metadata
dataset_info:
features:
- name: audio
dtype:
audio:
sampling_rate: 22050
- name: text
dtype: string
- name: normalized_text
dtype: string
- name: audio_token
sequence: int64
splits:
- name: train
num_bytes: 3809778094.5
num_examples: 13100
download_size: 3789647068
dataset_size: 3809778094.5
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
This dataset uses focal codec to tokenize audio in JL speech dataset.