diff --git a/torchchat/cli/builder.py b/torchchat/cli/builder.py
index 5dbf48529..cd0d05056 100644
--- a/torchchat/cli/builder.py
+++ b/torchchat/cli/builder.py
@@ -388,11 +388,9 @@ def _load_model_default(builder_args: BuilderArgs) -> Model:
             builder_args.device
         ):
             model = Model.from_params(builder_args.params_path)
-        state_dict = flamingo_meta_to_tune(checkpoint)
-        model.model.load_state_dict(state_dict)
-    else:
-        checkpoint = {"model." + k: v for k, v in checkpoint.items()}
-        model.load_state_dict(checkpoint, assign=True, strict=True)
+        checkpoint = flamingo_meta_to_tune(checkpoint)
+
+    model.load_state_dict(checkpoint, assign=True, strict=True)
 
     return model
 
diff --git a/torchchat/model.py b/torchchat/model.py
index 79bd1f188..089fef9aa 100644
--- a/torchchat/model.py
+++ b/torchchat/model.py
@@ -442,6 +442,8 @@ def __init__(self, config: ModelArgs) -> None:
         # It should be assigned in the actual model implementation, if any.
         self.text_transformer_args = None
 
+        self._register_load_state_dict_pre_hook(self._load_model_state_dict)
+
     def build_model(self) -> nn.Module:
         """
         Builds a model based on the provided configuration.
@@ -468,6 +470,35 @@ def _replace_known_params(self, params):
                 params[key] = patterns[value]
         return params
     
+    def _load_model_state_dict(
+        self,
+        state_dict,
+        prefix,
+        local_metadata,
+        strict,
+        missing_keys,
+        unexpected_keys,
+        error_msgs,
+    ):
+        """
+        Updates the loaded internal model state dictionary to match the Model class structure.
+        Note that this is a temporary solution and will be removed once the model structure is finalized.
+        Args:
+            state_dict (dict): The state dictionary to load.
+            prefix (str): The prefix of the model.
+            local_metadata (dict): Local metadata.
+            strict (bool): Whether to strictly enforce that the keys in the state dictionary match the keys in the model.
+            missing_keys (list): List of missing keys.
+            unexpected_keys (list): List of unexpected keys.
+            error_msgs (list): List of error messages.
+        Returns:
+            dict: The updated state dictionary.            
+        """
+        for key in list(state_dict.keys()):
+            new_key = "model." + key
+            state_dict[new_key] = state_dict.pop(key)
+        return state_dict
+
     @abstractmethod
     def forward(self, *args, **kwargs):
         raise NotImplementedError("forward method is not implemented")
diff --git a/torchchat/utils/gguf_loader.py b/torchchat/utils/gguf_loader.py
index 309ff807c..772c6f7dd 100644
--- a/torchchat/utils/gguf_loader.py
+++ b/torchchat/utils/gguf_loader.py
@@ -47,7 +47,6 @@ def _convert_gguf_tensor_name_to_llama_nn(gguf_name: str) -> str:
     result = copy.deepcopy(gguf_name)
     for gguf_string, replacement in _name_replacements:
         result = result.replace(gguf_string, replacement)
-    result = "model." + result
     return result
 
 
@@ -55,6 +54,10 @@ def _fqn_lookup(fqn: str, module: torch.nn.Module) -> Any:
     if fqn == "":
         return module
     atoms = fqn.split(".")
+    # Expose the root module to users.
+    # Note this is temporary, and will be removed once we removed Model.model
+    if isinstance(module, Model):
+        module = module.model
     curr = module
     for a in atoms:
         curr = getattr(curr, a)