update gpu control

app.py

@@ -141,7 +141,6 @@ with gr.Blocks(theme = gr.themes.Soft()) as demo:
     input = gr.Textbox(visible=False)  # placeholder
     gr_md = gr.Markdown(mark_text + md_text_initial)
 
-    @spaces.GPU(duration=60)
     def update_with_rag_md(message, llm_results_use = 5, database_choice = index_info, llm_model_picked = 'Trust-Align-Qwen2.5'):
         chat_round = [
             {"role": "user", 

utils.py

@@ -1,3 +1,4 @@
+import spaces
 import datetime
 import string
 
@@ -163,8 +164,9 @@ def load_llama_guard(model_id = "meta-llama/Llama-Guard-3-1B"):
     UNSAFE_TOKEN_ID = llama_guard_tokenizer.convert_tokens_to_ids("unsafe")
     
     return llama_guard, llama_guard_tokenizer, UNSAFE_TOKEN_ID
-    
 
+
+@spaces.GPU(duration=60)
 def moderate(chat, model, tokenizer, UNSAFE_TOKEN_ID):
     
     prompt = tokenizer.apply_chat_template(chat, return_tensors="pt", tokenize=False)
@@ -179,7 +181,6 @@ def moderate(chat, model, tokenizer, UNSAFE_TOKEN_ID):
         return_dict_in_generate=True,
         pad_token_id=tokenizer.eos_token_id,
         output_logits=True,  # get logits
-        do_sample=False
     )
     ######
     # Get generated text