diff --git a/dacapo/experiments/architectures/cnnectome_unet.py b/dacapo/experiments/architectures/cnnectome_unet.py
index b1ad1a821..0b03b4ecc 100644
--- a/dacapo/experiments/architectures/cnnectome_unet.py
+++ b/dacapo/experiments/architectures/cnnectome_unet.py
@@ -8,6 +8,7 @@
 
 logger = logging.getLogger(__name__)
 
+
 class CNNectomeUNet(Architecture):
     """
     A U-Net architecture for 3D or 4D data. The U-Net expects 3D or 4D tensors
@@ -181,7 +182,7 @@ def __init__(self, architecture_config):
     @property
     def skip_gate(self):
         return self._skip_gate
-    
+
     @skip_gate.setter
     def skip_gate(self, skip):
         self._skip_gate = skip
@@ -1081,7 +1082,7 @@ def __init__(
         crop_factor=None,
         next_conv_kernel_sizes=None,
         activation=None,
-        skip_gate = True,
+        skip_gate=True,
     ):
         """
         Upsample module. This module performs upsampling of the input tensor
diff --git a/dacapo/experiments/architectures/cnnectome_unet_config.py b/dacapo/experiments/architectures/cnnectome_unet_config.py
index b2d84bff1..d6a17c6e7 100644
--- a/dacapo/experiments/architectures/cnnectome_unet_config.py
+++ b/dacapo/experiments/architectures/cnnectome_unet_config.py
@@ -133,5 +133,7 @@ class CNNectomeUNetConfig(ArchitectureConfig):
     )
     skip_gate: bool = attr.ib(
         default=True,
-        metadata={"help_text": "Whether to use skip gates. using skip gates concatenates the left feature map with the right feature map which helps for training. disabling the skip gate will make the model like a encoder-decoder model. example pipeline: start with skip gate false, we can train with only raw data. then we can train with skip gate true to fine tune the model with groundtruth."},
+        metadata={
+            "help_text": "Whether to use skip gates. using skip gates concatenates the left feature map with the right feature map which helps for training. disabling the skip gate will make the model like a encoder-decoder model. example pipeline: start with skip gate false, we can train with only raw data. then we can train with skip gate true to fine tune the model with groundtruth."
+        },
     )