Update pipeline_stable_diffusion_3_ipa.py

pipeline_stable_diffusion_3_ipa.py

@@ -1210,8 +1210,22 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
         zeroes_tensor = torch.zeros_like(clip_image_embeds_stack_list)
         print('zeros shape: ', zeroes_tensor.shape)
         clip_image_embeds = torch.cat([zeroes_tensor, clip_image_embeds_stack_list], dim=0)
-        print('embeds shape: ', clip_image_embeds.shape)
-        
+        print('embeds shape old: ', clip_image_embeds.shape)
+
+
+
+        clip_image_embeds_cat_list = torch.cat(image_prompt_embeds_list).mean(dim=0)
+        print('catted embeds list with mean: ',image_prompt_embeds.shape)
+        bs_embed, seq_len, _ = image_prompt_embeds.shape
+        clip_image_embeds_cat_list_repeat = clip_image_embeds_cat_list.repeat(1, 1, 1)
+        print('catted embeds repeat: ',clip_image_embeds_cat_list_repeat.shape)
+        clip_image_embeds_view = clip_image_embeds_cat_list_repeat.view(bs_embed * 1, seq_len, -1)
+        print('catted viewed: ',clip_image_embeds_view.shape)
+        zeros_tensor = torch.zeros_like(clip_image_embeds_view)
+        print('zeros: ',zeros_tensor.shape)
+        clip_image_embeds = torch.cat([zeros_tensor, clip_image_embeds_view], dim=0)
+        print('embeds shape new: ', clip_image_embeds.shape)
+
         # 1. Stack the image embeddings
         stacked_image_embeds = torch.cat(image_prompt_embeds_list, dim=1)
         print('shape 1: ', stacked_image_embeds.shape)