Merge branch 'branch-24.12' into branch-24.12-update-bench-algos

.pre-commit-config.yaml

@@ -68,7 +68,7 @@ repos:
         types: [python]
         language: python
         pass_filenames: false
-        additional_dependencies: ["networkx>=3.3"]
+        additional_dependencies: ["networkx>=3.4"]
   - repo: local
     hooks:
       - id: nx-cugraph-readme-update
@@ -78,4 +78,4 @@ repos:
         types_or: [python, markdown]
         language: python
         pass_filenames: false
-        additional_dependencies: ["networkx>=3.3"]
+        additional_dependencies: ["networkx>=3.4"]

ci/test_wheel.sh

@@ -4,7 +4,6 @@
 set -eoxu pipefail
 
 package_name=$1
-package_dir=$2
 
 python_package_name=$(echo ${package_name}|sed 's/-/_/g')
 

ci/test_wheel_cugraph-dgl.sh

@@ -4,24 +4,16 @@
 set -eoxu pipefail
 
 package_name="cugraph-dgl"
-package_dir="python/cugraph-dgl"
-
-python_package_name=$(echo ${package_name}|sed 's/-/_/g')
 
 mkdir -p ./dist
 RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 
-# Download wheels built during this job.
+# Download the pylibcugraph, cugraph, and cugraph-dgl built in the previous step
 RAPIDS_PY_WHEEL_NAME="pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-deps
 RAPIDS_PY_WHEEL_NAME="cugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-deps
-python -m pip install ./local-deps/*.whl
-
-# use 'ls' to expand wildcard before adding `[extra]` requires for pip
 RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-download-wheels-from-s3 ./dist
-# pip creates wheels using python package names
-python -m pip install $(ls ./dist/${python_package_name}*.whl)[test]
-
 
+# determine pytorch and DGL sources
 PKG_CUDA_VER="$(echo ${CUDA_VERSION} | cut -d '.' -f1,2 | tr -d '.')"
 PKG_CUDA_VER_MAJOR=${PKG_CUDA_VER:0:2}
 if [[ "${PKG_CUDA_VER_MAJOR}" == "12" ]]; then
@@ -32,8 +24,15 @@ fi
 PYTORCH_URL="https://download.pytorch.org/whl/cu${PYTORCH_CUDA_VER}"
 DGL_URL="https://data.dgl.ai/wheels/torch-2.3/cu${PYTORCH_CUDA_VER}/repo.html"
 
-rapids-logger "Installing PyTorch and DGL"
-rapids-retry python -m pip install torch==2.3.0 --index-url ${PYTORCH_URL}
-rapids-retry python -m pip install dgl==2.4.0 --find-links ${DGL_URL}
+# echo to expand wildcard before adding `[extra]` requires for pip
+python -m pip install \
+    -v \
+    --extra-index-url "${PYTORCH_URL}" \
+    --find-links "${DGL_URL}" \
+    "$(echo ./local-deps/pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
+    "$(echo ./local-deps/cugraph_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
+    "$(echo ./dist/cugraph_dgl_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
+    'dgl==2.4.0' \
+    'torch>=2.3.0,<2.4'
 
 python -m pytest python/cugraph-dgl/tests

ci/test_wheel_cugraph-equivariant.sh

@@ -4,19 +4,14 @@
 set -eoxu pipefail
 
 package_name="cugraph-equivariant"
-package_dir="python/cugraph-equivariant"
-
-python_package_name=$(echo ${package_name}|sed 's/-/_/g')
 
 mkdir -p ./dist
 RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 
-# use 'ls' to expand wildcard before adding `[extra]` requires for pip
+# Download the cugraph-equivariant built in the previous step
 RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-download-wheels-from-s3 ./dist
-# pip creates wheels using python package names
-python -m pip install $(ls ./dist/${python_package_name}*.whl)[test]
-
 
+# determine pytorch source
 PKG_CUDA_VER="$(echo ${CUDA_VERSION} | cut -d '.' -f1,2 | tr -d '.')"
 PKG_CUDA_VER_MAJOR=${PKG_CUDA_VER:0:2}
 if [[ "${PKG_CUDA_VER_MAJOR}" == "12" ]]; then
@@ -26,8 +21,12 @@ else
 fi
 PYTORCH_URL="https://download.pytorch.org/whl/cu${PYTORCH_CUDA_VER}"
 
-rapids-logger "Installing PyTorch and e3nn"
-rapids-retry python -m pip install torch --index-url ${PYTORCH_URL}
-rapids-retry python -m pip install e3nn
+# echo to expand wildcard before adding `[extra]` requires for pip
+python -m pip install \
+    -v \
+    --extra-index-url "${PYTORCH_URL}" \
+    "$(echo ./dist/cugraph_equivariant_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
+    'e3nn' \
+    'torch>=2.3.0,<2.4'
 
 python -m pytest python/cugraph-equivariant/cugraph_equivariant/tests

ci/test_wheel_cugraph-pyg.sh

@@ -4,45 +4,44 @@
 set -eoxu pipefail
 
 package_name="cugraph-pyg"
-package_dir="python/cugraph-pyg"
-
-python_package_name=$(echo ${package_name}|sed 's/-/_/g')
 
 mkdir -p ./dist
 RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 
-# Download wheels built during this job.
+# Download the pylibcugraph, cugraph, and cugraph-pyg built in the previous step
 RAPIDS_PY_WHEEL_NAME="pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-deps
 RAPIDS_PY_WHEEL_NAME="cugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-deps
-python -m pip install ./local-deps/*.whl
-
-# use 'ls' to expand wildcard before adding `[extra]` requires for pip
 RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-download-wheels-from-s3 ./dist
-# pip creates wheels using python package names
-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
 
+# determine pytorch and pyg sources
 if [[ "${CUDA_VERSION}" == "11.8.0" ]]; then
   PYTORCH_URL="https://download.pytorch.org/whl/cu118"
   PYG_URL="https://data.pyg.org/whl/torch-2.3.0+cu118.html"
 else
   PYTORCH_URL="https://download.pytorch.org/whl/cu121"
   PYG_URL="https://data.pyg.org/whl/torch-2.3.0+cu121.html"
 fi
-rapids-logger "Installing PyTorch and PyG dependencies"
-rapids-retry python -m pip install torch==2.3.0 --index-url ${PYTORCH_URL}
-rapids-retry python -m pip install "torch-geometric>=2.5,<2.6"
-rapids-retry python -m pip install \
-  ogb \
-  pyg_lib \
-  torch_scatter \
-  torch_sparse \
-  -f ${PYG_URL}
+
+# echo to expand wildcard before adding `[extra]` requires for pip
+python -m pip install \
+    -v \
+    --extra-index-url "${PYTORCH_URL}" \
+    --find-links "${PYG_URL}" \
+    "$(echo ./local-deps/pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
+    "$(echo ./local-deps/cugraph_${RAPIDS_PY_CUDA_SUFFIX}*.whl)" \
+    "$(echo ./dist/cugraph_pyg_${RAPIDS_PY_CUDA_SUFFIX}*.whl)[test]" \
+    'ogb' \
+    'pyg_lib' \
+    'torch>=2.3.0,<2.4' \
+    'torch-geometric>=2.5,<2.6' \
+    'torch_scatter' \
+    'torch_sparse'
+
+# RAPIDS_DATASET_ROOT_DIR is used by test scripts
+export RAPIDS_DATASET_ROOT_DIR="$(realpath datasets)"
+
+# Used to skip certain examples in CI due to memory limitations
+export CI_RUN=1
 
 rapids-logger "pytest cugraph-pyg (single GPU)"
 pushd python/cugraph-pyg/cugraph_pyg

ci/test_wheel_cugraph.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 set -eoxu pipefail
 
@@ -8,4 +8,4 @@ RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 RAPIDS_PY_WHEEL_NAME="pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-pylibcugraph-dep
 python -m pip install --no-deps ./local-pylibcugraph-dep/pylibcugraph*.whl
 
-./ci/test_wheel.sh cugraph python/cugraph
+./ci/test_wheel.sh cugraph

ci/test_wheel_nx-cugraph.sh

@@ -8,4 +8,4 @@ RAPIDS_PY_CUDA_SUFFIX="$(rapids-wheel-ctk-name-gen ${RAPIDS_CUDA_VERSION})"
 RAPIDS_PY_WHEEL_NAME="pylibcugraph_${RAPIDS_PY_CUDA_SUFFIX}" rapids-download-wheels-from-s3 ./local-deps
 python -m pip install ./local-deps/*.whl
 
-./ci/test_wheel.sh nx-cugraph python/nx-cugraph
+./ci/test_wheel.sh nx-cugraph

ci/test_wheel_pylibcugraph.sh

@@ -1,6 +1,6 @@
 #!/bin/bash
-# Copyright (c) 2023, NVIDIA CORPORATION.
+# Copyright (c) 2023-2024, NVIDIA CORPORATION.
 
 set -eoxu pipefail
 
-./ci/test_wheel.sh pylibcugraph python/pylibcugraph
+./ci/test_wheel.sh pylibcugraph

cpp/src/detail/graph_partition_utils.cuh

@@ -43,7 +43,7 @@ struct compute_gpu_id_from_ext_vertex_t {
 
   __host__ __device__ int operator()(vertex_t v) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     auto vertex_partition_id = static_cast<int>(hash_func(v) % comm_size);
     return partition_manager::compute_global_comm_rank_from_vertex_partition_id(
       major_comm_size, minor_comm_size, vertex_partition_id);
@@ -58,7 +58,7 @@ struct compute_gpu_id_from_ext_edge_id_t {
 
   __host__ __device__ int operator()(edge_t e) const
   {
-    cuco::detail::MurmurHash3_32<edge_t> hash_func{};
+    cuco::murmurhash3_32<edge_t> hash_func{};
     auto vertex_partition_id = static_cast<int>(hash_func(e) % comm_size);
     return partition_manager::compute_global_comm_rank_from_vertex_partition_id(
       major_comm_size, minor_comm_size, vertex_partition_id);
@@ -88,7 +88,7 @@ struct compute_vertex_partition_id_from_ext_vertex_t {
 
   __host__ __device__ int operator()(vertex_t v) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return hash_func(v) % comm_size;
   }
 };
@@ -114,7 +114,7 @@ struct compute_gpu_id_from_ext_edge_endpoints_t {
 
   __host__ __device__ int operator()(vertex_t major, vertex_t minor) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     auto major_vertex_partition_id = static_cast<int>(hash_func(major) % comm_size);
     auto minor_vertex_partition_id = static_cast<int>(hash_func(minor) % comm_size);
     auto major_comm_rank           = major_vertex_partition_id % major_comm_size;
@@ -126,7 +126,7 @@ struct compute_gpu_id_from_ext_edge_endpoints_t {
   __host__ __device__ int operator()(
     thrust::tuple<vertex_t, vertex_t> pair /* major, minor */) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     auto major_vertex_partition_id = static_cast<int>(hash_func(thrust::get<0>(pair)) % comm_size);
     auto minor_vertex_partition_id = static_cast<int>(hash_func(thrust::get<1>(pair)) % comm_size);
     auto major_comm_rank           = major_vertex_partition_id % major_comm_size;
@@ -192,15 +192,15 @@ struct compute_edge_partition_id_from_ext_edge_endpoints_t {
 
   __host__ __device__ int operator()(vertex_t major, vertex_t minor) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return (hash_func(major) % comm_size) * minor_comm_size +
            (hash_func(minor) % comm_size) / major_comm_size;
   }
 
   __host__ __device__ int operator()(
     thrust::tuple<vertex_t, vertex_t> pair /* major, minor */) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return (hash_func(thrust::get<0>(pair)) % comm_size) * minor_comm_size +
            (hash_func(thrust::get<1>(pair)) % comm_size) / major_comm_size;
   }

cpp/src/structure/remove_multi_edges_impl.cuh

@@ -28,6 +28,7 @@
 
 #include <rmm/device_uvector.hpp>
 
+#include <cuda/std/cstddef>
 #include <thrust/binary_search.h>
 #include <thrust/distance.h>
 #include <thrust/iterator/zip_iterator.h>
@@ -53,8 +54,8 @@ struct hash_src_dst_pair {
     vertex_t pair[2];
     pair[0] = thrust::get<0>(t);
     pair[1] = thrust::get<1>(t);
-    cuco::detail::MurmurHash3_32<vertex_t*> hash_func{};
-    return hash_func.compute_hash(reinterpret_cast<std::byte*>(pair), 2 * sizeof(vertex_t)) %
+    cuco::murmurhash3_32<vertex_t*> hash_func{};
+    return hash_func.compute_hash(reinterpret_cast<cuda::std::byte*>(pair), 2 * sizeof(vertex_t)) %
            num_groups;
   }
 };

cpp/tests/prims/mg_count_if_v.cu

@@ -48,7 +48,7 @@ struct test_predicate {
   test_predicate(int mod_count) : mod(mod_count) {}
   __device__ bool operator()(vertex_t, const vertex_t& val)
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return (0 == (hash_func(val) % mod));
   }
 };

cpp/tests/prims/mg_per_v_pair_transform_dst_nbr_intersection.cu

@@ -137,7 +137,7 @@ class Tests_MGPerVPairTransformDstNbrIntersection
       cugraph::get_dataframe_buffer_begin(mg_vertex_pair_buffer),
       cugraph::get_dataframe_buffer_end(mg_vertex_pair_buffer),
       [comm_rank, num_vertices = mg_graph_view.number_of_vertices()] __device__(size_t i) {
-        cuco::detail::MurmurHash3_32<size_t>
+        cuco::murmurhash3_32<size_t>
           hash_func{};  // use hash_func to generate arbitrary vertex pairs
         auto v0 = static_cast<vertex_t>(hash_func(i + comm_rank) % num_vertices);
         auto v1 = static_cast<vertex_t>(hash_func(i + num_vertices + comm_rank) % num_vertices);

cpp/tests/prims/mg_per_v_pair_transform_dst_nbr_weighted_intersection.cu

@@ -163,7 +163,7 @@ class Tests_MGPerVPairTransformDstNbrIntersection
       cugraph::get_dataframe_buffer_begin(mg_vertex_pair_buffer),
       cugraph::get_dataframe_buffer_end(mg_vertex_pair_buffer),
       [comm_rank, num_vertices = mg_graph_view.number_of_vertices()] __device__(size_t i) {
-        cuco::detail::MurmurHash3_32<size_t>
+        cuco::murmurhash3_32<size_t>
           hash_func{};  // use hash_func to generate arbitrary vertex pairs
         auto v0 = static_cast<vertex_t>(hash_func(i + comm_rank) % num_vertices);
         auto v1 = static_cast<vertex_t>(hash_func(i + num_vertices + comm_rank) % num_vertices);

cpp/tests/prims/mg_transform_reduce_v.cu

@@ -53,7 +53,7 @@ struct v_op_t {
 
   __device__ auto operator()(vertex_t, vertex_t val) const
   {
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return cugraph::test::detail::make_property_value<property_t>(hash_func(val) % mod);
   }
 };

cpp/tests/utilities/property_generator_kernels.cuh

@@ -60,7 +60,7 @@ struct vertex_property_transform {
   {
     static_assert(cugraph::is_thrust_tuple_of_arithmetic<property_t>::value ||
                   std::is_arithmetic_v<property_t>);
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return make_property_value<property_t>(hash_func(v) % mod);
   }
 };
@@ -74,7 +74,7 @@ struct edge_property_transform {
   {
     static_assert(cugraph::is_thrust_tuple_of_arithmetic<property_t>::value ||
                   std::is_arithmetic_v<property_t>);
-    cuco::detail::MurmurHash3_32<vertex_t> hash_func{};
+    cuco::murmurhash3_32<vertex_t> hash_func{};
     return make_property_value<property_t>(hash_func(src + dst) % mod);
   }
 };

python/cugraph/cugraph/structure/graph_primtypes.pxd

@@ -1,4 +1,4 @@
-# Copyright (c) 2019-2023, NVIDIA CORPORATION.
+# Copyright (c) 2019-2024, 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
@@ -21,7 +21,7 @@ from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport pair
 from libcpp.vector cimport vector
 from pylibraft.common.handle cimport *
-from rmm._lib.device_buffer cimport device_buffer
+from rmm.librmm.device_buffer cimport device_buffer
 
 cdef extern from "cugraph/legacy/graph.hpp" namespace "cugraph::legacy":
 

python/cugraph/cugraph/structure/graph_primtypes.pyx

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2023, NVIDIA CORPORATION.
+# Copyright (c) 2020-2024, 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
@@ -20,7 +20,7 @@ import numpy as np
 from libc.stdint cimport uintptr_t
 from libcpp.utility cimport move
 
-from rmm._lib.device_buffer cimport DeviceBuffer
+from rmm.pylibrmm.device_buffer cimport DeviceBuffer
 from cudf.core.buffer import as_buffer
 import cudf
 

python/cugraph/cugraph/structure/graph_utilities.pxd

@@ -21,7 +21,7 @@ from libcpp.memory cimport unique_ptr
 from libcpp.utility cimport pair
 from libcpp.vector cimport vector
 
-from rmm._lib.device_buffer cimport device_buffer
+from rmm.librmm.device_buffer cimport device_buffer
 
 from pylibraft.common.handle cimport handle_t
 

python/nx-cugraph/_nx_cugraph/__init__.py

@@ -36,7 +36,7 @@
     "backend_name": "cugraph",
     "project": "nx-cugraph",
     "package": "nx_cugraph",
-    "url": f"https://rapids.ai/nx-cugraph",
+    "url": "https://rapids.ai/nx-cugraph",
     "short_summary": "GPU-accelerated backend.",
     # "description": "TODO",
     "functions": {
@@ -180,7 +180,7 @@
         "ego_graph": "Weighted ego_graph with negative cycles is not yet supported. `NotImplementedError` will be raised if there are negative `distance` edge weights.",
         "eigenvector_centrality": "`nstart` parameter is not used, but it is checked for validity.",
         "from_pandas_edgelist": "cudf.DataFrame inputs also supported; value columns with str is unsuppported.",
-        "generic_bfs_edges": "`neighbors` and `sort_neighbors` parameters are not yet supported.",
+        "generic_bfs_edges": "`neighbors` parameter is not yet supported.",
         "katz_centrality": "`nstart` isn't used (but is checked), and `normalized=False` is not supported.",
         "louvain_communities": "`seed` parameter is currently ignored, and self-loops are not yet supported.",
         "pagerank": "`dangling` parameter is not supported, but it is checked for validity.",