Fix bus error when using GPT2 on M1 macs (#34031)

There's a bug on M1 macs with transformer >= 4.43.0 and torch >= 2.1.0, where if a model has tied embeddings, then the fast loading from #31771 causes a bus error when the model is actually run. This can be solved by disabling `_supports_param_buffer_assignment` for these models.

More info in comments in #33357

src/transformers/models/gpt2/modeling_gpt2.py

@@ -890,6 +890,8 @@ class GPT2DoubleHeadsModelOutput(ModelOutput):
     GPT2_START_DOCSTRING,
 )
 class GPT2Model(GPT2PreTrainedModel):
+    _supports_param_buffer_assignment = False
+
     def __init__(self, config):
         super().__init__(config)