diff --git a/aucmedi/ensemble/stacking.py b/aucmedi/ensemble/stacking.py
index 43dfed87..dfa4bebd 100644
--- a/aucmedi/ensemble/stacking.py
+++ b/aucmedi/ensemble/stacking.py
@@ -190,24 +190,6 @@ def train(self, training_generator, epochs=20, iterations=None,
             data_train = (*ps_sampling[0], None)
             data_val = (*ps_sampling[1], None)
 
-        # Gather DataGenerator parameters
-        datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
-                         "batch_size": temp_dg.batch_size,
-                         "data_aug": temp_dg.data_aug,
-                         "seed": temp_dg.seed,
-                         "subfunctions": temp_dg.subfunctions,
-                         "shuffle": temp_dg.shuffle,
-                         "standardize_mode": temp_dg.standardize_mode,
-                         "resize": temp_dg.resize,
-                         "grayscale": temp_dg.grayscale,
-                         "prepare_images": temp_dg.prepare_images,
-                         "sample_weights": temp_dg.sample_weights,
-                         "image_format": temp_dg.image_format,
-                         "loader": temp_dg.sample_loader,
-                         "workers": temp_dg.workers,
-                         "kwargs": temp_dg.kwargs
-        }
-
         # Gather training parameters
         parameters_training = {"epochs": epochs,
                                "iterations": iterations,
@@ -248,6 +230,24 @@ def train(self, training_generator, epochs=20, iterations=None,
                 "multiprocessing": self.model_list[i].multiprocessing,
             }
 
+            # Gather DataGenerator parameters
+            datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
+                             "batch_size": temp_dg.batch_size,
+                             "data_aug": temp_dg.data_aug,
+                             "seed": temp_dg.seed,
+                             "subfunctions": temp_dg.subfunctions,
+                             "shuffle": temp_dg.shuffle,
+                             "standardize_mode": self.model_list[i].meta_standardize,
+                             "resize": self.model_list[i].meta_input,
+                             "grayscale": temp_dg.grayscale,
+                             "prepare_images": temp_dg.prepare_images,
+                             "sample_weights": temp_dg.sample_weights,
+                             "image_format": temp_dg.image_format,
+                             "loader": temp_dg.sample_loader,
+                             "workers": temp_dg.workers,
+                             "kwargs": temp_dg.kwargs
+            }
+
             # Start training process
             process_queue = mp.Queue()
             process_train = mp.Process(target=__training_process__,
@@ -304,24 +304,6 @@ def train_metalearner(self, training_generator):
         if len(ps_sampling[0]) == 3 : data_ensemble = ps_sampling[2]
         else : data_ensemble = (*ps_sampling[2], None)
 
-        # Gather DataGenerator parameters
-        datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
-                         "batch_size": temp_dg.batch_size,
-                         "data_aug": temp_dg.data_aug,
-                         "seed": temp_dg.seed,
-                         "subfunctions": temp_dg.subfunctions,
-                         "shuffle": temp_dg.shuffle,
-                         "standardize_mode": temp_dg.standardize_mode,
-                         "resize": temp_dg.resize,
-                         "grayscale": temp_dg.grayscale,
-                         "prepare_images": temp_dg.prepare_images,
-                         "sample_weights": temp_dg.sample_weights,
-                         "image_format": temp_dg.image_format,
-                         "loader": temp_dg.sample_loader,
-                         "workers": temp_dg.workers,
-                         "kwargs": temp_dg.kwargs
-        }
-
         # Identify path to model directory
         if isinstance(self.cache_dir, tempfile.TemporaryDirectory):
             path_model_dir = self.cache_dir.name
@@ -351,6 +333,24 @@ def train_metalearner(self, training_generator):
                 "multiprocessing": self.model_list[i].multiprocessing,
             }
 
+            # Gather DataGenerator parameters
+            datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
+                             "batch_size": temp_dg.batch_size,
+                             "data_aug": temp_dg.data_aug,
+                             "seed": temp_dg.seed,
+                             "subfunctions": temp_dg.subfunctions,
+                             "shuffle": temp_dg.shuffle,
+                             "standardize_mode": self.model_list[i].meta_standardize,
+                             "resize": self.model_list[i].meta_input,
+                             "grayscale": temp_dg.grayscale,
+                             "prepare_images": temp_dg.prepare_images,
+                             "sample_weights": temp_dg.sample_weights,
+                             "image_format": temp_dg.image_format,
+                             "loader": temp_dg.sample_loader,
+                             "workers": temp_dg.workers,
+                             "kwargs": temp_dg.kwargs
+            }
+
             # Start inference process for model i
             process_queue = mp.Queue()
             process_pred = mp.Process(target=__prediction_process__,
@@ -418,24 +418,6 @@ def predict(self, prediction_generator, return_ensemble=False):
         # Extract data
         data_test = (temp_dg.samples, temp_dg.labels, temp_dg.metadata)
 
-        # Gather DataGenerator parameters
-        datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
-                         "batch_size": temp_dg.batch_size,
-                         "data_aug": temp_dg.data_aug,
-                         "seed": temp_dg.seed,
-                         "subfunctions": temp_dg.subfunctions,
-                         "shuffle": temp_dg.shuffle,
-                         "standardize_mode": temp_dg.standardize_mode,
-                         "resize": temp_dg.resize,
-                         "grayscale": temp_dg.grayscale,
-                         "prepare_images": temp_dg.prepare_images,
-                         "sample_weights": temp_dg.sample_weights,
-                         "image_format": temp_dg.image_format,
-                         "loader": temp_dg.sample_loader,
-                         "workers": temp_dg.workers,
-                         "kwargs": temp_dg.kwargs
-        }
-
         # Identify path to model directory
         if isinstance(self.cache_dir, tempfile.TemporaryDirectory):
             path_model_dir = self.cache_dir.name
@@ -464,6 +446,24 @@ def predict(self, prediction_generator, return_ensemble=False):
                 "multiprocessing": self.model_list[i].multiprocessing,
             }
 
+            # Gather DataGenerator parameters
+            datagen_paras = {"path_imagedir": temp_dg.path_imagedir,
+                             "batch_size": temp_dg.batch_size,
+                             "data_aug": temp_dg.data_aug,
+                             "seed": temp_dg.seed,
+                             "subfunctions": temp_dg.subfunctions,
+                             "shuffle": temp_dg.shuffle,
+                             "standardize_mode": self.model_list[i].meta_standardize,
+                             "resize": self.model_list[i].meta_input,
+                             "grayscale": temp_dg.grayscale,
+                             "prepare_images": temp_dg.prepare_images,
+                             "sample_weights": temp_dg.sample_weights,
+                             "image_format": temp_dg.image_format,
+                             "loader": temp_dg.sample_loader,
+                             "workers": temp_dg.workers,
+                             "kwargs": temp_dg.kwargs
+            }
+
             # Start inference process for model i
             process_queue = mp.Queue()
             process_pred = mp.Process(target=__prediction_process__,