fix: 🐛 Fix norm_layer and residual block bugs

src/leibnetz/leibnet.py

@@ -340,7 +340,7 @@ def mps(self):
         else:
             logger.error('Unable to move model to Apple Silicon ("mps")')
 
-    def forward(self, inputs: dict[str, torch.Tensor]):
+    def forward(self, inputs: dict[str, dict[str, Sequence[int | float]]]):
         # function for forwarding data through the network
         # inputs is a dictionary of tensors
         # outputs is a dictionary of tensors

src/leibnetz/nets/unet.py

@@ -12,6 +12,8 @@ def build_unet(
     output_nc=1,
     base_nc=12,
     nc_increase_factor=2,
+    norm_layer=None,
+    residual=False,
 ):
     # define downsample nodes
     downsample_factors = np.array(downsample_factors)
@@ -28,6 +30,8 @@ def build_unet(
                 base_nc * nc_increase_factor**i,
                 kernel_sizes,
                 identifier=output_key,
+                norm_layer=norm_layer,
+                residual=residual,
             ),
         )
         c += 1
@@ -53,6 +57,8 @@ def build_unet(
             base_nc * nc_increase_factor ** (i + 1),
             kernel_sizes,
             identifier=output_key,
+            norm_layer=norm_layer,
+            residual=residual,
         )
     )
     input_key = output_key
@@ -80,6 +86,8 @@ def build_unet(
                 base_nc * nc_increase_factor**i,
                 kernel_sizes,
                 identifier=output_key,
+                norm_layer=norm_layer,
+                residual=residual,
             )
         )
         input_key = output_key
@@ -94,6 +102,8 @@ def build_unet(
             # kernel_sizes,
             [(1,) * len(top_resolution)],
             identifier="output",
+            norm_layer=norm_layer,  # TODO: remove?
+            residual=residual,
         )
     )
 

src/leibnetz/nodes/node_ops.py

@@ -73,7 +73,7 @@ def __init__(
 
         for i, kernel_size in enumerate(kernel_sizes):
             if norm_layer is not None:
-                layers.append(norm_layer(output_nc))
+                layers.append(norm_layer(input_nc))
 
             layers.append(self.activation)
 
@@ -115,7 +115,7 @@ def __init__(
     def crop(self, x, shape):
         """Center-crop x to match spatial dimensions given by shape."""
 
-        x_target_size = x.size()[: -self.dims] + np.array(shape)
+        x_target_size = x.shape[: -self.dims] + tuple(shape)
 
         offset = tuple((a - b) // 2 for a, b in zip(x.size(), x_target_size))
 
@@ -132,3 +132,4 @@ def forward(self, x):
                 init_x = self.crop(self.x_init_map(x), res.size()[-self.dims :])
             else:
                 init_x = self.x_init_map(x)
+            return res + init_x