metadata
dataset_info:
features:
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: sentence
dtype: string
- name: locale
dtype: string
splits:
- name: train
num_bytes: 87168599429
num_examples: 100000
- name: test
num_bytes: 863906806
num_examples: 1000
download_size: 79678350387
dataset_size: 88032506235
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
Build from mozilla commonvoice 13 using the script commited in this repo. Used to teach a model to ignore languages that are not french