🚨 Don't use cache in non-generative models (#38751)

* deprecate for 1 version

* style

* fix some tests

* fix esm

* skip for now, GC requires positional args but we have keyword args

* remove transpose for scores in modified models only

* skip fx trace tests

tests/models/layoutlm/test_modeling_layoutlm.py

@@ -243,7 +243,7 @@ class LayoutLMModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase
         if is_torch_available()
         else {}
     )
-    fx_compatible = True
+    fx_compatible = False  # Cannot support if `can_return_tuple`
 
     def setUp(self):
         self.model_tester = LayoutLMModelTester(self)