Trying to empty cache and load the model once

app.py

@@ -7,13 +7,15 @@ import torch
 # Use a pipeline as a high-level helper
 from transformers import pipeline
 
+torch.cuda.empty_cache()
 
+print("RUNNING PIPE")
+pipe = pipeline("text-generation", model="NousResearch/Hermes-3-Llama-3.1-8B", max_new_tokens=200, device=0)
+print("PIPE DONE") 
 
 @spaces.GPU(duration=120)
 def llama3_1_8B(question):
-    print("RUNNING PIPE")
-    pipe = pipeline("text-generation", model="NousResearch/Hermes-3-Llama-3.1-8B", max_new_tokens=200, device=0)
-    print("PIPE DONE") 
+    
 
     messages = [
     {"role": "user", "content": question},