Rename to layer

language_interpolation/dual_convolutional_network.py

@@ -8,7 +8,7 @@
 from torch import Tensor
 
 
-class DualConvolutionalNetwork(torch.nn.Module):
+class DualConvolutionalLayer(torch.nn.Module):
     def __init__(
         self,
         n: str,
@@ -53,14 +53,18 @@ def forward(self, x: Tensor):
         """
         x has shape [B, L, D]
         """
-        
+
         xshape = x.shape
         nx = x.reshape(xshape[0]*xshape[1],xshape[2])
 
         val = self.input_layer(nx)
         val = val.reshape(x.shape[0],x.shape[1],-1)
 
+        # Gradients apparently automatically accumulate, though probably want
+        # some normalization here
+        depth = 0
         while val.shape[1] > 1:
+            depth+=1
             if val.shape[1] % 2 == 1:
                 # Add padding to the end, hope this doesn't bust anything
                 val = torch.cat([val, torch.zeros(val.shape[0], 1, val.shape[2])], dim=1)

tests/test_dual_convolution.py

@@ -1,11 +1,11 @@
 import pytest
 
-from language_interpolation.dual_convolutional_network import DualConvolutionalNetwork
+from language_interpolation.dual_convolutional_network import DualConvolutionalLayer
 import torch
 
 
 def test_dual_convolution():
-    net = DualConvolutionalNetwork(n=3, in_width=1, out_width=10, hidden_layers=2, hidden_width=10, in_segments=128, segments=5)
+    net = DualConvolutionalLayer(n=3, in_width=1, out_width=10, hidden_layers=2, hidden_width=10, in_segments=128, segments=5)
     x = torch.rand(10, 15, 1) # character level
     res = net(x)
-    print('res', res)
+    print('res', res.shape)