diff --git a/test/data/test_functional.py b/test/data/test_functional.py
index 174f641733..3e32835a9e 100644
--- a/test/data/test_functional.py
+++ b/test/data/test_functional.py
@@ -80,6 +80,14 @@ def test_sentencepiece_tokenizer(self):
         self.assertEqual(list(spm_generator([test_sample]))[0],
                          ref_results)
 
+    def test_sentencepiece_unsupported_input_type(self):
+        with self.assertRaisesRegex(
+            TypeError,
+            'Unsupported type for spm argument: dict. '
+            'Supported types are: str, io.BufferedReader'
+        ):
+            load_sp_model(dict())
+
     # TODO(Nayef211): remove decorator once https://github.com/pytorch/pytorch/issues/38207 is closed
     @unittest.skipIf(platform.system() == "Windows", "Test is known to fail on Windows.")
     def test_BasicEnglishNormalize(self):
diff --git a/torchtext/data/functional.py b/torchtext/data/functional.py
index 3333c2dc63..6e20c8e667 100644
--- a/torchtext/data/functional.py
+++ b/torchtext/data/functional.py
@@ -58,7 +58,12 @@ def load_sp_model(spm):
     elif isinstance(spm, io.BufferedReader):
         return torch.ops.torchtext.load_sp_model_string(spm.read())
     else:
-        raise RuntimeError('the input of the load_sp_model func is not supported.')
+        raise TypeError(
+            f'Unsupported type for spm argument: {type(spm).__name__}. ' +
+            'Supported types are: ' +
+            ', '.join([
+                'str', 'io.BufferedReader'
+            ]))
 
 
 def sentencepiece_numericalizer(sp_model):