Fix missing initializations for models created in 2024 (#38987)

* fix GroundingDino

* fix SuperGlue

* fix GroundingDino

* fix MambaModel

* fix OmDetTurbo

* fix SegGpt

* fix Qwen2Audio

* fix Mamba2

* fix DabDetr

* fix Dac

* fix FalconMamba

* skip timm initialization

* fix Encodec and MusicgenMelody

* fix Musicgen

* skip timm initialization test

* fix OmDetTurbo

* clean the code

Co-authored-by: Cyril Vallez <cyril.vallez@gmail.com>

* add reviewed changes

* add back timm

* style

* better check for parametrizations

---------

Co-authored-by: Cyril Vallez <cyril.vallez@gmail.com>

tests/test_modeling_common.py

@@ -855,7 +855,7 @@ class ModelTesterMixin:
             # For now, skip everything older than 2025 and "important models" (too much models to patch otherwise)
             # Use `supports_cache_class` as a proxy to judge "important" models in order to prioritize them
             # TODO: relax this as we patch more and more models
-            if addition_year < 2025 and not model_class._supports_cache_class:
+            if addition_year < 2024 and not model_class._supports_cache_class:
                 self.skipTest(reason=f"{model_class} is not a priorited model for now.")
 
             # Monkey patch the method to add a seed (we do it on PreTrainedModel._initialize_weights, which wraps
@@ -895,6 +895,11 @@ class ModelTesterMixin:
                 model_from_config.state_dict().items(), model_from_pretrained.state_dict().items()
             ):
                 self.assertEqual(k1, k2, "The keys from each model should be the same")
+
+                # In case using torch.nn.utils.parametrizations on a module, we should skip the resulting keys
+                if re.search(r"\.parametrizations\..*?\.original[01]", k1):
+                    continue
+
                 # Since we added the seed, they should be exactly the same (i.e. using allclose maybe be wrong due
                 # to very low std in init function)
                 if not (v1 == v2).all():