diff --git a/examples/cars/train.py b/examples/cars/train.py
index ddc18d39..2da4137c 100644
--- a/examples/cars/train.py
+++ b/examples/cars/train.py
@@ -22,7 +22,6 @@ def train(
     shuffle: bool,
     save_dir: str,
 ):
-
     model = Model(
         lr=lr,
         mining=mining,
diff --git a/quaterion/dataset/similarity_data_loader.py b/quaterion/dataset/similarity_data_loader.py
index 0e775f03..9bc1af0b 100644
--- a/quaterion/dataset/similarity_data_loader.py
+++ b/quaterion/dataset/similarity_data_loader.py
@@ -31,7 +31,6 @@ class SimilarityDataLoader(DataLoader, Generic[T_co]):
     """
 
     def __init__(self, dataset: Dataset, **kwargs):
-
         if "collate_fn" not in kwargs:
             kwargs["collate_fn"] = self.__class__.pre_collate_fn
         self._original_dataset = dataset
diff --git a/quaterion/loss/cos_face_loss.py b/quaterion/loss/cos_face_loss.py
index 7980aeb5..4de05435 100644
--- a/quaterion/loss/cos_face_loss.py
+++ b/quaterion/loss/cos_face_loss.py
@@ -27,7 +27,6 @@ def __init__(
         margin: Optional[float] = 0.35,
         scale: Optional[float] = 64.0,
     ):
-
         super(GroupLoss, self).__init__()
 
         self.kernel = nn.Parameter(torch.FloatTensor(embedding_size, num_groups))
@@ -36,7 +35,6 @@ def __init__(
         self.margin = margin
 
     def forward(self, embeddings: Tensor, groups: LongTensor) -> Tensor:
-
         """Compute loss value
         Args:
             embeddings: shape: (batch_size, vector_length) - Output embeddings from the
diff --git a/quaterion/loss/online_contrastive_loss.py b/quaterion/loss/online_contrastive_loss.py
index 8ae841a7..d57c8c75 100644
--- a/quaterion/loss/online_contrastive_loss.py
+++ b/quaterion/loss/online_contrastive_loss.py
@@ -93,7 +93,6 @@ def forward(
         )
 
         if self._mining == "all":
-
             num_positive_pairs = anchor_positive_mask.sum()
             positive_loss = anchor_positive_dists.sum() / torch.max(
                 num_positive_pairs, torch.tensor(1e-16)
@@ -106,7 +105,6 @@ def forward(
             ).sum() / torch.max(num_negative_pairs, torch.tensor(1e-16))
 
         else:  # batch-hard pair mining
-
             # get the hardest positive for each anchor
             # shape: (batch_size,)
             hardest_positive_dists = anchor_positive_dists.max(dim=1)[0]
diff --git a/quaterion/train/cache/cache_model.py b/quaterion/train/cache/cache_model.py
index 10fbd6fe..7aafe4f6 100644
--- a/quaterion/train/cache/cache_model.py
+++ b/quaterion/train/cache/cache_model.py
@@ -24,7 +24,6 @@ def __init__(
         self,
         encoders: Dict[str, CacheEncoder],
     ):
-
         super().__init__()
         self.encoders = encoders
         for key, encoder in self.encoders.items():
diff --git a/quaterion/train/trainable_model.py b/quaterion/train/trainable_model.py
index d0110c75..e8eadbe1 100644
--- a/quaterion/train/trainable_model.py
+++ b/quaterion/train/trainable_model.py
@@ -381,7 +381,6 @@ def _maybe_compute_xbm_loss(
         ):
             loss_obj = self.loss  # Assign to tmp variable for better type inference
             if isinstance(loss_obj, GroupLoss):
-
                 memory_embeddings, memory_groups = self._xbm_buffer.get()
                 memory_loss = loss_obj.xbm_loss(
                     embeddings, targets["groups"], memory_embeddings, memory_groups
diff --git a/tests/test_import.py b/tests/test_import.py
index 233a6bfa..24541994 100644
--- a/tests/test_import.py
+++ b/tests/test_import.py
@@ -1,3 +1,2 @@
 def test_import_main_classes():
-
     from quaterion import Quaterion, TrainableModel