Thanks for sharing ! Just one question about the data files
#15
by
lhoestq
HF staff
- opened
Thanks for sharing this useful dataset :)
I noticed one has to download all the files and concatenate them to get the original gzipped JSON Lines file.
Have you considered making different JSON Lines files that can be used on their own instead of splitting one big JSON Lines file ? From my experience it is generally quite practical when loading in parallel or streaming the data
This would be a good idea. I am interested in using a small subset of the dataset for training and having to download the entire dataset is quite a large undertaking just to access a subset of the dataset.