Update app.py

app.py

@@ -72,7 +72,7 @@ def predict(image):
     prompt_ids = torch.tensor(prompt_ids).unsqueeze(0)
     
     text_ids = vqa_model.generate(pixel_values=pixel_values, input_ids=prompt_ids, max_length=50)
-    text = vqa_processor.batch_decode(text_ids, skip_special_tokens=True)
+    text = vqa_processor.batch_decode(text_ids, skip_special_tokens=True)[0][len(prompt):]
     
     audio = tts(text)