Datasets:

Languages:
English
Size:
n>1T
ArXiv:
License:

Dolma1.7 books and c4 error

#33
by Chasell - opened

When using Megatron's preprocess_data.py to process the Dolma 1.7's Books and C4 datasets, I encountered errors with all the Books datasets (books-0001, books-0002, books-0003) and some of the C4 datasets (e.g., c4-0001, c4-0006). It appears there may be issues with these specific datasets. Is there something I might have missed?
屏幕截图 2024-05-10 103354.png
屏幕截图 2024-05-10 103517.png

Sign up or log in to comment