diff --git a/egs/ljspeech/TTS/matcha/onnx_pretrained.py b/egs/ljspeech/TTS/matcha/onnx_pretrained.py
index f1569af3e..6d92b16eb 100755
--- a/egs/ljspeech/TTS/matcha/onnx_pretrained.py
+++ b/egs/ljspeech/TTS/matcha/onnx_pretrained.py
@@ -154,7 +154,7 @@ def main():
     params = get_parser().parse_args()
     logging.info(vars(params))
 
-    model = OnnxModel(params.acoustic_model, tokens)
+    model = OnnxModel(params.acoustic_model, params.tokens)
     vocoder = OnnxHifiGANModel(params.vocoder)
     text = params.input_text
     x = model.tokenizer.texts_to_token_ids([text], add_sos=True, add_eos=True)