Datasets:

speechcolab

/

gigaspeech

like 101

Follow

SpeechColab 24

Hey @wangweifei !

If you use the dataset using the .load_dataset method, all the data for the relevant config will be downloaded to your cache:

from datasets import load_dataset

gigaspeech = load_dataset("speechcolab/gigaspeech",  config="xs")  # to download and extract all data for the xs split to your cache under ~/.cache/huggingface/datasets

Not only this, but the data will be extracted and prepared into train/dev/test splits for you 🤗

You can also specify a cache directory with the argument cache_dir=to download the dataset to a specific directory.

If you would simply like to download the raw data without extracting it or preparing the splits (train/dev/test), you can clone this repo:

git lfs install
git clone https://huggingface.co/datasets/speechcolab/gigaspeech

Note that this will download all the data for the dataset (10k hours) and will not extract it or prepare splits.