fixed minor bug

Author: Your Name

modelopt/torch/opt/conversion.py

@@ -579,10 +579,17 @@ def restore(model: ModelLike, f: str | os.PathLike | BinaryIO, **kwargs) -> nn.M
 
     Returns:
         The model with original weights and stored architecture.
+
+    .. note::
+        Note that wrappers such as DistributedDataParallel are `not` supported during the restore
+        process. Please wrap the model after the restore process.
     """
     # initialize ModelLikeModule if needed.
     model = model if isinstance(model, nn.Module) else ModelLikeModule(model)
 
+    # check if the model is in a wrapper; we dont support restoring with wrappers
+    model = unwrap_model(model, raise_error=True)
+
     # load checkpoint
     kwargs.setdefault("map_location", "cpu")
     kwargs.setdefault("weights_only", False)