Update README.md

README.md

@@ -19,15 +19,9 @@ configs:
   - split: train
     path: data/train-*
 ---
-# Dataset: Pile - Ubuntu Irc
 
-This dataset is a subset of Ubuntu Irc data from The Pile dataset, containing text examples 
-from this specific source.
-
-## Dataset Source
-
-This dataset is derived from the uncopyrighted portion of The Pile dataset, specifically accessed through:
-- [monology/pile-uncopyrighted](https://huggingface.co/datasets/monology/pile-uncopyrighted/tree/main/train)
+## Dataset Creation Process
+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.
 
 ## Citations
 
@@ -48,8 +42,7 @@ year={2022}
 }
 ```
 
-## Dataset Details
+**Note**: Some subsets contain fewer than 100,000 rows due to the sparsity of certain categories in the original dataset. In these cases, the rows may not be the first encountered but are unevenly distributed due to parallelization during collection.
 
-This dataset contains examples from the Ubuntu Irc portion of The Pile. The data maintains 
-the same format as in the original dataset but is filtered to only include examples from 
-this specific source.
+## Future Updates
+Future updates may include completing the smaller subsets to reach the 100,000 row target, which could result in breaking changes for those specific subsets.