Update app.py

app.py

@@ -3,6 +3,7 @@ from transformers import TextIteratorStreamer
 from threading import Thread
 from transformers import StoppingCriteria, StoppingCriteriaList
 import torch
+import spaces
 import os 
 model_name = "microsoft/Phi-3-medium-128k-instruct"
 from transformers import AutoModelForCausalLM, AutoTokenizer
@@ -16,6 +17,7 @@ class StopOnTokens(StoppingCriteria):
             if input_ids[0][-1] == stop_id:
                 return True
         return False
+@spaces.GPU()
 def predict(message, history):
     history_transformer_format = history + [[message, ""]]
     stop = StopOnTokens()