Update app.py

app.py

@@ -90,7 +90,7 @@ def setup_llm(model_name="google/flan-t5-small"):
                 max_new_tokens=300,
                 temperature=0.3,
                 do_sample=True,
-                device_map="auto" if st.secrets.get("DEVICE", "cpu") == "gpu" else None
+                device=-1
             )
             
             llm = HuggingFacePipeline(pipeline=pipe)