search

rom1504 / cc2dataset

Easily convert common crawl to a dataset of caption and document. Image/text Audio/text Video/text, ...
MIT License
307 stars 23 forks source link

implement multi steps processing to limit disk space need #10

Closed rom1504 closed 1 year ago

rom1504 commented 1 year ago

add a --multi_steps option that will first prepare n groups, then process each individually, then redo the dedup