Datasets:

Modalities:
Tabular
Formats:
json
Languages:
English
Size:
< 1K
Libraries:
Datasets
pandas
tejasexpress commited on
Commit
4fcad1b
·
verified ·
1 Parent(s): 02737c4

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +19 -0
README.md ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ The SlimPajama_300B is a 300B token sample of de-duplicated Slim Pajama dataset tokenized using the [EleutherAI/gpt-neox-20b](https://huggingface.co/EleutherAI/gpt-neox-20b) tokenizer
2
+
3
+ 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:
4
+ ```bash
5
+ cat C4_part_* > C4.bin
6
+ cat CommonCrawl_part_* > CommonCrawl.bin
7
+ ```
8
+
9
+ #### Data Distribution
10
+
11
+ | Data source | Composition |
12
+ | ------------- | ------------------------------- |
13
+ | Commoncrawl | 0.5208 |
14
+ | C4 | 0.2668 |
15
+ | GitHub | 0.0522 |
16
+ | Books | 0.0420 |
17
+ | ArXiv | 0.0442 |
18
+ | Wikpedia | 0.0399 |
19
+ | StackExchange | 0.0337 |