Pipeline should be agnostic (huggingface#12656)

tests/test_pipelines_question_answering.py

@@ -14,6 +14,7 @@
 
 import unittest
 
+from transformers import is_tf_available, is_torch_available
 from transformers.data.processors.squad import SquadExample
 from transformers.pipelines import Pipeline, QuestionAnsweringArgumentHandler, pipeline
 from transformers.testing_utils import slow
@@ -57,14 +58,15 @@ def get_pipelines(self):
                 task=self.pipeline_task,
                 model=model,
                 tokenizer=model,
-                framework="pt",
+                framework="pt" if is_torch_available() else "tf",
                 **self.pipeline_loading_kwargs,
             )
             for model in self.small_models
         ]
         return question_answering_pipelines
 
     @slow
+    @unittest.skipIf(not is_torch_available() and not is_tf_available(), "Either torch or TF must be installed.")
     def test_high_topk_small_context(self):
         self.pipeline_running_kwargs.update({"topk": 20})
         valid_inputs = [