-
Notifications
You must be signed in to change notification settings - Fork 302
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Adds parameterized benchmarks for uniform_neighbor_sampling
, updates benchmarks
dir for future additions
#3048
Adds parameterized benchmarks for uniform_neighbor_sampling
, updates benchmarks
dir for future additions
#3048
Conversation
…re_params_product() function to PLC.
…2-initial_service_benchmarks
…est to use specific set of input param combinations.
…uniform_neighbor_sampling options
… parameter for with_replacement.
…2-initial_service_benchmarks
…2-initial_service_benchmarks
…fo(), added tests to verify new meta-data is correct.
… with the server, added tests which use/test the new builtin test extensions.
…s (via rapids-pytest-benchmark plugin) easier to read, changes to allow RemoteGraph to use either a server-side Graph or PropertyGraph.
…r LocalCUDACluster for MG service, re-wrote start list generation to work with both SG and MG, added server extension for generating start list, changed scale from 24 to 23 to prevent OOM, added separate function for creating MG graphs, added debug print to show result sizes, fixes to cugraph MG Graph for incorrect column names, added util to cugraph for starting and stopping dask client.
…2-initial_service_benchmarks
…2-initial_service_benchmarks
…chmarks for different libraries.
…ir to pytest-based to avoid name conflict.
… can be more easily shared, bug fix in benchmark extension testing util.
…params, removed funcs unused in each file.
…arkers to pytest.ini, added pythonpath and test paths to pytest.ini, gave benchmark files unique names so pytest can collect them across different bench dirs.
…ception handler for loading extensions, run server from a temp dir to avoid collisions with 'cugraph' namespace package in benchmarks dir.
…2-initial_benchmark_reorg
…, added markers for each batch size.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In addition to looking this over, I was able to use the code in the PR to run cugraph-pyg benchmarks, so it is all good. 👍
…2-initial_benchmark_reorg
Codecov ReportBase: 58.86% // Head: 58.50% // Decreases project coverage by
Additional details and impacted files@@ Coverage Diff @@
## branch-23.02 #3048 +/- ##
================================================
- Coverage 58.86% 58.50% -0.37%
================================================
Files 131 133 +2
Lines 7848 7895 +47
================================================
- Hits 4620 4619 -1
- Misses 3228 3276 +48
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report at Codecov. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Have requested a clarification based onMultiGraph, everything else looks good.
benchmarks/cugraph/pytest-based/bench_cugraph_uniform_neighbor_sample.py
Outdated
Show resolved
Hide resolved
benchmarks/cugraph/pytest-based/bench_cugraph_uniform_neighbor_sample.py
Outdated
Show resolved
Hide resolved
graph_props = pylibcugraph.experimental.GraphProperties( | ||
is_symmetric=False, is_multigraph=False) | ||
st = time.time() | ||
G = pylibcugraph.experimental.SGGraph( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
G = pylibcugraph.experimental.SGGraph( | |
G = pylibcugraph.experimental.MGGraph( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this needs to stay SGGraph
since we're not incorporating dask/MPI/etc. anywhere to distribute the data to separate MGGraphs
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Small additon
…2-initial_benchmark_reorg
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
rerun tests reason: PG test failures should be fixed from this merged PR. |
from @BradReesWork offline:
will address those when a PR for this branch is pushed next. |
@gpucibot merge |
closes #3034
This PR updates the top-level
benchmarks
dir to better organize benchmarks (and future benchmarks) based on package name. Additional parameterized benchmarks foruniform_neighbor_sampling
were added as well as re-usable fixtures and pytest param objects and utilities.Other updates include:
genFixteurParamsProduct
and refactored so only one copy is used for both cugraph and PLCThe "breaking" label is used because the
benchmarks
dir re-org changes the location of existing benchmarks, which could break scripts that call them.