Merge pull request #1970 from rapidsai/branch-21.12

[RELEASE] cugraph v21.12

CHANGELOG.md

@@ -1,6 +1,91 @@
-# cuGraph 21.10.00 (Date TBD)
+# cuGraph 21.12.00 (Date TBD)
 
-Please see https://github.com/rapidsai/cugraph/releases/tag/v21.10.00a for the latest changes to this development branch.
+Please see https://github.com/rapidsai/cugraph/releases/tag/v21.12.00a for the latest changes to this development branch.
+
+# cuGraph 21.10.00 (7 Oct 2021)
+
+## 🚨 Breaking Changes
+
+- remove tsp implementation from 21.10 ([#1812](https://github.com/rapidsai/cugraph/pull/1812)) [@ChuckHastings](https://github.com/ChuckHastings)
+- multi seeds BFS with one seed per component ([#1591](https://github.com/rapidsai/cugraph/pull/1591)) [@afender](https://github.com/afender)
+
+## 🐛 Bug Fixes
+
+- make_zip_iterator should be on a make_tuple ([#1857](https://github.com/rapidsai/cugraph/pull/1857)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Removed NetworkX requirement for type checks, fixed docstring, added new docstrings, import cleanups ([#1853](https://github.com/rapidsai/cugraph/pull/1853)) [@rlratzel](https://github.com/rlratzel)
+- Temporarily disable input argument checks for a currently disabled feature ([#1840](https://github.com/rapidsai/cugraph/pull/1840)) [@seunghwak](https://github.com/seunghwak)
+- Changed value of the expensive check param to `false` in `populate_graph_container` ([#1839](https://github.com/rapidsai/cugraph/pull/1839)) [@rlratzel](https://github.com/rlratzel)
+- Accommodate cudf change to is_string_dtype method ([#1827](https://github.com/rapidsai/cugraph/pull/1827)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Changed code to disable `k_truss` on CUDA 11.4 differently ([#1811](https://github.com/rapidsai/cugraph/pull/1811)) [@rlratzel](https://github.com/rlratzel)
+- Clean-up artifacts from the multi-source BFS PR ([#1591) (#1804](https://github.com/rapidsai/cugraph/pull/1591) (#1804)) [@seunghwak](https://github.com/seunghwak)
+- MG WCC bug fix ([#1802](https://github.com/rapidsai/cugraph/pull/1802)) [@seunghwak](https://github.com/seunghwak)
+- Fix MG Louvain test compile errors ([#1797](https://github.com/rapidsai/cugraph/pull/1797)) [@seunghwak](https://github.com/seunghwak)
+- force_atlas2 to support nx hypercube_graph ([#1779](https://github.com/rapidsai/cugraph/pull/1779)) [@jnke2016](https://github.com/jnke2016)
+- Bug louvain reverted fix ([#1766](https://github.com/rapidsai/cugraph/pull/1766)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Bug dask cudf personalization ([#1764](https://github.com/rapidsai/cugraph/pull/1764)) [@Iroy30](https://github.com/Iroy30)
+
+## 📖 Documentation
+
+- updated to new doc theme ([#1793](https://github.com/rapidsai/cugraph/pull/1793)) [@BradReesWork](https://github.com/BradReesWork)
+- Change python docs to pydata theme ([#1785](https://github.com/rapidsai/cugraph/pull/1785)) [@galipremsagar](https://github.com/galipremsagar)
+- Initial doc update for running the python E2E benchmarks in a MNMG environment. ([#1781](https://github.com/rapidsai/cugraph/pull/1781)) [@rlratzel](https://github.com/rlratzel)
+
+## 🚀 New Features
+
+- C++ benchmarking for additional algorithms ([#1762](https://github.com/rapidsai/cugraph/pull/1762)) [@seunghwak](https://github.com/seunghwak)
+
+## 🛠️ Improvements
+
+- Updating cuco to latest ([#1859](https://github.com/rapidsai/cugraph/pull/1859)) [@BradReesWork](https://github.com/BradReesWork)
+- fix benchmark exit status ([#1850](https://github.com/rapidsai/cugraph/pull/1850)) [@jnke2016](https://github.com/jnke2016)
+- add try/catch for python-louvain ([#1842](https://github.com/rapidsai/cugraph/pull/1842)) [@BradReesWork](https://github.com/BradReesWork)
+- Pin max dask and distributed versions to 2021.09.1 ([#1841](https://github.com/rapidsai/cugraph/pull/1841)) [@galipremsagar](https://github.com/galipremsagar)
+- add compiler version checks to cmake to fail early ([#1836](https://github.com/rapidsai/cugraph/pull/1836)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Make sure we keep the rapids-cmake and cugraph cal version in sync ([#1830](https://github.com/rapidsai/cugraph/pull/1830)) [@robertmaynard](https://github.com/robertmaynard)
+- Remove obsolete file ([#1829](https://github.com/rapidsai/cugraph/pull/1829)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Improve memory scaling for low average vertex degree graphs & many GPUs ([#1823](https://github.com/rapidsai/cugraph/pull/1823)) [@seunghwak](https://github.com/seunghwak)
+- Added the reduction op input parameter to host_scalar_(all)reduce utility functions. ([#1822](https://github.com/rapidsai/cugraph/pull/1822)) [@seunghwak](https://github.com/seunghwak)
+- Count if e test ([#1821](https://github.com/rapidsai/cugraph/pull/1821)) [@kaatish](https://github.com/kaatish)
+- Added Sorensen algorithm to Python API ([#1820](https://github.com/rapidsai/cugraph/pull/1820)) [@jnke2016](https://github.com/jnke2016)
+- Updated to enforce only supported dtypes, changed to use legacy connected_components API ([#1817](https://github.com/rapidsai/cugraph/pull/1817)) [@rlratzel](https://github.com/rlratzel)
+- Group return values of renumber_edgelist and input parameters of graph_t & graph_view_t constructors. ([#1816](https://github.com/rapidsai/cugraph/pull/1816)) [@seunghwak](https://github.com/seunghwak)
+- remove tsp implementation from 21.10 ([#1812](https://github.com/rapidsai/cugraph/pull/1812)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Changed pylibcugraph connected_components APIs to use duck typing for CAI inputs, added doc placeholders ([#1810](https://github.com/rapidsai/cugraph/pull/1810)) [@rlratzel](https://github.com/rlratzel)
+- Add new new raft symlink path to .gitignore ([#1808](https://github.com/rapidsai/cugraph/pull/1808)) [@trxcllnt](https://github.com/trxcllnt)
+- Initial version of `pylibcugraph` conda package and CI build script updates ([#1806](https://github.com/rapidsai/cugraph/pull/1806)) [@rlratzel](https://github.com/rlratzel)
+- Also building cpp MG tests as part of conda/CI libcugraph builds ([#1805](https://github.com/rapidsai/cugraph/pull/1805)) [@rlratzel](https://github.com/rlratzel)
+- Split many files to separate SG from MG template instantiations ([#1803](https://github.com/rapidsai/cugraph/pull/1803)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Graph primitives memory scaling improvements for low average vertex degree graphs and many GPUs (Part 1) ([#1801](https://github.com/rapidsai/cugraph/pull/1801)) [@seunghwak](https://github.com/seunghwak)
+- Pylibcugraph connected components ([#1800](https://github.com/rapidsai/cugraph/pull/1800)) [@Iroy30](https://github.com/Iroy30)
+- Transform Reduce E test ([#1798](https://github.com/rapidsai/cugraph/pull/1798)) [@kaatish](https://github.com/kaatish)
+- Update with rapids cmake new features ([#1790](https://github.com/rapidsai/cugraph/pull/1790)) [@robertmaynard](https://github.com/robertmaynard)
+- Update thrust/RMM deprecated calls ([#1789](https://github.com/rapidsai/cugraph/pull/1789)) [@dantegd](https://github.com/dantegd)
+- Update UCX-Py to 0.22 ([#1788](https://github.com/rapidsai/cugraph/pull/1788)) [@pentschev](https://github.com/pentschev)
+- Initial version of `pylibcugraph` source tree and build script updates ([#1787](https://github.com/rapidsai/cugraph/pull/1787)) [@rlratzel](https://github.com/rlratzel)
+- Fix Forward-Merge Conflicts ([#1786](https://github.com/rapidsai/cugraph/pull/1786)) [@ajschmidt8](https://github.com/ajschmidt8)
+- add conda environment for CUDA 11.4 ([#1784](https://github.com/rapidsai/cugraph/pull/1784)) [@seunghwak](https://github.com/seunghwak)
+- Temporarily pin RMM while refactor removes deprecated calls ([#1775](https://github.com/rapidsai/cugraph/pull/1775)) [@dantegd](https://github.com/dantegd)
+- MNMG memory footprint improvement for low average vertex degree graphs (part 2) ([#1774](https://github.com/rapidsai/cugraph/pull/1774)) [@seunghwak](https://github.com/seunghwak)
+- Fix unused variables/parameters warnings ([#1772](https://github.com/rapidsai/cugraph/pull/1772)) [@seunghwak](https://github.com/seunghwak)
+- MNMG memory footprint improvement for low average vertex degree graphs (part 1) ([#1769](https://github.com/rapidsai/cugraph/pull/1769)) [@seunghwak](https://github.com/seunghwak)
+- Transform reduce v test ([#1768](https://github.com/rapidsai/cugraph/pull/1768)) [@kaatish](https://github.com/kaatish)
+- Move experimental source files and a few implementation headers ([#1763](https://github.com/rapidsai/cugraph/pull/1763)) [@ChuckHastings](https://github.com/ChuckHastings)
+- updating notebooks ([#1761](https://github.com/rapidsai/cugraph/pull/1761)) [@BradReesWork](https://github.com/BradReesWork)
+- consolidate tests to use the fixture dask_client ([#1758](https://github.com/rapidsai/cugraph/pull/1758)) [@jnke2016](https://github.com/jnke2016)
+- Move all new graph objects out of experimental namespace ([#1757](https://github.com/rapidsai/cugraph/pull/1757)) [@ChuckHastings](https://github.com/ChuckHastings)
+- C++ benchmarking for MG PageRank ([#1755](https://github.com/rapidsai/cugraph/pull/1755)) [@seunghwak](https://github.com/seunghwak)
+- Move legacy implementations into legacy directories ([#1752](https://github.com/rapidsai/cugraph/pull/1752)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Remove hardcoded Pagerank dtype ([#1751](https://github.com/rapidsai/cugraph/pull/1751)) [@jnke2016](https://github.com/jnke2016)
+- Add python end to end benchmark and create new directories ([#1750](https://github.com/rapidsai/cugraph/pull/1750)) [@jnke2016](https://github.com/jnke2016)
+- Modify MNMG louvain to support an empty vertex partition ([#1744](https://github.com/rapidsai/cugraph/pull/1744)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Fea renumbering test ([#1742](https://github.com/rapidsai/cugraph/pull/1742)) [@ChuckHastings](https://github.com/ChuckHastings)
+- Fix auto-merger for Branch 21.10 coming from 21.08 ([#1740](https://github.com/rapidsai/cugraph/pull/1740)) [@galipremsagar](https://github.com/galipremsagar)
+- Use the new RAPIDS.cmake to fetch rapids-cmake ([#1734](https://github.com/rapidsai/cugraph/pull/1734)) [@robertmaynard](https://github.com/robertmaynard)
+- Biased Random Walks for GNN ([#1732](https://github.com/rapidsai/cugraph/pull/1732)) [@aschaffer](https://github.com/aschaffer)
+- Updated MG python tests to run in single and multi-node environments ([#1731](https://github.com/rapidsai/cugraph/pull/1731)) [@rlratzel](https://github.com/rlratzel)
+- ENH Replace gpuci_conda_retry with gpuci_mamba_retry ([#1720](https://github.com/rapidsai/cugraph/pull/1720)) [@dillon-cullinan](https://github.com/dillon-cullinan)
+- Apply modifications to account for RAFT changes ([#1707](https://github.com/rapidsai/cugraph/pull/1707)) [@viclafargue](https://github.com/viclafargue)
+- multi seeds BFS with one seed per component ([#1591](https://github.com/rapidsai/cugraph/pull/1591)) [@afender](https://github.com/afender)
 
 # cuGraph 21.08.00 (4 Aug 2021)
 
@@ -224,7 +309,6 @@ Please see https://github.com/rapidsai/cugraph/releases/tag/v21.10.00a for the l
 - Create labeler.yml (#1318) @jolorunyomi
 - Updates to support nightly MG test automation (#1308) @rlratzel
 - Add C++ graph functions (coarsen_grpah, renumber_edgelist, relabel) and primitvies (transform_reduce_by_adj_matrix_row_key, transform_reduce_by_adj_matrix_col_key, copy_v_transform_reduce_key_aggregated_out_nbr) (#1257) @seunghwak
->>>>>>> upstream/branch-0.18
 
 # cuGraph 0.17.0 (10 Dec 2020)
 ## New Features

README.md

@@ -43,6 +43,7 @@ As of Release 21.08 - including 21.08 nightly
 
 
 ## Supported Algorithms
+_Italic_ algorithms are planned for future releases.
 
 | Category     | Algorithm                              | Scale        |  Notes              |
 | ------------ | -------------------------------------- | ------------ | ------------------- |
@@ -66,6 +67,10 @@ As of Release 21.08 - including 21.08 nightly
 | Core         |                                        |              |                     |
 |              | K-Core                                 | Single-GPU   |                     |
 |              | Core Number                            | Single-GPU   |                     |
+| _Flow_       |                                        |              |                     |
+|              | _MaxFlow_                              | ---          |                     |
+| _Influence_  |                                        |              |                     |
+|              | _Influence Maximization_               | ---          |                     |
 | Layout       |                                        |              |                     |
 |              | Force Atlas 2                          | Single-GPU   |                     |
 | Linear Assignment|                                    |              |                     |
@@ -78,8 +83,11 @@ As of Release 21.08 - including 21.08 nightly
 |              | Jaccard Similarity                     | Single-GPU   |                     |
 |              | Weighted Jaccard Similarity            | Single-GPU   |                     |
 |              | Overlap Similarity                     | Single-GPU   |                     |
+|              | Sorensen Coefficient                   | Single-GPU   |                     |
+|              | _Local Clustering Coefficient_         |   ---        |                     |
 | Sampling     |                                        |              |                     |
-|              | Random Walks (RW)                      | Single-GPU   |                     |
+|              | Random Walks (RW)                      | Single-GPU   | Biased and Uniform  |
+|              | _node2vec_                             |   ---        |                    |
 | Traversal    |                                        |              |                     |
 |              | Breadth First Search (BFS)             | Multi-GPU    | with cutoff support <br/> [C++ README](cpp/src/traversal/README.md#BFS) |
 |              | Single Source Shortest Path (SSSP)     | Multi-GPU    | [C++ README](cpp/src/traversal/README.md#SSSP) |
@@ -89,16 +97,18 @@ As of Release 21.08 - including 21.08 nightly
 | Other        |                                        |              |                     |
 |              | Renumbering                            | Multi-GPU    | multiple columns, any data type  |
 |              | Symmetrize                             | Multi-GPU    |                     |
+| Data Generator  |                                     |              |                     |
+|              | RMAT                                   | Multi-GPU    |                     |
+|              | _Barabasi-Albert_                      |  ---         |                     |
 |  |  |
 
 </br></br>
 ## Supported Graph
 | Type            |  Description                                        |
 | --------------- | --------------------------------------------------- |
-| Graph           | An undirected Graph                                 |
-| DiGraph         | A Directed Graph                                    |
+| Graph           | An undirected Graph is default                      |
+|                 | directed=True yields a Directed Graph               |
 | Multigraph      | A Graph with multiple edges between a vertex pair   |
-| MultiDigraph    | A Directed Graph with multiple edges between a vertex pair   |
 |  |  |
 
 ALL Algorithms support Graphs and MultiGraph (directed and undirected)
@@ -156,6 +166,9 @@ conda install -c nvidia -c rapidsai -c numba -c conda-forge cugraph cudatoolkit=
 
 # CUDA 11.2
 conda install -c nvidia -c rapidsai -c numba -c conda-forge cugraph cudatoolkit=11.2
+
+# CUDA 11.4
+conda install -c nvidia -c rapidsai -c numba -c conda-forge cugraph cudatoolkit=11.4
 ```
 
 Note: This conda installation only applies to Linux and Python versions 3.7/3.8.

benchmarks/python_e2e/cugraph_dask_funcs.py

@@ -23,6 +23,7 @@
 from cugraph.comms import comms as Comms
 from cugraph.dask.common.mg_utils import get_visible_devices
 from cugraph.generators import rmat
+import tempfile
 
 
 def generate_edgelist(scale,
@@ -165,8 +166,12 @@ def setup(dask_scheduler_file=None):
         client = Client(scheduler_file=dask_scheduler_file)
 
     else:
-        cluster = LocalCUDACluster()
+        tempdir_object = tempfile.TemporaryDirectory()
+        cluster = LocalCUDACluster(local_directory=tempdir_object.name)
         client = Client(cluster)
+        # add the obj to the client so it doesn't get deleted until
+        # the 'client' obj gets cleaned up
+        client.tempdir_object = tempdir_object
         client.wait_for_workers(len(get_visible_devices()))
 
     Comms.initialize(p2p=True)

benchmarks/python_e2e/main.py

@@ -23,6 +23,31 @@
 import cugraph_dask_funcs
 from benchmark import BenchmarkRun
 
+import json
+from pathlib import Path
+
+
+def store_results_json(benchmark_dir=None,
+                       algo_name=None,
+                       algo_time=None,
+                       n_gpus=None,
+                       scale=None):
+    """
+    Store all benchmark results in json files
+    """
+    benchmark_result = {}
+    benchmark_result['funcName'] = algo_name
+    benchmark_result['result'] = algo_time
+    benchmark_result['argNameValuePairs'] = [('scale', scale), ('ngpus', n_gpus)]
+
+    json_object = json.dumps(benchmark_result, indent=4)
+
+    benchmark_dir_path = Path(benchmark_dir)
+
+    with open(f"{benchmark_dir_path}/benchmark_result_scale_{scale}_ngpus_{n_gpus}_{algo_name}.json", "w") as outfile:
+        outfile.write(json_object)
+
+
 
 def log(s, end="\n"):
     print(s, end=end)
@@ -36,6 +61,7 @@ def run(algos,
         unweighted=False,
         symmetric=False,
         edgefactor=None,
+        benchmark_dir=None,
         dask_scheduler_file=None):
     """
     Run the nightly benchmark on cugraph.
@@ -47,6 +73,9 @@ def run(algos,
         funcs = cugraph_funcs
     else:
         funcs = cugraph_dask_funcs
+        # For MNMG, find the number of GPUS after the client and
+        # and the cluster started
+        n_gpus = None
 
     # Setup the benchmarks to run based on algos specified, or all.
     # Values are either callables, or tuples of (callable, args) pairs.
@@ -72,6 +101,11 @@ def run(algos,
     # output files/reports, etc.
     log("calling setup...", end="")
     setup_objs = funcs.setup(dask_scheduler_file)
+
+    # If the number of GPUs is None, This is a MNMG run
+    # Extract the number of gpus from the client
+    if n_gpus is None:
+        n_gpus = len(setup_objs[0].scheduler_info()['workers'])
     log("done.")
 
     try:
@@ -95,6 +129,13 @@ def run(algos,
                                 )
         success = benchmark.run()
 
+        algo_name = benchmark.results[1].name
+        algo_name = f"benchmarks.{algo_name}"
+        algo_time = benchmark.results[1].runtime
+        # Generate json files containing the benchmark results
+        if benchmark_dir is not None:
+            store_results_json(benchmark_dir, algo_name, algo_time, n_gpus, scale)
+
         # Report results
         print(generate_console_report(benchmark.results))
         if csv_results_file:
@@ -136,6 +177,9 @@ def run(algos,
     ap.add_argument("--edgefactor", type=int, default=16,
                     help="edge factor for the graph edgelist generator "
                     "(num_edges=num_verts*EDGEFACTOR).")
+    ap.add_argument("--benchmark-dir", type=str, default=None,
+                    help="directory to store the results in json files")
+
     args = ap.parse_args()
 
     exitcode = run(algos=args.algo,
@@ -145,6 +189,7 @@ def run(algos,
                    unweighted=args.unweighted,
                    symmetric=args.symmetric_graph,
                    edgefactor=args.edgefactor,
+                   benchmark_dir=args.benchmark_dir,
                    dask_scheduler_file=args.dask_scheduler_file)
 
     sys.exit(exitcode)

benchmarks/python_pytest_based/bench_algos.py

@@ -37,7 +37,6 @@ def setFixtureParamNames(*args, **kwargs):
 
 from .params import FIXTURE_PARAMS
 
-
 ###############################################################################
 # Helpers
 def createGraph(csvFileName, graphType=None):