Update modeling_clip.py

modeling_clip.py

@@ -485,9 +485,9 @@ class JinaCLIPModel(JinaCLIPPreTrainedModel):
                 else:
                     raise ValueError("Unsupported image format")
     
-                processed_inputs.append(self.preprocess(image))
+                processed_inputs.append(image)
     
-            processed_inputs = torch.stack(processed_inputs).to(device)
+            processed_inputs = self.preprocess(processed_inputs).to(device)
             embeddings = self.get_image_features(processed_inputs)
             
             if normalize_embeddings: