Datasets:

Modalities:
Tabular
Text
Formats:
parquet
Languages:
English
ArXiv:
DOI:
Libraries:
Datasets
Dask
License:

Code Datasets

#30
by Taylor658 - opened

Based on the filtering steps applied is it recommended to not use code datasets like GitHub Repositories and Google’s Code Dataset for training due to the fact that these datasets will require in many cases extensive filtering and preprocessing as well and to stick to more finely curated datasets like The Stack V2?

HuggingFaceFW org

Not super familiar with those datasets but we do recommend The Stack V2 yes

ok thanks for feedback

Taylor658 changed discussion status to closed

Sign up or log in to comment