question : structure of dataset in localdisk

#3
by ZBW - opened

from datasets import load_dataset

without streaming

ds = load_dataset('espnet/yodas', 'en000', cache_dir='/media/yodas')
i got 2 folders: downloads and espnet__yodas
in downloads:

img1.png
so wav folder without ASR text & ASR text without wav folder ?????????????????????

ESPnet org

Since the audio content is very large, we split it into an audio directory with audio compressed into tar.gz and a text directory containing the corresponding transcription for each subset. You can see its structure here:
https://huggingface.co/datasets/espnet/yodas/tree/main/data/en000

tks for your reply, other question is whats in this folder: downloads/extracted . there are many uncompressed audio folders without txt files

ESPnet org

hi, since the original structure is to store audio/text separately in different directories, it is expected that the uncompressed audio folder does not contain text files. You can still find the corresponding transcription in the text folder though

Sign up or log in to comment