Sourced from: https://gist.github.com/bartowski1182/eb213dccb3571f863da82e99418f81e8
calibration_datav3.txt
is the original data (64181 tokens)calibration_datav3_small.txt
is up to line 1241 of the original data (30173 tokens)
Using the full dataset will take the longest but also provide the most accurate calibration.