Update app.py

app.py

@@ -33,6 +33,8 @@ def predict(image, text):
     inputs = processor(image, input_text, return_tensors="pt").to(model.device)
     outputs = model.generate(**inputs, max_new_tokens=250)
     response = processor.decode(outputs[0], skip_special_tokens=True)
+    # Split the response at the first occurrence of "assistant" and return only the part after it for a clean output.
+    response = response.split("assistant", 1)[1].strip()
     return f"\n{response}"
 
 # Gradio