---
license: mit
---
I filtered short conversations out of the dataset. 

UPCOMING: I am working on fixing newlines in the whole dataset and then also doing quality classification to find better samples. As of 4th of October around 50% of dataset (around 500MB jsonl file) has fixed newlines.