Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Remove unecessary clones in the wan2.2 VAE. (#9083)

comfyanonymousc60dc4177c16
diff --git a/comfy/ldm/wan/vae2_2.py b/comfy/ldm/wan/vae2_2.py
index c2c150e10..b9c2d1a26 100644
--- a/comfy/ldm/wan/vae2_2.py
+++ b/comfy/ldm/wan/vae2_2.py
@@ -136,7 +136,7 @@ class ResidualBlock(nn.Module):
if in_dim != out_dim else nn.Identity())
def forward(self, x, feat_cache=None, feat_idx=[0]):
- h = self.shortcut(x)
+ old_x = x
for layer in self.residual:
if isinstance(layer, CausalConv3d) and feat_cache is not None:
idx = feat_idx[0]
@@ -156,7 +156,7 @@ class ResidualBlock(nn.Module):
feat_idx[0] += 1
else:
x = layer(x)
- return x + h
+ return x + self.shortcut(old_x)
def patchify(x, patch_size):
@@ -327,7 +327,7 @@ class Down_ResidualBlock(nn.Module):
self.downsamples = nn.Sequential(*downsamples)
def forward(self, x, feat_cache=None, feat_idx=[0]):
- x_copy = x.clone()
+ x_copy = x
for module in self.downsamples:
x = module(x, feat_cache, feat_idx)
@@ -369,7 +369,7 @@ class Up_ResidualBlock(nn.Module):
self.upsamples = nn.Sequential(*upsamples)
def forward(self, x, feat_cache=None, feat_idx=[0], first_chunk=False):
- x_main = x.clone()
+ x_main = x
for module in self.upsamples:
x_main = module(x_main, feat_cache, feat_idx)
if self.avg_shortcut is not None: