Trained on 5000 rows from databricks-dolly-15k-id and a certain version of the original english dataset for 3 epochs