The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add emb_patch support to UNetModel forward (#4779)
Jedrzej Kosinskif04229b84d3d
diff --git a/comfy/ldm/modules/diffusionmodules/openaimodel.py b/comfy/ldm/modules/diffusionmodules/openaimodel.pyindex 6535e899c..2902073d5 100644--- a/comfy/ldm/modules/diffusionmodules/openaimodel.py+++ b/comfy/ldm/modules/diffusionmodules/openaimodel.py@@ -842,6 +842,11 @@ class UNetModel(nn.Module):t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False).to(x.dtype)emb = self.time_embed(t_emb)+ if "emb_patch" in transformer_patches:+ patch = transformer_patches["emb_patch"]+ for p in patch:+ emb = p(emb, self.model_channels, transformer_options)+if self.num_classes is not None:assert y.shape[0] == x.shape[0]emb = emb + self.label_emb(y)