What's the source data to build this chunked data? Can you provide a link?
#1
by ZetangForward - opened
See title, thx.
Hi. This is based on DKYoon/SlimPajama-6B. The sequences are concat/chunked to 32k token length with llama3's tokenizer.
ZetangForward changed discussion title from What's the ousce data to make this chunked data? Can you provide a link? to What's the sousce data to make this chunked data? Can you provide a link?
ZetangForward changed discussion title from What's the sousce data to make this chunked data? Can you provide a link? to What's the source data to build this chunked data? Can you provide a link?