diff --git a/tests/python/pytorch/graphbolt/gb_test_utils.py b/tests/python/pytorch/graphbolt/gb_test_utils.py
index bfcb14d426e6..5b9f6b5cbd22 100644
--- a/tests/python/pytorch/graphbolt/gb_test_utils.py
+++ b/tests/python/pytorch/graphbolt/gb_test_utils.py
@@ -9,9 +9,10 @@
 import torch
 
 
-def rand_csc_graph(N, density):
+def rand_csc_graph(N, density, bidirection_edge=False):
     adj = sp.random(N, N, density)
-    adj = adj + adj.T
+    if bidirection_edge:
+        adj = adj + adj.T
     adj = adj.tocsc()
 
     indptr = torch.LongTensor(adj.indptr)
diff --git a/tests/python/pytorch/graphbolt/impl/test_negative_sampler.py b/tests/python/pytorch/graphbolt/impl/test_negative_sampler.py
index cdd61fa232bb..2fb74b2f8a93 100644
--- a/tests/python/pytorch/graphbolt/impl/test_negative_sampler.py
+++ b/tests/python/pytorch/graphbolt/impl/test_negative_sampler.py
@@ -32,7 +32,7 @@ def test_NegativeSampler_invoke():
 
 def test_UniformNegativeSampler_invoke():
     # Instantiate graph and required datapipes.
-    graph = gb_test_utils.rand_csc_graph(100, 0.05)
+    graph = gb_test_utils.rand_csc_graph(100, 0.05, bidirection_edge=True)
     num_seeds = 30
     item_set = gb.ItemSet(
         torch.arange(0, 2 * num_seeds).reshape(-1, 2), names="node_pairs"
@@ -69,7 +69,7 @@ def _verify(negative_sampler):
 @pytest.mark.parametrize("negative_ratio", [1, 5, 10, 20])
 def test_Uniform_NegativeSampler(negative_ratio):
     # Construct FusedCSCSamplingGraph.
-    graph = gb_test_utils.rand_csc_graph(100, 0.05)
+    graph = gb_test_utils.rand_csc_graph(100, 0.05, bidirection_edge=True)
     num_seeds = 30
     item_set = gb.ItemSet(
         torch.arange(0, num_seeds * 2).reshape(-1, 2), names="node_pairs"
diff --git a/tests/python/pytorch/graphbolt/test_base.py b/tests/python/pytorch/graphbolt/test_base.py
index e83ec734d2a4..409f22f832a0 100644
--- a/tests/python/pytorch/graphbolt/test_base.py
+++ b/tests/python/pytorch/graphbolt/test_base.py
@@ -29,7 +29,7 @@ def test_CopyToWithMiniBatches():
     N = 16
     B = 2
     itemset = gb.ItemSet(torch.arange(N), names="seed_nodes")
-    graph = gb_test_utils.rand_csc_graph(100, 0.15)
+    graph = gb_test_utils.rand_csc_graph(100, 0.15, bidirection_edge=True)
 
     features = {}
     keys = [("node", None, "a"), ("node", None, "b")]
diff --git a/tests/python/pytorch/graphbolt/test_feature_fetcher.py b/tests/python/pytorch/graphbolt/test_feature_fetcher.py
index 61c3e8df468e..3437bdbd580a 100644
--- a/tests/python/pytorch/graphbolt/test_feature_fetcher.py
+++ b/tests/python/pytorch/graphbolt/test_feature_fetcher.py
@@ -8,7 +8,7 @@
 
 def test_FeatureFetcher_invoke():
     # Prepare graph and required datapipes.
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     a = torch.tensor(
         [[random.randint(0, 10)] for _ in range(graph.total_num_nodes)]
     )
@@ -40,7 +40,7 @@ def test_FeatureFetcher_invoke():
 
 
 def test_FeatureFetcher_homo():
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     a = torch.tensor(
         [[random.randint(0, 10)] for _ in range(graph.total_num_nodes)]
     )
@@ -65,7 +65,7 @@ def test_FeatureFetcher_homo():
 
 
 def test_FeatureFetcher_with_edges_homo():
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     a = torch.tensor(
         [[random.randint(0, 10)] for _ in range(graph.total_num_nodes)]
     )
diff --git a/tests/python/pytorch/graphbolt/test_minibatch_transformer.py b/tests/python/pytorch/graphbolt/test_minibatch_transformer.py
index bad866449978..726d92b4daea 100644
--- a/tests/python/pytorch/graphbolt/test_minibatch_transformer.py
+++ b/tests/python/pytorch/graphbolt/test_minibatch_transformer.py
@@ -7,7 +7,7 @@ def test_dgl_minibatch_converter():
     N = 32
     B = 4
     itemset = gb.ItemSet(torch.arange(N), names="seed_nodes")
-    graph = gb_test_utils.rand_csc_graph(200, 0.15)
+    graph = gb_test_utils.rand_csc_graph(200, 0.15, bidirection_edge=True)
 
     features = {}
     keys = [("node", None, "a"), ("node", None, "b")]
diff --git a/tests/python/pytorch/graphbolt/test_multi_process_dataloader.py b/tests/python/pytorch/graphbolt/test_multi_process_dataloader.py
index 9aaa626d6c82..d095b5cbaa22 100644
--- a/tests/python/pytorch/graphbolt/test_multi_process_dataloader.py
+++ b/tests/python/pytorch/graphbolt/test_multi_process_dataloader.py
@@ -14,7 +14,7 @@ def test_DataLoader():
     N = 40
     B = 4
     itemset = dgl.graphbolt.ItemSet(torch.arange(N), names="seed_nodes")
-    graph = gb_test_utils.rand_csc_graph(200, 0.15)
+    graph = gb_test_utils.rand_csc_graph(200, 0.15, bidirection_edge=True)
     features = {}
     keys = [("node", None, "a"), ("node", None, "b")]
     features[keys[0]] = dgl.graphbolt.TorchBasedFeature(torch.randn(200, 4))
diff --git a/tests/python/pytorch/graphbolt/test_single_process_dataloader.py b/tests/python/pytorch/graphbolt/test_single_process_dataloader.py
index 86ca22541e39..d7accb89750f 100644
--- a/tests/python/pytorch/graphbolt/test_single_process_dataloader.py
+++ b/tests/python/pytorch/graphbolt/test_single_process_dataloader.py
@@ -11,7 +11,7 @@ def test_DataLoader():
     N = 32
     B = 4
     itemset = dgl.graphbolt.ItemSet(torch.arange(N), names="seed_nodes")
-    graph = gb_test_utils.rand_csc_graph(200, 0.15)
+    graph = gb_test_utils.rand_csc_graph(200, 0.15, bidirection_edge=True)
 
     features = {}
     keys = [("node", None, "a"), ("node", None, "b")]
diff --git a/tests/python/pytorch/graphbolt/test_subgraph_sampler.py b/tests/python/pytorch/graphbolt/test_subgraph_sampler.py
index 871c090beabf..7411a8c9cc5c 100644
--- a/tests/python/pytorch/graphbolt/test_subgraph_sampler.py
+++ b/tests/python/pytorch/graphbolt/test_subgraph_sampler.py
@@ -22,7 +22,7 @@ def test_SubgraphSampler_invoke():
 
 @pytest.mark.parametrize("labor", [False, True])
 def test_NeighborSampler_invoke(labor):
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     itemset = gb.ItemSet(torch.arange(10), names="seed_nodes")
     item_sampler = gb.ItemSampler(itemset, batch_size=2)
     num_layer = 2
@@ -43,7 +43,7 @@ def test_NeighborSampler_invoke(labor):
 
 @pytest.mark.parametrize("labor", [False, True])
 def test_NeighborSampler_fanouts(labor):
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     itemset = gb.ItemSet(torch.arange(10), names="seed_nodes")
     item_sampler = gb.ItemSampler(itemset, batch_size=2)
     num_layer = 2
@@ -67,7 +67,7 @@ def test_NeighborSampler_fanouts(labor):
 
 @pytest.mark.parametrize("labor", [False, True])
 def test_SubgraphSampler_Node(labor):
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     itemset = gb.ItemSet(torch.arange(10), names="seed_nodes")
     item_sampler = gb.ItemSampler(itemset, batch_size=2)
     num_layer = 2
@@ -84,7 +84,7 @@ def to_link_batch(data):
 
 @pytest.mark.parametrize("labor", [False, True])
 def test_SubgraphSampler_Link(labor):
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     itemset = gb.ItemSet(torch.arange(0, 20).reshape(-1, 2), names="node_pairs")
     item_sampler = gb.ItemSampler(itemset, batch_size=2)
     num_layer = 2
@@ -96,7 +96,7 @@ def test_SubgraphSampler_Link(labor):
 
 @pytest.mark.parametrize("labor", [False, True])
 def test_SubgraphSampler_Link_With_Negative(labor):
-    graph = gb_test_utils.rand_csc_graph(20, 0.15)
+    graph = gb_test_utils.rand_csc_graph(20, 0.15, bidirection_edge=True)
     itemset = gb.ItemSet(torch.arange(0, 20).reshape(-1, 2), names="node_pairs")
     item_sampler = gb.ItemSampler(itemset, batch_size=2)
     num_layer = 2