diff --git a/examples/sparse/appnp.py b/examples/sparse/appnp.py
index db54cddabca2..7c40e8461f75 100644
--- a/examples/sparse/appnp.py
+++ b/examples/sparse/appnp.py
@@ -56,7 +56,7 @@ def evaluate(g, pred):
     return val_acc, test_acc
 
 
-def train(g, A_hat, X, model):
+def train(model, g, A_hat, X):
     label = g.ndata["label"]
     train_mask = g.ndata["train_mask"]
     optimizer = Adam(model.parameters(), lr=1e-2, weight_decay=5e-4)
@@ -114,4 +114,4 @@ def train(g, A_hat, X, model):
     model = APPNP(in_size, out_size).to(dev)
 
     # Kick off training.
-    train(g, A_hat, X, model)
+    train(model, g, A_hat, X)
diff --git a/examples/sparse/sgc.py b/examples/sparse/sgc.py
index 41a397cf581c..c83eb26e3520 100644
--- a/examples/sparse/sgc.py
+++ b/examples/sparse/sgc.py
@@ -32,7 +32,7 @@ def evaluate(g, pred):
     return val_acc, test_acc
 
 
-def train(g, X_sgc, model):
+def train(model, g, X_sgc):
     label = g.ndata["label"]
     train_mask = g.ndata["train_mask"]
     optimizer = Adam(model.parameters(), lr=2e-1, weight_decay=5e-6)
@@ -91,4 +91,4 @@ def train(g, X_sgc, model):
     model = nn.Linear(in_size, out_size).to(dev)
 
     # Kick off training.
-    train(g, X_sgc, model)
+    train(model, g, X_sgc)
diff --git a/examples/sparse/sign.py b/examples/sparse/sign.py
index ae7d1b3a9c7a..a272fe6157ba 100644
--- a/examples/sparse/sign.py
+++ b/examples/sparse/sign.py
@@ -57,7 +57,7 @@ def evaluate(g, pred):
     return val_acc, test_acc
 
 
-def train(g, X_sign, model):
+def train(model, g, X_sign):
     label = g.ndata["label"]
     train_mask = g.ndata["train_mask"]
     optimizer = Adam(model.parameters(), lr=3e-3)
@@ -124,4 +124,4 @@ def train(g, X_sign, model):
     model = SIGN(in_size, out_size, r).to(dev)
 
     # Kick off training.
-    train(g, X_sign, model)
+    train(model, g, X_sign)