diff --git a/datasets/TTSDataset.py b/datasets/TTSDataset.py
index ecf8e9ea..0305da7f 100644
--- a/datasets/TTSDataset.py
+++ b/datasets/TTSDataset.py
@@ -102,6 +102,7 @@ class MyDataset(Dataset):
                                                                  cache_path)
         if self.enable_eos_bos:
             phonemes = pad_with_eos_bos(phonemes)
+            phonemes = np.asarray(phonemes, dtype=np.int32)
 
         return phonemes
 
diff --git a/utils/text/__init__.py b/utils/text/__init__.py
index 332163d2..77cc23a5 100644
--- a/utils/text/__init__.py
+++ b/utils/text/__init__.py
@@ -47,7 +47,7 @@ def text2phone(text, language):
 
 
 def pad_with_eos_bos(phoneme_sequence):
-    return [_PHONEMES_TO_ID[_bos]] + phoneme_sequence + [_PHONEMES_TO_ID[_eos]]
+    return [_PHONEMES_TO_ID[_bos]] + list(phoneme_sequence) + [_PHONEMES_TO_ID[_eos]]
 
 
 def phoneme_to_sequence(text, cleaner_names, language, enable_eos_bos=False):