Update pipeline_stable_diffusion_3_ipa_clip.py

pipeline_stable_diffusion_3_ipa_clip.py

@@ -865,16 +865,11 @@ class StableDiffusion3Pipeline(DiffusionPipeline, SD3LoraLoaderMixin, FromSingle
     @torch.inference_mode()
     def init_ipadapter(self, ip_adapter_path, image_encoder_path, nb_token, output_dim=2432):
         from transformers import SiglipVisionModel, SiglipImageProcessor
-        from transformers import CLIPVisionModelWithProjection, CLIPImageProcessor
         state_dict = torch.load(ip_adapter_path, map_location="cpu")
 
         device, dtype = self.transformer.device, self.transformer.dtype
-        #image_encoder = SiglipVisionModel.from_pretrained(image_encoder_path)
-        image_encoder = CLIPVisionModelWithProjection.from_pretrained(image_encoder_path)
-        
-        #image_processor = SiglipImageProcessor.from_pretrained(image_encoder_path)
-        image_processor = CLIPImageProcessor.from_pretrained(image_encoder_path)
-        
+        image_encoder = SiglipVisionModel.from_pretrained(image_encoder_path)
+        image_processor = SiglipImageProcessor.from_pretrained(image_encoder_path)
         image_encoder.eval()
         image_encoder.to(device, dtype=dtype)
         self.image_encoder = image_encoder