config attribute not foud error for FluxImagetoImage Pipeline for multi controlnet solved (#9586)

Co-authored-by: YiYi Xu <yixu310@gmail.com>

Author: 2 people

src/diffusers/pipelines/flux/pipeline_flux_controlnet_image_to_image.py

@@ -903,9 +903,12 @@ def __call__(
 
                 timestep = t.expand(latents.shape[0]).to(latents.dtype)
 
-                guidance = (
-                    torch.tensor([guidance_scale], device=device) if self.controlnet.config.guidance_embeds else None
-                )
+                if isinstance(self.controlnet, FluxMultiControlNetModel):
+                    use_guidance = self.controlnet.nets[0].config.guidance_embeds
+                else:
+                    use_guidance = self.controlnet.config.guidance_embeds
+
+                guidance = torch.tensor([guidance_scale], device=device) if use_guidance else None
                 guidance = guidance.expand(latents.shape[0]) if guidance is not None else None
 
                 if isinstance(controlnet_keep[i], list):