Datasets:

facebook

/

voxpopuli

like 56

I've been trying with several languages (but not in a multilingual setting) to obtain just a split of the data, e.g. "test" or "test[:50]".
In all cases (I didn't try with "en_accented") the script download all splits (train + validation + test) and then just selects the required part, instead of just downloading the particular split.

Example of the script

from datasets import load_dataset, Audio
data_files = {"train": "data/es/test/test_part_0.tar.gz"}
language_code = "es"
dataset = load_dataset("facebook/voxpopuli", language_code,split="test[:50]")
dataset

Thanks!