex_t, typename edge_t, bool multi_gpu>
+edge_property_t<graph_view_t<vertex_t, edge_t, false, multi_gpu>, edge_t> edge_triangle_count(
+  raft::handle_t const& handle, graph_view_t<vertex_t, edge_t, false, multi_gpu> const& graph_view);
+
 /*
  * @brief Compute K-Truss.
  *
@@ -2368,6 +2386,32 @@ rmm::device_uvector<vertex_t> vertex_coloring(
   graph_view_t<vertex_t, edge_t, false, multi_gpu> const& graph_view,
   raft::random::RngState& rng_state);
 
+/*
+ * @brief Approximate Weighted Matching
+ *
+ * A matching in an undirected graph G = (V, E) is a pairing of adjacent vertices
+ * such that each vertex is matched with at most one other vertex, the objective
+ * being to match as many vertices as possible or to maximise the sum of the
+ * weights of the matched edges. Here we provide an implementation of an
+ * approximation algorithm to the weighted Maximum matching. See
+ * https://web.archive.org/web/20081031230449id_/http://www.ii.uib.no/~fredrikm/fredrik/papers/CP75.pdf
+ * for further information.
+ *
+ * @tparam vertex_t Type of vertex identifiers. Needs to be an integral type.
+ * @tparam edge_t Type of edge identifiers. Needs to be an integral type.
+ * @tparam multi_gpu Flag indicating whether template instantiation should target single-GPU (false)
+ * @param[in] handle RAFT handle object to encapsulate resources (e.g. CUDA stream, communicator,
+ * and handles to various CUDA libraries) to run graph algorithms.
+ * @param[in] graph_view Graph view object.
+ * @param[in] edge_weight_view View object holding edge weights for @p graph_view.
+ * @return A tuple of device vector of matched vertex ids and sum of the weights of the matched
+ * edges.
+ */
+template <typename vertex_t, typename edge_t, typename weight_t, bool multi_gpu>
+std::tuple<rmm::device_uvector<vertex_t>, weight_t> approximate_weighted_matching(
+  raft::handle_t const& handle,
+  graph_view_t<vertex_t, edge_t, false, multi_gpu> const& graph_view,
+  edge_property_view_t<edge_t, weight_t const*> edge_weight_view);
 }  // namespace cugraph
 
 /**