520jefferson commited on
Commit
d5b28b1
1 Parent(s): 042acd6

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -1
README.md CHANGED
@@ -16,6 +16,7 @@ size_categories:
16
 
17
  ## Dataset Summary
18
  SkyPile-150B is created to serve as a large-scale Chinese dataset for pre-training of large language models. It is taken from publicly available Chinese Internet web pages. Through strict filtering and extensive deduplication then sensitive data filtering, At the same time, fastText and Bert are used to filter low-quality data.
 
19
  The public part of SkyPile-150B contains approximately 166M individual web pages, with an average of over 1k Chinese characters, a total of approximately 150B tokens, and 592G plain text data.
20
 
21
  ## Language
@@ -29,4 +30,4 @@ Because SkyPile-150B is built on publicly available web pages, it may contain se
29
  Since toxic or biased data is prevalent on the Internet, although we filter it using specific URL filtering methods, users should be aware.
30
 
31
  ## Social Impact of Dataset
32
- With the open source release of SkyPile-150B, we aim to increase access to high-quality web data which has typically been held private by model developers. We believe this release will in turn improve the accessibility and the spread of performant large language models.
 
16
 
17
  ## Dataset Summary
18
  SkyPile-150B is created to serve as a large-scale Chinese dataset for pre-training of large language models. It is taken from publicly available Chinese Internet web pages. Through strict filtering and extensive deduplication then sensitive data filtering, At the same time, fastText and Bert are used to filter low-quality data.
19
+
20
  The public part of SkyPile-150B contains approximately 166M individual web pages, with an average of over 1k Chinese characters, a total of approximately 150B tokens, and 592G plain text data.
21
 
22
  ## Language
 
30
  Since toxic or biased data is prevalent on the Internet, although we filter it using specific URL filtering methods, users should be aware.
31
 
32
  ## Social Impact of Dataset
33
+ With the open source release of SkyPile-150B, we aim to increase access to high-quality web data which has typically been held private by model developers. We believe this release will in turn improve the accessibility and the spread of performant large language models.