Why wav2vec2-base-960h is trained without using attention mask?

[YWMditto]

I have seen the code of Wav2Vec2FeatureExtractor in transformers, and it said the model wav2vec2-base-960h is trained without using attention mask.

I wonder why and how the model is trained without using attention mask to mask the pad place.

Does not it make mistakes when allowing the pad place into compute?