diff --git a/paddlespeech/cli/README.md b/paddlespeech/cli/README.md
index bd6572f194edb685ddb20f0cad5cb7687591f98a..56afb939c2a36f998a70993f44f4d402f6538a3d 100644
--- a/paddlespeech/cli/README.md
+++ b/paddlespeech/cli/README.md
@@ -6,4 +6,4 @@
  `paddlespeech help`
 
  ## ASR
- `paddlespeech asr --input ./test_audio.wav --device gpu`
+ `paddlespeech asr --input ./test_audio.wav`
diff --git a/paddlespeech/cli/asr/infer.py b/paddlespeech/cli/asr/infer.py
index 605163803a7e5b3e3b816060d7d37fd804d6a3a8..e5c64e9ab38d7dc9b4e421b67043c1be1b33fbc4 100644
--- a/paddlespeech/cli/asr/infer.py
+++ b/paddlespeech/cli/asr/infer.py
@@ -87,7 +87,7 @@ class ASRExecutor(BaseExecutor):
         self.parser.add_argument(
             '--device',
             type=str,
-            default='cpu',
+            default=paddle.get_device(),
             help='Choose device to execute model inference.')
 
     def _get_pretrained_path(self, tag: str) -> os.PathLike: