return attention mask in int32 (#13543)

src/transformers/models/speech_to_text/feature_extraction_speech_to_text.py

@@ -240,12 +240,12 @@ class Speech2TextFeatureExtractor(SequenceFeatureExtractor):
 
         attention_mask = padded_inputs.get("attention_mask")
         if attention_mask is not None:
-            padded_inputs["attention_mask"] = [np.asarray(array, dtype=np.bool) for array in attention_mask]
+            padded_inputs["attention_mask"] = [np.asarray(array, dtype=np.int32) for array in attention_mask]
 
         # Utterance-level cepstral mean and variance normalization
         if self.do_ceptral_normalize:
             attention_mask = (
-                np.array(attention_mask, dtype=np.bool)
+                np.array(attention_mask, dtype=np.int32)
                 if self._get_padding_strategies(padding, max_length=max_length) is not PaddingStrategy.DO_NOT_PAD
                 else None
             )

src/transformers/models/wav2vec2/feature_extraction_wav2vec2.py

@@ -86,7 +86,7 @@ class Wav2Vec2FeatureExtractor(SequenceFeatureExtractor):
         Every array in the list is normalized to have zero mean and unit variance
         """
         if attention_mask is not None:
-            attention_mask = np.array(attention_mask, np.bool)
+            attention_mask = np.array(attention_mask, np.int32)
             normed_input_values = []
 
             for vector, length in zip(input_values, attention_mask.sum(-1)):
@@ -216,7 +216,7 @@ class Wav2Vec2FeatureExtractor(SequenceFeatureExtractor):
         # convert attention_mask to correct format
         attention_mask = padded_inputs.get("attention_mask")
         if attention_mask is not None:
-            padded_inputs["attention_mask"] = [np.asarray(array, dtype=np.bool) for array in attention_mask]
+            padded_inputs["attention_mask"] = [np.asarray(array, dtype=np.int32) for array in attention_mask]
 
         # zero-mean and unit-variance normalization
         if self.do_normalize: