diff --git a/examples/graphbolt/pyg/multigpu/node_classification.py b/examples/graphbolt/pyg/multigpu/node_classification.py
index 30e076ce50f8..306a5cd2b618 100644
--- a/examples/graphbolt/pyg/multigpu/node_classification.py
+++ b/examples/graphbolt/pyg/multigpu/node_classification.py
@@ -391,6 +391,7 @@ def run(rank, world_size, args, dataset):
     # Set up multiprocessing environment.
     torch.cuda.set_device(rank)
     dist.init_process_group(
+        backend='cpu:gloo,cuda:nccl',
         init_method="tcp://127.0.0.1:12345",
         rank=rank,
         world_size=world_size,
diff --git a/tests/python/pytorch/graphbolt/test_dataloader.py b/tests/python/pytorch/graphbolt/test_dataloader.py
index 5843264516fc..6f56a3bb54b0 100644
--- a/tests/python/pytorch/graphbolt/test_dataloader.py
+++ b/tests/python/pytorch/graphbolt/test_dataloader.py
@@ -89,6 +89,7 @@ def test_gpu_sampling_DataLoader(
             else "tcp://127.0.0.1:12345"
         )
         thd.init_process_group(
+            backend='cpu:gloo,cuda:nccl',
             init_method=init_method,
             world_size=1,
             rank=0,