Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
jqhoogland commited on
Commit
86b5077
·
verified ·
1 Parent(s): d27ed71

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -15
README.md CHANGED
@@ -19,15 +19,9 @@ dataset_info:
19
  download_size: 116713449
20
  dataset_size: 222211313
21
  ---
22
- # Dataset: Pile - StackExchange
23
 
24
- This dataset is a subset of StackExchange data from The Pile dataset, containing text examples
25
- from this specific source.
26
-
27
- ## Dataset Source
28
-
29
- This dataset is derived from the uncopyrighted portion of The Pile dataset, specifically accessed through:
30
- - [monology/pile-uncopyrighted](https://huggingface.co/datasets/monology/pile-uncopyrighted/tree/main/train)
31
 
32
  ## Citations
33
 
@@ -46,10 +40,4 @@ author={Biderman, Stella and Bicheno, Kieran and Gao, Leo},
46
  journal={arXiv preprint arXiv:2201.07311},
47
  year={2022}
48
  }
49
- ```
50
-
51
- ## Dataset Details
52
-
53
- This dataset contains examples from the StackExchange portion of The Pile. The data maintains
54
- the same format as in the original dataset but is filtered to only include examples from
55
- this specific source.
 
19
  download_size: 116713449
20
  dataset_size: 222211313
21
  ---
 
22
 
23
+ ## Dataset Creation Process
24
+ These subsets were created by streaming over the rows from [`monology/pile-uncopyrighted`](https://huggingface.co/datasets/monology/pile-uncopyrighted) and filtering by the meta column. Each subset is generally limited to the first 100,000 qualifying rows encountered.
 
 
 
 
 
25
 
26
  ## Citations
27
 
 
40
  journal={arXiv preprint arXiv:2201.07311},
41
  year={2022}
42
  }
43
+ ```