docs: fix return type annotation of get_default_model_revision (#35982

)

src/transformers/pipelines/base.py

@@ -384,7 +384,7 @@ def get_framework(model, revision: Optional[str] = None):
 
 def get_default_model_and_revision(
     targeted_task: Dict, framework: Optional[str], task_options: Optional[Any]
-) -> Union[str, Tuple[str, str]]:
+) -> Tuple[str, str]:
     """
     Select a default model to use for a given task. Defaults to pytorch if ambiguous.
 
@@ -401,7 +401,9 @@ def get_default_model_and_revision(
 
     Returns
 
-        `str` The model string representing the default model for this pipeline
+        Tuple:
+            - `str` The model string representing the default model for this pipeline.
+            - `str` The revision of the model.
     """
     if is_torch_available() and not is_tf_available():
         framework = "pt"

tests/pipelines/test_pipelines_common.py

@@ -796,7 +796,7 @@ def test_register_pipeline(self):
             pipeline_class=PairClassificationPipeline,
             pt_model=AutoModelForSequenceClassification if is_torch_available() else None,
             tf_model=TFAutoModelForSequenceClassification if is_tf_available() else None,
-            default={"pt": "hf-internal-testing/tiny-random-distilbert"},
+            default={"pt": ("hf-internal-testing/tiny-random-distilbert", "2ef615d")},
             type="text",
         )
         assert "custom-text-classification" in PIPELINE_REGISTRY.get_supported_tasks()
@@ -806,7 +806,9 @@ def test_register_pipeline(self):
         self.assertEqual(task_def["tf"], (TFAutoModelForSequenceClassification,) if is_tf_available() else ())
         self.assertEqual(task_def["type"], "text")
         self.assertEqual(task_def["impl"], PairClassificationPipeline)
-        self.assertEqual(task_def["default"], {"model": {"pt": "hf-internal-testing/tiny-random-distilbert"}})
+        self.assertEqual(
+            task_def["default"], {"model": {"pt": ("hf-internal-testing/tiny-random-distilbert", "2ef615d")}}
+        )
 
         # Clean registry for next tests.
         del PIPELINE_REGISTRY.supported_tasks["custom-text-classification"]