Update app.py

app.py

@@ -88,7 +88,7 @@ def run_example(image, text_input=None, model_id="Qwen/Qwen2-VL-7B-Instruct"):
     inputs = inputs.to("cuda")
     
     # Inference: Generation of the output
-    generated_ids = model.generate(**inputs, max_new_tokens=4096)
+    generated_ids = model.generate(**inputs, max_new_tokens=1024)
     generated_ids_trimmed = [
         out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
     ]