SpeechT5/SpeechUT/dataset/MuSTC/en_es/config_enes.yaml

bpe_tokenizer:
  bpe: sentencepiece
  sentencepiece_model: spm_unigram10000.model

sampling_alpha: 1.0
shuffle: false
use_audio_input: true
use_sample_rate: 16000

vocab_filename: dict.spm.txt

# required by speech_to_text task but never used  
input_channels: 1
input_feat_per_channel: 1