Enable data streaming
#1
by
lhoestq
HF staff
- opened
using dl_manager.iter_archive in TAR archives
This allows to stream the data using
from datasets import load_dataset
ds = load_dataset("dell-research-harvard/AmericanStories", streaming=True) # pass revision="refs/pr/1" to try the code from this PR
for example in ds["1774"]:
print(example)
This should also fix the dataset viewer, which currently shows an error
Thanks for your contribution! This functionality has been added via a separate commit. The above streaming code now runs as expected.
tombryan
changed pull request status to
closed