diff --git a/test/onnx/test_pytorch_onnx_onnxruntime.py b/test/onnx/test_pytorch_onnx_onnxruntime.py
index 7aef2feed22b9..72968f3334262 100644
--- a/test/onnx/test_pytorch_onnx_onnxruntime.py
+++ b/test/onnx/test_pytorch_onnx_onnxruntime.py
@@ -13148,7 +13148,7 @@ def forward(self, x):
         model = torch.ao.quantization.convert(model)
 
         # Set fixed input to avoid flaky test.
-        input = _construct_tensor_for_quantization_test((3, 2), offset=-384, max_val=12)
+        input = _construct_tensor_for_quantization_test((3, 4), offset=-384, max_val=12)
         self.run_test(model, input)
 
     @skipIfUnsupportedMinOpsetVersion(10)