UPDATE

app.py

@@ -38,7 +38,7 @@ generate_kwargs = dict(
 llama_guard, llama_guard_tokenizer, UNSAFE_TOKEN_ID = load_llama_guard("meta-llama/Llama-Guard-3-1B")
 
 ## RAG MODEL
-RAG = RAGPretrainedModel.from_index("colbert/indexes/arxiv_colbert", n_gpu=1)
+RAG = RAGPretrainedModel.from_index("colbert/indexes/arxiv_colbert", n_gpu=0)
 
 try:
   gr.Info("Setting up retriever, please wait...")
@@ -141,7 +141,7 @@ 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)
+    # @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",