Upload app.py

app.py

@@ -20,7 +20,7 @@ def tinyllama_chat(message, history):
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     model = AutoModelForCausalLM.from_pretrained(
         model_id,
-        device_map="cuda",
+        device_map=None, 
         torch_dtype=dtype,
     )