Datasets:

Modalities:
Tabular
Formats:
json
Languages:
English
Size:
< 1K
Libraries:
Datasets
pandas
Search is not available for this dataset
Wikipedia
float64
Book
float64
C4
float64
StackExchange
float64
Github
float64
CommonCrawl
float64
ArXiv
float64
11.963032
12.579061
79.872896
10.088908
15.634722
155.887114
13.252597

The SlimPajama_300B is a 300B token sample of de-duplicated Slim Pajama dataset tokenized using the EleutherAI/gpt-neox-20b tokenizer

Due to file size constraints, C4 and CommonCrawl has been uploaded in multiple chunks, you can use the following commands to merge them back into a single file:

cat C4_part_* > C4.bin
cat CommonCrawl_part_* > CommonCrawl.bin

Data Distribution

Data source Composition
Commoncrawl 0.5208
C4 0.2668
GitHub 0.0522
Books 0.0420
ArXiv 0.0442
Wikpedia 0.0399
StackExchange 0.0337
Downloads last month
128