[core/ GC / tests] Stronger GC tests (#27124)

* stronger GC tests

* better tests and skip failing tests

* break down into 3 sub-tests

* break down into 3 sub-tests

* refactor a bit

* more refactor

* fix

* last nit

* credits contrib and suggestions

* credits contrib and suggestions

---------

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

tests/models/vilt/test_modeling_vilt.py

@@ -320,6 +320,18 @@ class ViltModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
             loss = model(**inputs).loss
             loss.backward()
 
+    @unittest.skip(
+        reason="This architecure seem to not compute gradients properly when using GC, check: https://github.com/huggingface/transformers/pull/27124"
+    )
+    def test_training_gradient_checkpointing_use_reentrant(self):
+        pass
+
+    @unittest.skip(
+        reason="This architecure seem to not compute gradients properly when using GC, check: https://github.com/huggingface/transformers/pull/27124"
+    )
+    def test_training_gradient_checkpointing_use_reentrant_false(self):
+        pass
+
     @unittest.skip(
         reason="""VilT samples image tokens from a multinomial distribution, resulting in not deterministic
                             hidden states"""