Update app.py

app.py

@@ -90,9 +90,7 @@ def predict(message, history, system_prompt, temperature, max_new_tokens, top_k,
 
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
 model = AutoModelForCausalLM.from_pretrained(
-    MODEL_ID,
-    device_map="auto",
-)
+    MODEL_ID)
 
 # Create Gradio interface
 gr.ChatInterface(