Update modeling_molmo to work with transformers 4.56.2

Avoid errors like this:
```
[rank0]: past_length = past_key_values[0][0].size(-2)
[rank0]: AttributeError: 'NoneType' object has no attribute 'size'
```

modeling_molmo.py

@@ -2226,6 +2226,11 @@ class MolmoForCausalLM(PreTrainedModel):
     def prepare_inputs_for_generation(
         self, input_ids: torch.LongTensor, past_key_values: Optional[List[Tuple]] = None, **kwargs
     ):
+        has_past_key_values = past_key_values is not None and any (x != (None, None) for x in past_key_values)
+        if not has_past_key_values:
+            # Override a list of (None, None) tuples with a single None value as expected by the rest of the code
+            past_key_values = None
+
         if past_key_values:
             # This is because we want the model to only process the last generated token.
             input_ids = input_ids[:, -1:]