diff --git a/dwi_ml/models/main_models.py b/dwi_ml/models/main_models.py
index 23c1f717..942aba46 100644
--- a/dwi_ml/models/main_models.py
+++ b/dwi_ml/models/main_models.py
@@ -450,7 +450,8 @@ def forward(self, inputs, target_streamlines: List[torch.tensor]):
 
 class MainModelOneInput(MainModelAbstract):
     def prepare_batch_one_input(self, streamlines, subset: MultisubjectSubset,
-                                subj_idx, input_group_idx, prepare_mask=False):
+                                subj_idx, input_group_idx, prepare_mask=False,
+                                clear_cache=True):
         """
         These params are passed by either the batch loader or the propagator,
         which manage the data.
@@ -491,10 +492,11 @@ def prepare_batch_one_input(self, streamlines, subset: MultisubjectSubset,
         if isinstance(self, ModelWithNeighborhood):
             # Adding neighborhood.
             subj_x_data, coords_torch = interpolate_volume_in_neighborhood(
-                data_tensor, flat_subj_x_coords, self.neighborhood_vectors)
+                data_tensor, flat_subj_x_coords, self.neighborhood_vectors,
+                clear_cache=clear_cache)
         else:
             subj_x_data, coords_torch = interpolate_volume_in_neighborhood(
-                data_tensor, flat_subj_x_coords, None)
+                data_tensor, flat_subj_x_coords, None, clear_cache=clear_cache)
 
         # Split the flattened signal back to streamlines
         lengths = [len(s) for s in streamlines]
diff --git a/dwi_ml/training/batch_loaders.py b/dwi_ml/training/batch_loaders.py
index cd9c1f46..59db8d82 100644
--- a/dwi_ml/training/batch_loaders.py
+++ b/dwi_ml/training/batch_loaders.py
@@ -346,6 +346,7 @@ def __init__(self, input_group_name, **kw):
                              .format(input_group_name,
                                      self.dataset.volume_groups))
         self.input_group_idx = idx
+        self.clear_cache = True
 
     @property
     def params_for_checkpoint(self):
@@ -404,7 +405,7 @@ def load_batch_inputs(self, batch_streamlines: List[torch.tensor],
             # before adding streamline to batch.
             subbatch_x_data = self.model.prepare_batch_one_input(
                     streamlines, self.context_subset, subj,
-                    self.input_group_idx)
+                    self.input_group_idx, clear_cache=self.clear_cache)
 
             batch_x_data.extend(subbatch_x_data)