diff --git a/src/dwi_ml/general/data/dataset/multi_subject_containers.py b/src/dwi_ml/general/data/dataset/multi_subject_containers.py
index 64b55b92..27e2344f 100644
--- a/src/dwi_ml/general/data/dataset/multi_subject_containers.py
+++ b/src/dwi_ml/general/data/dataset/multi_subject_containers.py
@@ -448,12 +448,20 @@ def load_data(self, load_training=True, load_validation=True,
             if compress == 'Not defined by user':
                 compress = None
 
-            # Loading the first training subject's group information.
+            # Loading the first subject's group information.
             # Others should fit.
-            one_subj = hdf_handle.attrs['training_subjs'][0]
+            if len(hdf_handle.attrs['training_subjs']) > 0:
+                first_subj = hdf_handle.attrs['training_subjs'][0]
+            elif len(hdf_handle.attrs['validation_subjs']) > 0:
+                first_subj = hdf_handle.attrs['validation_subjs'][0]
+            elif len(hdf_handle.attrs['testing_subjs']) > 0:
+                first_subj = hdf_handle.attrs['testing_subjs'][0]
+            else:
+                raise ValueError("No subject found in the hdf5 file")
+
             (poss_volume_groups, nb_features, poss_strea_groups,
              contains_connectivity) = prepare_groups_info(
-                one_subj, hdf_handle, ref_group_info=None)
+                first_subj, hdf_handle, ref_group_info=None)
             logger.debug("Possible volume groups are: {}"
                          .format(poss_volume_groups))
             logger.debug("Number of features in each of these groups: {}"