diff --git a/python/mxnet/optimizer/optimizer.py b/python/mxnet/optimizer/optimizer.py
index c3a1f3374a94..bc87777e40fc 100644
--- a/python/mxnet/optimizer/optimizer.py
+++ b/python/mxnet/optimizer/optimizer.py
@@ -1942,7 +1942,7 @@ def __init__(self, optimizer):
 
     def __call__(self, index, grad, weight):
         """Updates weight given gradient and index."""
-        allow_np = self.optimizer.allow_np_array
+        allow_np = self.optimizer.allow_np_array if hasattr(self.optimizer, "allow_np_array") else is_np_array()
         if not isinstance(index, (list, tuple)):
             indices = [index]
             grads = [_as_classic(grad, allow_np)]
diff --git a/tests/python/unittest/test_numpy_gluon.py b/tests/python/unittest/test_numpy_gluon.py
index f24eb6a325bf..af5425336699 100644
--- a/tests/python/unittest/test_numpy_gluon.py
+++ b/tests/python/unittest/test_numpy_gluon.py
@@ -113,6 +113,15 @@ def hybrid_forward(self, F, pred, label):
         trainer.step(1)
 
 
+@with_seed()
+@use_np
+def test_optimizer_backward_compat():
+    optimizer = mx.optimizer.SGD()
+    delattr(optimizer, "allow_np_array")
+    updater = mx.optimizer.Updater(optimizer)
+    updater(0, np.ones((0, 0)), np.zeros((0, 0)))
+
+
 @with_seed()
 @use_np
 def test_np_loss_ndarray():