Update README.md
Browse files
README.md
CHANGED
|
@@ -19,3 +19,18 @@ configs:
|
|
| 19 |
- split: train
|
| 20 |
path: data/train-*
|
| 21 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
- split: train
|
| 20 |
path: data/train-*
|
| 21 |
---
|
| 22 |
+
|
| 23 |
+
This dataset contains the following (filtered) subsets of [the Common Pile](https://huggingface.co/common-pile):
|
| 24 |
+
* DOAB
|
| 25 |
+
* Wikimedia
|
| 26 |
+
* CC Common Crawl
|
| 27 |
+
* ArXiv Papers
|
| 28 |
+
* Data Provenance Initiative
|
| 29 |
+
* PressBooks
|
| 30 |
+
* LibreTexts
|
| 31 |
+
* News
|
| 32 |
+
* OERCommons
|
| 33 |
+
* Public Domain Review
|
| 34 |
+
|
| 35 |
+
It totals around 40B tokens and is perceived to contain the highest quality sources from the original Common Pile.
|
| 36 |
+
It is intended to be compared to other methods of filtering the Common Pile (ie edu classifiers or ScalingFilter).
|