Fix for self._model device

Signed-off-by: Jael Gu <mengjia.gu@zilliz.com>

auto_transformers.py

@@ -37,6 +37,7 @@ warnings.filterwarnings('ignore')
 os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
 t_logging.set_verbosity_error()
 
+
 def create_model(model_name, checkpoint_path, device):
     model = AutoModel.from_pretrained(model_name).to(device)
     if hasattr(model, 'pooler') and model.pooler:
@@ -50,6 +51,7 @@ def create_model(model_name, checkpoint_path, device):
     model.eval()
     return model
 
+
 # @accelerate
 class Model:
     def __init__(self, model_name, checkpoint_path, device):
@@ -131,7 +133,7 @@ class AutoTransformers(NNOperator):
 
     @property
     def _model(self):
-        return self.model.model
+        return self.model.model.to('cpu')
 
     def save_model(self, model_type: str = 'pytorch', output_file: str = 'default'):
         if output_file == 'default':
@@ -165,7 +167,7 @@ class AutoTransformers(NNOperator):
         elif model_type == 'onnx':
             from transformers.onnx.features import FeaturesManager
             from transformers.onnx import export
-            self._model = self._model.to('cpu')
+
             model_kind, model_onnx_config = FeaturesManager.check_supported_model_or_raise(
                 self._model, feature='default')
             onnx_config = model_onnx_config(self._model.config)