Update README.md

README.md

@@ -19,7 +19,7 @@ license: mit
 ## Uses
 
 <!-- Address questions around how the dataset is intended to be used. -->
-This dataset comprises ~250B bytes of high quality text data. Should be used to train *English-only*, *very small*, *byte-level* models. 
+This dataset comprises ~250B bytes of high quality text data. Should be used to train very small (<50M parameters) byte-level English models for educational purposes and to iterate on experimental LLM architectures quickly. 
 
 ## Dataset Structure