Update pipeline_stable_diffusion_3_ipa.py

pipeline_stable_diffusion_3_ipa.py

@@ -1164,7 +1164,6 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
                 clip_image_embeds_2 = clip_image_embeds_2.to(device, dtype=dtype)
                 clip_image_embeds_2 = self.image_encoder(clip_image_embeds_2, output_hidden_states=True).hidden_states[-2]
             clip_image_embeds_2 = clip_image_embeds_2 * scale_2
-            print('encoder output size 2: ', clip_image_embeds_2.shape)
             image_prompt_embeds_list.append(clip_image_embeds_2)
         if clip_image_3 != None:
             print('Using tertiary image.')
@@ -1174,7 +1173,6 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
                 clip_image_embeds_3 = clip_image_embeds_3.to(device, dtype=dtype)
                 clip_image_embeds_3 = self.image_encoder(clip_image_embeds_3, output_hidden_states=True).hidden_states[-2]
             clip_image_embeds_3 = clip_image_embeds_3 * scale_3
-            print('encoder output size 3: ', clip_image_embeds_3.shape)
             image_prompt_embeds_list.append(clip_image_embeds_3)
         if clip_image_4 != None:
             print('Using quaternary image.')
@@ -1182,9 +1180,8 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
             with torch.inference_mode():
                 clip_image_embeds_4 = self.clip_image_processor(images=clip_image_4, return_tensors="pt").pixel_values
                 clip_image_embeds_4 = clip_image_embeds_4.to(device, dtype=dtype)
-                clip_image_embeds_2 = self.image_encoder(clip_image_embeds_4, output_hidden_states=True).hidden_states[-2]
+                clip_image_embeds_4 = self.image_encoder(clip_image_embeds_4, output_hidden_states=True).hidden_states[-2]
             clip_image_embeds_4 = clip_image_embeds_4 * scale_4
-            print('encoder output size 4: ', clip_image_embeds_4.shape)
             image_prompt_embeds_list.append(clip_image_embeds_4)
         if clip_image_5 != None:
             print('Using quinary image.')
@@ -1194,7 +1191,6 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
                 clip_image_embeds_5 = clip_image_embeds_5.to(device, dtype=dtype)
                 clip_image_embeds_5 = self.image_encoder(clip_image_embeds_5, output_hidden_states=True).hidden_states[-2]
             clip_image_embeds_5 = clip_image_embeds_5 * scale_5
-            print('encoder output size 5: ', clip_image_embeds_5.shape)
             image_prompt_embeds_list.append(clip_image_embeds_5)
             
         clip_image_embeds_cat_list = torch.cat(image_prompt_embeds_list).mean(dim=0)