diff --git a/nowcasting_gan/generators.py b/nowcasting_gan/generators.py
index 0732a96..ee06fdd 100644
--- a/nowcasting_gan/generators.py
+++ b/nowcasting_gan/generators.py
@@ -41,8 +41,9 @@ def __init__(
             in_channels=context_channels, out_channels=latent_channels, kernel_size=(1, 1)
         )
         self.g1 = GBlock(input_channels=latent_channels, output_channels=latent_channels)
-        self.up_g1 = UpsampleGBlock(input_channels = latent_channels, output_channels =
-        latent_channels // 2)
+        self.up_g1 = UpsampleGBlock(
+            input_channels=latent_channels, output_channels=latent_channels // 2
+        )
 
         self.convGRU2 = ConvGRU(
             input_channels=latent_channels // 2 + context_channels // 2,
@@ -53,8 +54,9 @@ def __init__(
             in_channels=context_channels // 2, out_channels=latent_channels // 2, kernel_size=(1, 1)
         )
         self.g2 = GBlock(input_channels=latent_channels // 2, output_channels=latent_channels // 2)
-        self.up_g2 = UpsampleGBlock(input_channels = latent_channels // 2, output_channels =
-        latent_channels // 4)
+        self.up_g2 = UpsampleGBlock(
+            input_channels=latent_channels // 2, output_channels=latent_channels // 4
+        )
 
         self.convGRU3 = ConvGRU(
             input_channels=latent_channels // 4 + context_channels // 4,
@@ -65,8 +67,9 @@ def __init__(
             in_channels=context_channels // 4, out_channels=latent_channels // 4, kernel_size=(1, 1)
         )
         self.g3 = GBlock(input_channels=latent_channels // 4, output_channels=latent_channels // 4)
-        self.up_g3 = UpsampleGBlock(input_channels = latent_channels // 4, output_channels =
-        latent_channels // 8)
+        self.up_g3 = UpsampleGBlock(
+            input_channels=latent_channels // 4, output_channels=latent_channels // 8
+        )
 
         self.convGRU4 = ConvGRU(
             input_channels=latent_channels // 8 + context_channels // 8,
@@ -77,15 +80,17 @@ def __init__(
             in_channels=context_channels // 8, out_channels=latent_channels // 8, kernel_size=(1, 1)
         )
         self.g4 = GBlock(input_channels=latent_channels // 8, output_channels=latent_channels // 8)
-        self.up_g4 = UpsampleGBlock(input_channels = latent_channels // 8, output_channels =
-        latent_channels // 16)
+        self.up_g4 = UpsampleGBlock(
+            input_channels=latent_channels // 8, output_channels=latent_channels // 16
+        )
 
         self.bn = torch.nn.BatchNorm2d(latent_channels // 16)
         self.relu = torch.nn.ReLU()
         self.conv_1x1 = spectral_norm(
             torch.nn.Conv2d(
-                in_channels=latent_channels // 16, out_channels=4 * output_channels,
-                kernel_size=(1,1)
+                in_channels=latent_channels // 16,
+                out_channels=4 * output_channels,
+                kernel_size=(1, 1),
             )
         )