update

Author: a-r-r-o-w

src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_modular.py

@@ -233,10 +233,11 @@ def encode_image(self, components, image, device, num_images_per_prompt, output_
 
     # modified from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_ip_adapter_image_embeds
     def prepare_ip_adapter_image_embeds(
-        self, components, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
+        self, components, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt, do_classifier_free_guidance
     ):
         image_embeds = []
-        negative_image_embeds = []
+        if do_classifier_free_guidance:
+            negative_image_embeds = []
         if ip_adapter_image_embeds is None:
             if not isinstance(ip_adapter_image, list):
                 ip_adapter_image = [ip_adapter_image]
@@ -255,18 +256,21 @@ def prepare_ip_adapter_image_embeds(
                 )
 
                 image_embeds.append(single_image_embeds[None, :])
-                negative_image_embeds.append(single_negative_image_embeds[None, :])
+                if do_classifier_free_guidance:
+                    negative_image_embeds.append(single_negative_image_embeds[None, :])
         else:
             for single_image_embeds in ip_adapter_image_embeds:
-                single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
-                negative_image_embeds.append(single_negative_image_embeds)
+                if do_classifier_free_guidance:
+                    single_negative_image_embeds, single_image_embeds = single_image_embeds.chunk(2)
+                    negative_image_embeds.append(single_negative_image_embeds)
                 image_embeds.append(single_image_embeds)
 
         ip_adapter_image_embeds = []
         for i, single_image_embeds in enumerate(image_embeds):
             single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
-            single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
-            single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
+            if do_classifier_free_guidance:
+                single_negative_image_embeds = torch.cat([negative_image_embeds[i]] * num_images_per_prompt, dim=0)
+                single_image_embeds = torch.cat([single_negative_image_embeds, single_image_embeds], dim=0)
 
             single_image_embeds = single_image_embeds.to(device=device)
             ip_adapter_image_embeds.append(single_image_embeds)