datablations/oscar-filter
this is the one where we build the suffix array for 25% Oscar and only deduplicate that part - by deduplication I mean removing any document which has an at least 100-char span overlapping with another document in the 25% chunk. This is very strict and preserves only about 20 million documents, so less then 5% of the full Oscar.
02.2k
this is the one where we build the suffix array for 25% Oscar and only deduplicate that part - by deduplication I mean removing any document which has an at least 100-char span overlapping with another document in the 25% chunk. This is very strict and preserves only about 20 million documents, so less then 5% of the full Oscar.