diff --git a/nlu/pipeline.py b/nlu/pipeline.py
index 27e68ddd..7a34fa90 100644
--- a/nlu/pipeline.py
+++ b/nlu/pipeline.py
@@ -1194,7 +1194,7 @@ def save(self, path, component='entire_pipeline', overwrite=False):
         print(f'Stored model in {path}')
         # else : print('Please fit untrained pipeline first or predict on a String to save it')
     def predict(self, data, output_level='', positions=False, keep_stranger_features=True, metadata=False,
-                multithread=True, drop_irrelevant_cols=True, verbose=False):
+                multithread=True, drop_irrelevant_cols=True, verbose=False,return_spark_df=False):
         '''
         Annotates a Pandas Dataframe/Pandas Series/Numpy Array/Spark DataFrame/Python List strings /Python String
 
@@ -1372,8 +1372,8 @@ def predict(self, data, output_level='', positions=False, keep_stranger_features
                 except:
                     print(
                         "If you use Modin, make sure you have installed 'pip install modin[ray]' or 'pip install modin[dask]' backend for Modin ")
-
-            return self.pythonify_spark_dataframe(sdf, self.output_different_levels,
+            if return_spark_df : return sdf
+            else :  return self.pythonify_spark_dataframe(sdf, self.output_different_levels,
                                                   keep_stranger_features=keep_stranger_features,
                                                   stranger_features=stranger_features, output_metadata=metadata,
                                                   index_provided=index_provided,