Update app.py

app.py

@@ -1,3 +1,4 @@
+import spaces
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
 from accelerate import init_empty_weights, load_checkpoint_and_dispatch
@@ -49,6 +50,7 @@ except Exception:
     pass
 
 # === RESPONSE GENERATOR ===
+@spaces.GPU()
 def generate_response(user_message, max_tokens, temperature, top_k, top_p, repetition_penalty, history_state):
     if not user_message.strip():
         return history_state, history_state