add upsample combiner feature for the unets

Author: lucidrains

README.md

@@ -1112,7 +1112,7 @@ For detailed information on training the diffusion prior, please refer to the [d
 - [x] allow for unet to be able to condition non-cross attention style as well
 - [x] speed up inference, read up on papers (ddim)
 - [x] add inpainting ability using resampler from repaint paper https://arxiv.org/abs/2201.09865
-- [ ] try out the nested unet from https://arxiv.org/abs/2005.09007 after hearing several positive testimonies from researchers, for segmentation anyhow
+- [x] add the final combination of upsample feature maps, used in unet squared, seems to have an effect in local experiments
 - [ ] interface out the vqgan-vae so a pretrained one can be pulled off the shelf to validate latent diffusion + DALL-E2
 
 ## Citations

dalle2_pytorch/dalle2_pytorch.py

@@ -1538,6 +1538,38 @@ def forward(self, x):
         fmaps = tuple(map(lambda conv: conv(x), self.convs))
         return torch.cat(fmaps, dim = 1)
 
+class UpsampleCombiner(nn.Module):
+    def __init__(
+        self,
+        dim,
+        *,
+        enabled = False,
+        dim_ins = tuple(),
+        dim_outs = tuple()
+    ):
+        super().__init__()
+        assert len(dim_ins) == len(dim_outs)
+        self.enabled = enabled
+
+        if not self.enabled:
+            self.dim_out = dim
+            return
+
+        self.fmap_convs = nn.ModuleList([Block(dim_in, dim_out) for dim_in, dim_out in zip(dim_ins, dim_outs)])
+        self.dim_out = dim + (sum(dim_outs) if len(dim_outs) > 0 else 0)
+
+    def forward(self, x, fmaps = None):
+        target_size = x.shape[-1]
+
+        fmaps = default(fmaps, tuple())
+
+        if not self.enabled or len(fmaps) == 0 or len(self.fmap_convs) == 0:
+            return x
+
+        fmaps = [resize_image_to(fmap, target_size) for fmap in fmaps]
+        outs = [conv(fmap) for fmap, conv in zip(fmaps, self.fmap_convs)]
+        return torch.cat((x, *outs), dim = 1)
+
 class Unet(nn.Module):
     def __init__(
         self,
@@ -1575,6 +1607,7 @@ def __init__(
         scale_skip_connection = False,
         pixel_shuffle_upsample = True,
         final_conv_kernel_size = 1,
+        combine_upsample_fmaps = False, # whether to combine the outputs of all upsample blocks, as in unet squared paper
         **kwargs
     ):
         super().__init__()
@@ -1710,7 +1743,8 @@ def __init__(
         self.ups = nn.ModuleList([])
         num_resolutions = len(in_out)
 
-        skip_connect_dims = [] # keeping track of skip connection dimensions
+        skip_connect_dims = []          # keeping track of skip connection dimensions
+        upsample_combiner_dims = []     # keeping track of dimensions for final upsample feature map combiner
 
         for ind, ((dim_in, dim_out), groups, layer_num_resnet_blocks, layer_self_attn) in enumerate(zip(in_out, resnet_groups, num_resnet_blocks, self_attn)):
             is_first = ind == 0
@@ -1752,14 +1786,27 @@ def __init__(
             elif sparse_attn:
                 attention = Residual(LinearAttention(dim_out, **attn_kwargs))
 
+            upsample_combiner_dims.append(dim_out)
+
             self.ups.append(nn.ModuleList([
                 ResnetBlock(dim_out + skip_connect_dim, dim_out, cond_dim = layer_cond_dim, time_cond_dim = time_cond_dim, groups = groups),
                 nn.ModuleList([ResnetBlock(dim_out + skip_connect_dim, dim_out, cond_dim = layer_cond_dim, time_cond_dim = time_cond_dim, groups = groups)  for _ in range(layer_num_resnet_blocks)]),
                 attention,
                 upsample_klass(dim_out, dim_in) if not is_last or memory_efficient else nn.Identity()
             ]))
 
-        self.final_resnet_block = ResnetBlock(dim * 2, dim, time_cond_dim = time_cond_dim, groups = top_level_resnet_group)
+        # whether to combine outputs from all upsample blocks for final resnet block
+
+        self.upsample_combiner = UpsampleCombiner(
+            dim = dim,
+            enabled = combine_upsample_fmaps,
+            dim_ins = upsample_combiner_dims,
+            dim_outs = (dim,) * len(upsample_combiner_dims)
+        )
+
+        # a final resnet block
+
+        self.final_resnet_block = ResnetBlock(self.upsample_combiner.dim_out + dim, dim, time_cond_dim = time_cond_dim, groups = top_level_resnet_group)
 
         out_dim_in = dim + (channels if lowres_cond else 0)
 
@@ -1953,7 +2000,8 @@ def forward(
 
         # go through the layers of the unet, down and up
 
-        hiddens = []
+        down_hiddens = []
+        up_hiddens = []
 
         for pre_downsample, init_block, resnet_blocks, attn, post_downsample in self.downs:
             if exists(pre_downsample):
@@ -1963,10 +2011,10 @@ def forward(
 
             for resnet_block in resnet_blocks:
                 x = resnet_block(x, t, c)
-                hiddens.append(x)
+                down_hiddens.append(x.contiguous())
 
             x = attn(x)
-            hiddens.append(x.contiguous())
+            down_hiddens.append(x.contiguous())
 
             if exists(post_downsample):
                 x = post_downsample(x)
@@ -1978,7 +2026,7 @@ def forward(
 
         x = self.mid_block2(x, t, mid_c)
 
-        connect_skip = lambda fmap: torch.cat((fmap, hiddens.pop() * self.skip_connect_scale), dim = 1)
+        connect_skip = lambda fmap: torch.cat((fmap, down_hiddens.pop() * self.skip_connect_scale), dim = 1)
 
         for init_block, resnet_blocks, attn, upsample in self.ups:
             x = connect_skip(x)
@@ -1989,8 +2037,12 @@ def forward(
                 x = resnet_block(x, t, c)
 
             x = attn(x)
+
+            up_hiddens.append(x.contiguous())
             x = upsample(x)
 
+        x = self.upsample_combiner(x, up_hiddens)
+
         x = torch.cat((x, r), dim = 1)
 
         x = self.final_resnet_block(x, t)

dalle2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '1.0.6'
+__version__ = '1.1.0'