Fix vocab to_token (#570)

* fix

* fix

Author: FrostML

paddlenlp/data/vocab.py

@@ -200,10 +200,10 @@ def to_tokens(self, indices):
         if not isinstance(indices, (list, tuple, np.ndarray)):
             indices = [indices]
             to_reduce = True
-        elif isinstance(indices, (list, tuple)):
+        if isinstance(indices, (list, tuple)):
             indices = np.asarray(indices)
 
-        if len(indices.shape) > 1:
+        if isinstance(indices, (np.ndarray)) and len(indices.shape) > 1:
             raise ValueError(
                 'Token indices is invalid. Expected 1D array, but received {}D array. '.
                 format(len(indices.shape)))