udpate branch

.pre-commit-config.yaml

@@ -17,7 +17,7 @@ repos:
     hooks:
       - id: black
         language_version: python3
-        args: [--target-version=py39]
+        args: [--target-version=py310]
         files: ^(python/.*|benchmarks/.*)$
         exclude: ^python/nx-cugraph/
   - repo: https://github.com/PyCQA/flake8
@@ -42,7 +42,7 @@ repos:
         types_or: [c, c++, cuda]
         args: ["-fallback-style=none", "-style=file", "-i"]
   - repo: https://github.com/rapidsai/pre-commit-hooks
-    rev: v0.3.1
+    rev: v0.4.0
     hooks:
       - id: verify-copyright
         files: |

benchmarks/cugraph/pytest-based/bench_cugraph_uniform_neighbor_sample.py

@@ -266,7 +266,7 @@ def uns_func(*args, **kwargs):
 @pytest.mark.managedmem_off
 @pytest.mark.poolallocator_on
 @pytest.mark.parametrize("batch_size", params.batch_sizes.values())
-@pytest.mark.parametrize("fanout", [params.fanout_10_25, params.fanout_5_10_15])
+@pytest.mark.parametrize("fanout", [params.fanout_10_25])
 @pytest.mark.parametrize(
     "with_replacement", [False], ids=lambda v: f"with_replacement={v}"
 )
@@ -287,6 +287,8 @@ def bench_cugraph_uniform_neighbor_sample(
         start_list=uns_args["start_list"],
         fanout_vals=uns_args["fanout"],
         with_replacement=uns_args["with_replacement"],
+        use_legacy_names=False,
+        with_edge_properties=True,
     )
     """
     dtmap = {"int32": 32 // 8, "int64": 64 // 8}

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -42,6 +42,7 @@ dependencies:
 - numpy>=1.23,<2.0a0
 - numpydoc
 - nvcc_linux-64=11.8
+- ogb
 - openmpi
 - packaging>=21
 - pandas
@@ -74,6 +75,7 @@ dependencies:
 - sphinxcontrib-websupport
 - thriftpy2!=0.5.0,!=0.5.1
 - torchdata
+- torchmetrics
 - ucx-proc=*=gpu
 - ucx-py==0.40.*,>=0.0.0a0
 - wget

conda/environments/all_cuda-125_arch-x86_64.yaml

@@ -47,6 +47,7 @@ dependencies:
 - numba>=0.57
 - numpy>=1.23,<2.0a0
 - numpydoc
+- ogb
 - openmpi
 - packaging>=21
 - pandas
@@ -79,6 +80,7 @@ dependencies:
 - sphinxcontrib-websupport
 - thriftpy2!=0.5.0,!=0.5.1
 - torchdata
+- torchmetrics
 - ucx-proc=*=gpu
 - ucx-py==0.40.*,>=0.0.0a0
 - wget

conda/recipes/nx-cugraph/meta.yaml

@@ -14,9 +14,7 @@ source:
 
 build:
   number: {{ GIT_DESCRIBE_NUMBER }}
-  build:
-      number: {{ GIT_DESCRIBE_NUMBER }}
-      string: py{{ py_version }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
+  string: py{{ py_version }}_{{ date_string }}_{{ GIT_DESCRIBE_HASH }}_{{ GIT_DESCRIBE_NUMBER }}
 
 requirements:
   host:

cpp/examples/developers/graph_operations/graph_operations.cu

@@ -131,7 +131,7 @@ create_graph(raft::handle_t const& handle,
   //
 
   if (multi_gpu) {
-    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore) =
+    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore, std::ignore) =
       cugraph::shuffle_external_edges<vertex_t, vertex_t, weight_t, int32_t>(handle,
                                                                              std::move(d_edge_srcs),
                                                                              std::move(d_edge_dsts),
@@ -215,10 +215,10 @@ void perform_example_graph_operations(
                                                                                   graph_view);
 
     cugraph::update_edge_src_property(
-      handle, graph_view, vertex_weights.begin(), src_vertex_weights_cache);
+      handle, graph_view, vertex_weights.begin(), src_vertex_weights_cache.mutable_view());
 
     cugraph::update_edge_dst_property(
-      handle, graph_view, vertex_weights.begin(), dst_vertex_weights_cache);
+      handle, graph_view, vertex_weights.begin(), dst_vertex_weights_cache.mutable_view());
 
     rmm::device_uvector<result_t> weighted_averages(
       size_of_the_vertex_partition_assigned_to_this_process, handle.get_stream());
@@ -259,10 +259,10 @@ void perform_example_graph_operations(
                                                                                   graph_view);
 
     cugraph::update_edge_src_property(
-      handle, graph_view, vertex_weights.begin(), src_vertex_weights_cache);
+      handle, graph_view, vertex_weights.begin(), src_vertex_weights_cache.mutable_view());
 
     cugraph::update_edge_dst_property(
-      handle, graph_view, vertex_weights.begin(), dst_vertex_weights_cache);
+      handle, graph_view, vertex_weights.begin(), dst_vertex_weights_cache.mutable_view());
 
     rmm::device_uvector<result_t> weighted_averages(
       size_of_the_vertex_partition_assigned_to_this_process, handle.get_stream());

cpp/examples/developers/vertex_and_edge_partition/vertex_and_edge_partition.cu

@@ -127,7 +127,7 @@ create_graph(raft::handle_t const& handle,
   //
 
   if (multi_gpu) {
-    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore) =
+    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore, std::ignore) =
       cugraph::shuffle_external_edges<vertex_t, vertex_t, weight_t, int32_t>(handle,
                                                                              std::move(d_edge_srcs),
                                                                              std::move(d_edge_dsts),

cpp/examples/users/multi_gpu_application/mg_graph_algorithms.cpp

@@ -123,7 +123,7 @@ create_graph(raft::handle_t const& handle,
   //
 
   if (multi_gpu) {
-    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore) =
+    std::tie(d_edge_srcs, d_edge_dsts, d_edge_wgts, std::ignore, std::ignore, std::ignore) =
       cugraph::shuffle_external_edges<vertex_t, vertex_t, weight_t, int32_t>(handle,
                                                                              std::move(d_edge_srcs),
                                                                              std::move(d_edge_dsts),
@@ -248,9 +248,8 @@ void run_graph_algorithms(
                             std::cout);
 }
 
-int main(int argc, char** argv)
+void run_tests()
 {
-  initialize_mpi_and_set_device(argc, argv);
   std::unique_ptr<raft::handle_t> handle = initialize_mg_handle();
 
   //
@@ -279,6 +278,7 @@ int main(int argc, char** argv)
       std::move(std::make_optional(edge_wgts)),
       renumber,
       is_symmetric);
+
   // Non-owning view of the graph object
   auto graph_view = graph.view();
 
@@ -292,5 +292,14 @@ int main(int argc, char** argv)
   run_graph_algorithms<vertex_t, edge_t, weight_t, store_transposed, multi_gpu>(
     *handle, graph_view, edge_weight_view);
 
+  handle.release();
+}
+
+int main(int argc, char** argv)
+{
+  initialize_mpi_and_set_device(argc, argv);
+
+  run_tests();
+
   RAFT_MPI_TRY(MPI_Finalize());
 }

cpp/include/cugraph/graph_functions.hpp

@@ -1178,7 +1178,8 @@ std::tuple<rmm::device_uvector<vertex_t>,
            rmm::device_uvector<vertex_t>,
            std::optional<rmm::device_uvector<weight_t>>,
            std::optional<rmm::device_uvector<edge_t>>,
-           std::optional<rmm::device_uvector<edge_type_t>>>
+           std::optional<rmm::device_uvector<edge_type_t>>,
+           std::vector<size_t>>
 shuffle_external_edges(raft::handle_t const& handle,
                        rmm::device_uvector<vertex_t>&& edge_srcs,
                        rmm::device_uvector<vertex_t>&& edge_dsts,

cpp/tests/CMakeLists.txt

@@ -490,7 +490,7 @@ ConfigureTest(SAMPLING_POST_PROCESSING_TEST sampling/sampling_post_processing_te
 
 ###################################################################################################
 # - NEGATIVE SAMPLING tests --------------------------------------------------------------------
-ConfigureTest(NEGATIVE_SAMPLING_TEST sampling/negative_sampling.cpp)
+ConfigureTest(NEGATIVE_SAMPLING_TEST sampling/negative_sampling.cpp PERCENT 100)
 
 ###################################################################################################
 # - Renumber tests --------------------------------------------------------------------------------

dependencies.yaml

@@ -480,10 +480,6 @@ dependencies:
     specific:
       - output_types: [conda]
         matrices:
-          - matrix:
-              py: "3.9"
-            packages:
-              - python=3.9
           - matrix:
               py: "3.10"
             packages:
@@ -494,7 +490,7 @@ dependencies:
               - python=3.11
           - matrix:
             packages:
-              - python>=3.9,<3.12
+              - python>=3.10,<3.12
   python_build_rapids:
     common:
       - output_types: [conda, pyproject, requirements]
@@ -717,6 +713,8 @@ dependencies:
           - &pytorch_unsuffixed pytorch>=2.0,<2.2.0a0
           - torchdata
           - pydantic
+          - ogb
+          - torchmetrics
 
     specific:
       - output_types: [requirements]

docs/cugraph/source/installation/getting_cugraph.md

@@ -45,7 +45,7 @@ conda install -c rapidsai -c conda-forge -c nvidia cugraph cuda-version=12.0
 
 Alternatively, use `cuda-version=11.8` for packages supporting CUDA 11.
 
-Note: This conda installation only applies to Linux and Python versions 3.9/3.10/3.11.
+Note: This conda installation only applies to Linux and Python versions 3.10/3.11.
 
 <br>
 

docs/cugraph/source/installation/source_build.md

@@ -12,8 +12,7 @@ __Compilers:__
 * `nvcc`          version 11.5+
 
 __CUDA:__
-* CUDA 11.2+
-* NVIDIA driver 470.42.01 or newer
+* CUDA 11.8+
 * NVIDIA GPU, Volta architecture or later, with [compute capability](https://developer.nvidia.com/cuda-gpus) 7.0+
 
 Further details and download links for these prerequisites are available on the
@@ -178,7 +177,7 @@ Run either the C++ or the Python tests with datasets
    make test
    ```
 
-Note: This conda installation only applies to Linux and Python versions 3.8/3.11.
+Note: This conda installation only applies to Linux and Python versions 3.10 and 3.11.
 
 ### (OPTIONAL) Set environment variable on activation
 

docs/cugraph/source/tutorials/basic_cugraph.md

@@ -4,8 +4,8 @@
 
 CuGraph is part of [Rapids](https://docs.rapids.ai/user-guide) and has the following system requirements:
  * NVIDIA GPU, Volta architecture or later, with [compute capability](https://developer.nvidia.com/cuda-gpus) 7.0+
- * CUDA 11.2, 11.4, 11.5, 11.8, 12.0 or 12.2
- * Python version 3.9, 3.10, or 3.11
+ * CUDA 11.2, 11.4, 11.5, 11.8, 12.0, 12.2, or 12.5
+ * Python version 3.10 or 3.11
  * NetworkX >= version 3.3 or newer in order to use use [NetworkX Configs](https://networkx.org/documentation/stable/reference/backends.html#module-networkx.utils.configs) **This is required for use of nx-cuGraph, [see below](#cugraph-using-networkx-code).**
 
 ## Installation

docs/cugraph/source/tutorials/cugraph_notebooks.md

@@ -55,10 +55,9 @@ Running the example in these notebooks requires:
   * Download via Docker, Conda (See [__Getting Started__](https://rapids.ai/start.html))
 
 * cuGraph is dependent on the latest version of cuDF.  Please install all components of RAPIDS
-* Python 3.8+
-* A system with an NVIDIA GPU:  Pascal architecture or better
+* Python 3.10+
+* A system with an NVIDIA GPU: Volta architecture or newer
 * CUDA 11.4+
-* NVIDIA driver 450.51+
 
 ## Copyright
 

docs/cugraph/source/wholegraph/installation/source_build.md

@@ -16,8 +16,7 @@ __Compiler__:
 
 __CUDA__:
 * CUDA 11.8+
-* NVIDIA driver 450.80.02+
-* Pascal architecture or better
+* Volta architecture or better
 
 You can obtain CUDA from [https://developer.nvidia.com/cuda-downloads](https://developer.nvidia.com/cuda-downloads).
 
@@ -177,7 +176,7 @@ Run either the C++ or the Python tests with datasets
     ```
 
 
-Note: This conda installation only applies to Linux and Python versions 3.8/3.10.
+Note: This conda installation only applies to Linux and Python versions 3.10 and 3.11.
 
 ## Creating documentation
 

notebooks/README.md

@@ -56,24 +56,23 @@ Running the example in these notebooks requires:
   * Download via Docker, Conda (See [__Getting Started__](https://rapids.ai/start.html))
 
 * cuGraph is dependent on the latest version of cuDF.  Please install all components of RAPIDS
-* Python 3.8+
-* A system with an NVIDIA GPU:  Pascal architecture or better
+* Python 3.10+
+* A system with an NVIDIA GPU: Volta architecture or newer
 * CUDA 11.4+
-* NVIDIA driver 450.51+
 
 ### QuickStart
 
 The easiest way to run the notebooks is to get the latest [rapidsai/notebooks](https://hub.docker.com/r/rapidsai/notebooks) docker image with matching cuda version and run a container based on the image.
 
 For example, get the latest (as of writing the document) nightly image (`a` after the version number indicates that an image is nightly) with cuda 12.0 using
 ```sh
-docker pull rapidsai/notebooks:24.10a-cuda12.0-py3.9
+docker pull rapidsai/notebooks:24.10a-cuda12.0-py3.10
 ```
 
 And, then run a container based on the image using
 
 ```sh
-docker run --rm  -it --pull always --gpus all --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864   -p 8888:8888 rapidsai/notebooks:24.10a-cuda12.0-py3.9
+docker run --rm  -it --pull always --gpus all --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864   -p 8888:8888 rapidsai/notebooks:24.10a-cuda12.0-py3.10
 ```
 You are all set. Run and edit cugraph notebooks from a browser at url
 http://127.0.0.1:8888/lab/tree/cugraph/cugraph_benchmarks
@@ -89,8 +88,8 @@ ssh -L  127.0.0.1:8888:127.0.0.1:8888 [USER_NAME@][REMOTE_HOST_NAME or REMOTE_HO
 and then run the container in your remote machine.
 
 ```sh
-docker pull rapidsai/notebooks:24.10a-cuda12.0-py3.9
-docker run --rm  -it --pull always --gpus all --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 -p 8888:8888 rapidsai/notebooks:24.10a-cuda12.0-py3.9
+docker pull rapidsai/notebooks:24.10a-cuda12.0-py3.10
+docker run --rm  -it --pull always --gpus all --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 -p 8888:8888 rapidsai/notebooks:24.10a-cuda12.0-py3.10
 ```
 
 You can run and edit cugraph notebooks at url http://127.0.0.1:8888/lab/tree/cugraph/cugraph_benchmarks as if they are running locally.

notebooks/demo/nx_cugraph_demo.ipynb

@@ -20,7 +20,7 @@
     "Using `nx-cugraph` with this notebook requires the following: \n",
     "- NVIDIA GPU, Pascal architecture or later\n",
     "- CUDA 11.2, 11.4, 11.5, 11.8, or 12.0\n",
-    "- Python versions 3.9, 3.10, or 3.11\n",
+    "- Python versions 3.10 or 3.11\n",
     "- NetworkX >= version 3.2\n",
     "  - _NetworkX 3.0 supports dispatching and is compatible with `nx-cugraph`, but this notebook will demonstrate features added in 3.2_\n",
     "  - At the time of this writing, NetworkX 3.2 is only available from source and can be installed by following the [development version install instructions](https://github.com/networkx/networkx/blob/main/INSTALL.rst#install-the-development-version).\n",

python/cugraph-dgl/cugraph_dgl/dataloading/dataloader.py

@@ -140,6 +140,10 @@ def __init__(
         self.__graph = graph
         self.__device = device
 
+    @property
+    def _batch_size(self):
+        return self.__batch_size
+
     @property
     def dataset(
         self,

python/cugraph-dgl/cugraph_dgl/dataloading/neighbor_sampler.py

@@ -194,7 +194,7 @@ def sample(
 
         if g.is_homogeneous:
             indices = torch.concat(list(indices))
-            ds.sample_from_nodes(indices, batch_size=batch_size)
+            ds.sample_from_nodes(indices.long(), batch_size=batch_size)
             return HomogeneousSampleReader(
                 ds.get_reader(), self.output_format, self.edge_dir
             )