Skip to content
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.

Sign up for GitHub

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

[REVIEW] dask personalization, fix df query #1237

Merged
merged 16 commits into from
Nov 20, 2020
Merged

[REVIEW] dask personalization, fix df query #1237

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
16 commits
Select commit Hold shift + click to select a range
dda79ba
dask personalization, fix df query
Iroy30 Oct 20, 2020
2bb3010
test updates for asymmetric graphs
Iroy30 Oct 26, 2020
25ef46d
changelog
Iroy30 Oct 26, 2020
cf0a80e
raise exception
Iroy30 Oct 26, 2020
47d2cd4
flake8
Iroy30 Oct 30, 2020
3782296
Merge branch 'branch-0.17' into fix_bugs
Iroy30 Oct 30, 2020
926c052
Remove DiGraph test
Iroy30 Nov 4, 2020
3770a2c
Remove DiGraph test
Iroy30 Nov 4, 2020
514e540
Update karate_undirected.csv
Iroy30 Nov 4, 2020
4efb203
Update and rename asymmetric_directed__tiny.csv to karate-asymmetric.csv
Iroy30 Nov 4, 2020
0acb7b7
fixed conflicts
Iroy30 Nov 10, 2020
7ca833b
merge branch 0.17
Iroy30 Nov 10, 2020
4fbaa7c
update tests
Iroy30 Nov 20, 2020
d69de72
update notebooks
Iroy30 Nov 20, 2020
73b42e4
flake8:
Iroy30 Nov 20, 2020
2e88e12
Merge branch 'branch-0.17' into fix_bugs
Iroy30 Nov 20, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions python/cugraph/dask/link_analysis/mg_pagerank_wrapper.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,8 +99,8 @@ def mg_pagerank(input_df,

if personalization is not None:
sz = personalization['vertex'].shape[0]
personalization['vertex'] = personalization['vertex'].astype(np.int32)
personalization['values'] = personalization['values'].astype(df['pagerank'].dtype)
personalization['vertex'] = personalization['vertex'].astype(vertex_t)
personalization['values'] = personalization['values'].astype(weight_t)
c_pers_vtx = personalization['vertex'].__cuda_array_interface__['data'][0]
c_pers_val = personalization['values'].__cuda_array_interface__['data'][0]

Expand Down
48 changes: 30 additions & 18 deletions python/cugraph/dask/link_analysis/pagerank.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,9 +119,6 @@ def pagerank(input_graph,
"""
from cugraph.structure.graph import null_check

if personalization is not None:
raise Exception("Personalization not supported")

nstart = None

client = default_client()
Expand All @@ -141,21 +138,36 @@ def pagerank(input_graph,
if input_graph.renumbered is True:
personalization = input_graph.add_internal_vertex_id(
personalization, "vertex", "vertex"
).compute()

result = [client.submit(call_pagerank,
Comms.get_session_id(),
wf[1],
num_verts,
num_edges,
vertex_partition_offsets,
alpha,
max_iter,
tol,
personalization,
nstart,
workers=[wf[0]])
for idx, wf in enumerate(data.worker_to_parts.items())]
)
p_data = get_distributed_data(personalization)

result = [client.submit(call_pagerank,
Comms.get_session_id(),
wf[1],
num_verts,
num_edges,
vertex_partition_offsets,
alpha,
max_iter,
tol,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we need to discuss more about tol.

The code below is what NetworkX does (compare err with N * tol). This requires setting tol to a smaller value if N gets large (tol here basically means tolerance for a single PageRank value not the entire set of PageRank vector).

https://github.com/networkx/networkx/blob/master/networkx/algorithms/link_analysis/pagerank_alg.py#L155

err = sum([abs(x[n] - xlast[n]) for n in x])
        if err < N * tol:
            return x

The new PageRank algorithm adopts this NetworkX logic to determine convergence, so should we follow the NetworkX logic (and set tol to a smaller value with a larger N) or better remove N * from the new PageRank code?

p_data.worker_to_parts[wf[0]][0],
nstart,
workers=[wf[0]])
for idx, wf in enumerate(data.worker_to_parts.items())]
else:
result = [client.submit(call_pagerank,
Comms.get_session_id(),
wf[1],
num_verts,
num_edges,
vertex_partition_offsets,
alpha,
max_iter,
tol,
personalization,
nstart,
workers=[wf[0]])
for idx, wf in enumerate(data.worker_to_parts.items())]
wait(result)
ddf = dask_cudf.from_delayed(result)
if input_graph.renumbered:
Expand Down
4 changes: 2 additions & 2 deletions python/cugraph/structure/graph.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1081,7 +1081,7 @@ def degrees(self, vertex_subset=None):
df = self.unrenumber(df, "vertex")

if vertex_subset is not None:
df = df.query("`vertex` in @vertex_subset")
df = df[df['vertex'].isin(vertex_subset)]

return df

Expand All @@ -1095,7 +1095,7 @@ def _degree(self, vertex_subset, x=0):
df = self.unrenumber(df, "vertex")

if vertex_subset is not None:
df = df.query("`vertex` in @vertex_subset")
df = df[df['vertex'].isin(vertex_subset)]

return df

Expand Down
2 changes: 1 addition & 1 deletion python/cugraph/tests/dask/test_mg_pagerank.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,7 @@ def personalize(v, personalization_perc):
return cu_personalization


PERSONALIZATION_PERC = [0]
PERSONALIZATION_PERC = [0, 50]


@pytest.fixture
Expand Down