Fix test_trompt.py (#373)

`stype_encoder_dicts` needs to be initialized inside `test_trompt()`.
otherwise, it will be persisted across different parametrization,
causing an unexpected error.

Author: weihua916

test/nn/models/test_trompt.py

@@ -7,24 +7,11 @@
 
 
 @pytest.mark.parametrize('batch_size', [0, 5])
-@pytest.mark.parametrize('stype_encoder_dicts', [
-    [
-        {
-            stype.numerical: LinearEncoder(),
-            stype.categorical: EmbeddingEncoder(),
-        },
-        {
-            stype.numerical: LinearEncoder(),
-            stype.categorical: EmbeddingEncoder(),
-        },
-        {
-            stype.numerical: LinearEncoder(),
-            stype.categorical: EmbeddingEncoder(),
-        },
-    ],
-    None,
+@pytest.mark.parametrize('use_stype_encoder_dicts', [
+    True,
+    False,
 ])
-def test_trompt(batch_size, stype_encoder_dicts):
+def test_trompt(batch_size, use_stype_encoder_dicts):
     batch_size = 10
     channels = 8
     out_channels = 1
@@ -33,6 +20,23 @@ def test_trompt(batch_size, stype_encoder_dicts):
     dataset: Dataset = FakeDataset(num_rows=10, with_nan=False)
     dataset.materialize()
     tensor_frame = dataset.tensor_frame[:batch_size]
+    if use_stype_encoder_dicts:
+        stype_encoder_dicts = [
+            {
+                stype.numerical: LinearEncoder(),
+                stype.categorical: EmbeddingEncoder(),
+            },
+            {
+                stype.numerical: LinearEncoder(),
+                stype.categorical: EmbeddingEncoder(),
+            },
+            {
+                stype.numerical: LinearEncoder(),
+                stype.categorical: EmbeddingEncoder(),
+            },
+        ]
+    else:
+        stype_encoder_dicts = None
     model = Trompt(
         channels=channels,
         out_channels=out_channels,