Cho mình xin dữ liệu sau khi đã lọc "71GB of long paragraphs for 1024-length model"

[justinphan3110]

You can try to download the original file from CC100 and filter them by the length of your choice. We didn't do any extra preprocess for ViT5.

!wget http://data.statmt.org/cc-100/vi.txt.xz
!unxz vi.txt.xz