Merge branch 'main' of https://huggingface.co/datasets/stanford-crfm/DSIR-filtered-pile-50M into main

README.md

@@ -9,13 +9,13 @@ size_categories:
 
 ## Dataset Description
 
-- **Repository:** 
+- **Repository:** https://github.com/p-lambda/dsir
 - **Paper:** 
 - **Point of Contact: Sang Michael Xie <[email protected]>** 
 
 ### Dataset Summary
 
-This dataset is a subset of The Pile, selected via the DSIR data selection method. The target distribution for DSIR is the Wikipedia and book-related subsets of The Pile.
+This dataset is a subset of The Pile, selected via the DSIR data selection method. The target distribution for DSIR is the Wikipedia and BookCorpus2 subsets of The Pile.
 
 
 ### Languages
@@ -24,7 +24,7 @@ English (EN)
 
 ## Dataset Structure
 
-A train set is provided, which can be split into validation and test if needed.
+A train set is provided (51M examples) with a small validation and test set (50k examples each).
 
 ### Data Instances
 
@@ -71,10 +71,11 @@ The dataset is biased towards choosing data from non-Wikipedia and non-Books sou
 Sang Michael Xie
 
 ### Citation Information
-
+```
 @article{xie2023data,
   author = {Sang Michael Xie and Shibani Santurkar and Tengyu Ma and Percy Liang},
   journal = {TODO},
   title = {Data Selection for Language Models via Importance Resampling},
   year = {2023},
 }
+```