Update app.py

app.py

@@ -38,7 +38,7 @@ concurrency_limit = 5
 # Load the model
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
-    device_map=device
+    device_map=device,
     torch_dtype=torch_dtype,
     trust_remote_code=True,
 )