Test: generate with torch.compile(model.forward) as a fast test (#34544)

tests/models/whisper/test_modeling_whisper.py

@@ -1602,6 +1602,11 @@ class WhisperModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
             with self.assertRaises(ValueError):
                 model(input_features=input_features, labels=labels)
 
+    # TODO (joao, eustache): fix me :)
+    @unittest.skip(reason="Whisper's custom generate is not consistent regarding the cache return types")
+    def test_generate_compile_model_forward(self):
+        pass
+
 
 @require_torch
 @require_torchaudio