metadata
dataset_info:
features:
- name: audio_tokens
sequence: int64
- name: raw_text
dtype: string
- name: transcript
dtype: string
- name: entities
sequence: string
- name: prompt
dtype: string
- name: input_ids
sequence: int64
- name: attention_mask
sequence: int64
splits:
- name: train
num_bytes: 2895627778
num_examples: 70397
download_size: 184378881
dataset_size: 2895627778
configs:
- config_name: default
data_files:
- split: train
path: data/train-*