Update llm.py

llm.py

@@ -9,6 +9,7 @@ from transformers import AutoTokenizer, AutoModelForCausalLM
 from transformers import TextIteratorStreamer
 import threading
 import torch
+import spaces
 
 import logging
 logger = logging.getLogger(__name__)
@@ -79,6 +80,7 @@ def build_messages(
 #
 # Stream response
 #
+@spaces.GPU
 @torch.inference_mode()
 def stream_response(
         messages: list[dict],