feat: 🚧 Begin FactorCompressedConvNode

src/leibnetz/nodes/factor_compressed_conv_node.py

@@ -1,13 +1,20 @@
 from leibnetz.nodes import Node
 
+
 class FactorCompressedConvNode(Node):
+    # Inspired by LoRA fine tuning method
+
+    # TODO: Do with codebook adaptive lookup
+    # increase ranks steadily until optimum found
     def __init__(
         self,
         input_keys,
         output_keys,
         input_nc,
         output_nc,
         kernel_sizes,
+        rank_A: int = 8,
+        rank_B: int = 8,
         output_key_channels=None,
         activation="ReLU",
         padding="valid",
@@ -17,6 +24,21 @@ def __init__(
         dropout_prob=None,
         identifier=None,
     ) -> None:
-    self.rank_A = ...
-    self.rank_B = ...
-    # matrices A and B are multiplied to compute the full weight matrix W for the forward pass
+        super().__init__(input_keys, output_keys, identifier)
+        # matrices A and B are multiplied to compute the full weight matrix W for the forward pass
+        self.rank_A = rank_A
+        self.rank_B = rank_B
+        self.output_key_channels = output_key_channels
+        self._type = __name__.split(".")[-1]
+        self.input_nc = input_nc
+        self.output_nc = output_nc
+        self.kernel_sizes = kernel_sizes
+        self.activation = activation
+        self.padding = padding
+        self.residual = residual
+        self.padding_mode = padding_mode
+        self.norm_layer = norm_layer
+        self.dropout_prob = dropout_prob
+        self.model = ...  # TODO
+        self.color = "#00FFF0"
+        self._convolution_crop = None