Update pipeline_stable_diffusion_3_ipa.py

pipeline_stable_diffusion_3_ipa.py

@@ -864,18 +864,12 @@ 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
         state_dict = torch.load(ip_adapter_path, map_location="cpu")
+
         device, dtype = self.transformer.device, self.transformer.dtype
-        if image_encoder_path=='laion/CLIP-ViT-H-14-laion2B-s32B-b79K':
-            from transformers import CLIPVisionModelWithProjection, CLIPImageProcessor
-            image_encoder = CLIPVisionModelWithProjection.from_pretrained(image_encoder_path)
-            image_processor = CLIPImageProcessor.from_pretrained(image_encoder_path)
-        else:
-            from transformers import SiglipVisionModel, SiglipImageProcessor
-            image_encoder = SiglipVisionModel.from_pretrained(image_encoder_path)
-            image_processor = SiglipImageProcessor.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