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 optimization that caused border.
comfyanonymousbed116a1f916
diff --git a/comfy/ldm/modules/diffusionmodules/model.py b/comfy/ldm/modules/diffusionmodules/model.pyindex 4c1f4ecef..431548483 100644--- a/comfy/ldm/modules/diffusionmodules/model.py+++ b/comfy/ldm/modules/diffusionmodules/model.py@@ -85,11 +85,10 @@ class Downsample(nn.Module):stride=2,padding=0)- def forward(self, x, already_padded=False):+ def forward(self, x):if self.with_conv:- if not already_padded:- pad = (0,1,0,1)- x = torch.nn.functional.pad(x, pad, mode="constant", value=0)+ pad = (0,1,0,1)+ x = torch.nn.functional.pad(x, pad, mode="constant", value=0)x = self.conv(x)else:x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)@@ -606,9 +605,6 @@ class Encoder(nn.Module):def forward(self, x):# timestep embeddingtemb = None- pad = (0,1,0,1)- x = torch.nn.functional.pad(x, pad, mode="constant", value=0)- already_padded = True# downsamplingh = self.conv_in(x)for i_level in range(self.num_resolutions):@@ -617,8 +613,7 @@ class Encoder(nn.Module):if len(self.down[i_level].attn) > 0:h = self.down[i_level].attn[i_block](h)if i_level != self.num_resolutions-1:- h = self.down[i_level].downsample(h, already_padded)- already_padded = False+ h = self.down[i_level].downsample(h)# middleh = self.mid.block_1(h, temb)