Merge branch 'lucidrains-master'

README.md

@@ -102,6 +102,9 @@ Tested with Python 3.6 and PyTorch 1.1.0. Should work with any version of Python
 
 ## Changelog
 
+#### Version 0.2.1
+- Added option to select the dimension along which the tensor is split ([Pull request](https://github.com/RobinBruegger/RevTorch/pull/2))
+
 #### Version 0.2.0
 - Fixed memory leak when not consuming output of the reversible block ([Issue](https://github.com/RobinBruegger/RevTorch/issues/1))
 

revtorch/revtorch.py

@@ -15,24 +15,25 @@ class ReversibleBlock(nn.Module):
         g_block (nn.Module): arbitrary subnetwork whos output shape is equal to its input shape
     '''
 
-    def __init__(self, f_block, g_block):
+    def __init__(self, f_block, g_block, split_along_dim=1):
         super(ReversibleBlock, self).__init__()
         self.f_block = f_block
         self.g_block = g_block
+        self.split_along_dim = split_along_dim
 
     def forward(self, x):
         """
         Performs the forward pass of the reversible block. Does not record any gradients.
         :param x: Input tensor. Must be splittable along dimension 1.
         :return: Output tensor of the same shape as the input tensor
         """
-        x1, x2 = torch.chunk(x, 2, dim=1)
+        x1, x2 = torch.chunk(x, 2, dim=self.split_along_dim)
         y1, y2 = None, None
         with torch.no_grad():
             y1 = x1 + self.f_block(x2)
             y2 = x2 + self.g_block(y1)
 
-        return torch.cat([y1, y2], dim=1)
+        return torch.cat([y1, y2], dim=self.split_along_dim)
 
     def backward_pass(self, y, dy):
         """
@@ -47,11 +48,11 @@ def backward_pass(self, y, dy):
         """
 
         # Split the arguments channel-wise
-        y1, y2 = torch.chunk(y, 2, dim=1)
+        y1, y2 = torch.chunk(y, 2, dim=self.split_along_dim)
         del y
         assert (not y1.requires_grad), "y1 must already be detached"
         assert (not y2.requires_grad), "y2 must already be detached"
-        dy1, dy2 = torch.chunk(dy, 2, dim=1)
+        dy1, dy2 = torch.chunk(dy, 2, dim=self.split_along_dim)
         del dy
         assert (not dy1.requires_grad), "dy1 must not require grad"
         assert (not dy2.requires_grad), "dy2 must not require grad"
@@ -100,8 +101,8 @@ def backward_pass(self, y, dy):
             x2.grad = None
 
             # Undo the channelwise split
-            x = torch.cat([x1, x2.detach()], dim=1)
-            dx = torch.cat([dx1, dx2], dim=1)
+            x = torch.cat([x1, x2.detach()], dim=self.split_along_dim)
+            dx = torch.cat([dx1, dx2], dim=self.split_along_dim)
 
         return x, dx