def test_mg_edge_betweenness_centrality(
graph_file,
directed,
subset_size,
normalized,
weight,
subset_seed,
result_dtype,
mg_device_count,
):
prepare_test()
skip_if_not_enough_devices(mg_device_count)
with MGContext(number_of_devices=mg_device_count, p2p=True):
sorted_df = calc_edge_betweenness_centrality(
graph_file,
directed=directed,
normalized=normalized,
k=subset_size,
weight=weight,
seed=subset_seed,
result_dtype=result_dtype,
multi_gpu_batch=True,
)
compare_scores(
sorted_df,
first_key="cu_bc",
second_key="ref_bc",
epsilon=DEFAULT_EPSILON,
)
def test_enable_batch_adjlist_replication_weights(graph_file, directed,
mg_device_count):
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(graph_file,
delimiter=' ',
names=['src', 'dst', 'value'],
dtype=['int32', 'int32', 'float32'])
G = cugraph.DiGraph() if directed else cugraph.Graph()
G.from_cudf_edgelist(df,
source='src',
destination='dst',
edge_attr='value')
with MGContext(mg_device_count):
G.enable_batch()
G.view_adj_list()
adjlist = G.adjlist
offsets = adjlist.offsets
indices = adjlist.indices
weights = adjlist.weights
for worker in G.batch_adjlists:
(rep_offsets, rep_indices, rep_weights) = G.batch_adjlists[worker]
assert offsets.equals(rep_offsets.result()), "Replication of " \
"adjlist offsets failed"
assert indices.equals(rep_indices.result()), "Replication of " \
"adjlist indices failed"
assert weights.equals(rep_weights.result()), "Replication of " \
"adjlist weights failed"
def test_enable_batch_adjlist_replication_no_weights(graph_file, directed,
mg_device_count):
gc.collect()
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(
graph_file,
delimiter=" ",
names=["src", "dst"],
dtype=["int32", "int32"],
)
G = cugraph.DiGraph() if directed else cugraph.Graph()
G.from_cudf_edgelist(df, source="src", destination="dst")
with MGContext(number_of_devices=mg_device_count, p2p=True):
G.enable_batch()
G.view_adj_list()
adjlist = G.adjlist
offsets = adjlist.offsets
indices = adjlist.indices
weights = adjlist.weights
for worker in G.batch_adjlists:
(rep_offsets, rep_indices, rep_weights) = G.batch_adjlists[worker]
assert offsets.equals(
rep_offsets.result()), ("Replication of "
"adjlist offsets failed")
assert indices.equals(
rep_indices.result()), ("Replication of "
"adjlist indices failed")
assert weights is None and rep_weights is None
def test_enable_batch_edgelist_replication(graph_file, directed,
mg_device_count):
skip_if_not_enough_devices(mg_device_count)
G = utils.generate_cugraph_graph_from_file(graph_file, directed)
with MGContext(mg_device_count):
G.enable_batch()
df = G.edgelist.edgelist_df
for worker in G.batch_edgelists:
replicated_df = G.batch_edgelists[worker].result()
assert df.equals(replicated_df), "Replication of edgelist failed"
def test_replicate_cudf_dataframe_no_weights(input_data_path, mg_device_count):
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(input_data_path,
delimiter=' ',
names=['src', 'dst'],
dtype=['int32', 'int32'])
with MGContext(mg_device_count):
worker_to_futures = replication.replicate_cudf_dataframe(df)
for worker in worker_to_futures:
replicated_df = worker_to_futures[worker].result()
assert df.equals(replicated_df), "There is a mismatch in one " \
"of the replications"
def test_enable_batch_context_no_context_views(graph_file, directed,
mg_device_count):
skip_if_not_enough_devices(mg_device_count)
G = utils.generate_cugraph_graph_from_file(graph_file, directed)
with MGContext(mg_device_count):
assert G.batch_enabled is False, "Internal property should be False"
G.enable_batch()
assert G.batch_enabled is True, "Internal property should be True"
assert G.batch_edgelists is not None, "The graph should have " \
"been created with an " \
"edgelist"
G.view_edge_list()
G.view_adj_list()
G.view_transposed_adj_list()
def test_replicate_cudf_dataframe_no_weights(input_data_path, mg_device_count):
gc.collect()
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(
input_data_path,
delimiter=" ",
names=["src", "dst"],
dtype=["int32", "int32"],
)
with MGContext(number_of_devices=mg_device_count, p2p=True):
worker_to_futures = replication.replicate_cudf_dataframe(df)
for worker in worker_to_futures:
replicated_df = worker_to_futures[worker].result()
assert df.equals(replicated_df), ("There is a mismatch in one "
"of the replications")
def test_replicate_cudf_series(input_data_path, mg_device_count):
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(input_data_path,
delimiter=' ',
names=['src', 'dst', 'value'],
dtype=['int32', 'int32', 'float32'])
with MGContext(mg_device_count):
for column in df.columns.values:
series = df[column]
worker_to_futures = replication.replicate_cudf_series(series)
for worker in worker_to_futures:
replicated_series = worker_to_futures[worker].result()
assert series.equals(replicated_series), "There is a " \
"mismatch in one of the replications"
# FIXME: If we do not clear this dictionary, when comparing
# results for the 2nd column, one of the workers still
# has a value from the 1st column
worker_to_futures = {}
def test_enable_batch_context_then_views(graph_file, directed,
mg_device_count):
gc.collect()
skip_if_not_enough_devices(mg_device_count)
G = utils.generate_cugraph_graph_from_file(graph_file, directed)
with MGContext(number_of_devices=mg_device_count, p2p=True):
assert G.batch_enabled is False, "Internal property should be False"
G.enable_batch()
assert G.batch_enabled is True, "Internal property should be True"
assert G.batch_edgelists is not None, ("The graph should have "
"been created with an "
"edgelist")
assert G.batch_adjlists is None
G.view_adj_list()
assert G.batch_adjlists is not None
assert G.batch_transposed_adjlists is None
G.view_transposed_adj_list()
assert G.batch_transposed_adjlists is not None
def test_replicate_cudf_series(input_data_path, mg_device_count):
gc.collect()
skip_if_not_enough_devices(mg_device_count)
df = cudf.read_csv(
input_data_path,
delimiter=" ",
names=["src", "dst", "value"],
dtype=["int32", "int32", "float32"],
)
with MGContext(number_of_devices=mg_device_count, p2p=True):
for column in df.columns.values:
series = df[column]
worker_to_futures = replication.replicate_cudf_series(series)
for worker in worker_to_futures:
replicated_series = worker_to_futures[worker].result()
assert series.equals(replicated_series), (
"There is a "
"mismatch in one of the replications")
# FIXME: If we do not clear this dictionary, when comparing
# results for the 2nd column, one of the workers still
# has a value from the 1st column
worker_to_futures = {}