Update modeling_gpt2vision.py

modeling_gpt2vision.py

@@ -69,9 +69,8 @@ class GPT2Vision(PreTrainedModel):
     def generate(self, question, image, max_new_tokens=30, **kwargs):
         # Process the image
         # Convert the image to a tensor and add a batch dimension
-        image_tensor = self.vision_encoder.image_transform(image).unsqueeze(0).to(self.device)
         with torch.no_grad():
-            img_features = self.vision_encoder(image_tensor,device=self.device)
+            img_features = self.vision_encoder(image,device=self.device)
         img_embs = self.mlp(img_features)
         
         # Tokenize the question