After using the parquet_build.py you will have a parquet folder which contains the parquet data from the 22.34 Gb Library_Collection_Inventory.csv in the collection folder. Notice that the data is stored in chunks such that each file is smaller than the 100 Mb limit that Github enforces. Due to the efficient storage and compression the 22.34 Gb .csv format of the data is reduced to 5.66 Gb. In the checkouts folder the data from 4.89 Gb Checkouts_By_Title__Physical_Items_.csv file will only require 0.97 Gb of space.
Github does ask that repositories not store more than 1 Gb of data with stronger warning if it is over 5 Gb so we have not included these Parquet files here. It is amazing that we went from over 25 Gb of data in the .csv format to about 6 Gb of data in the .parquet format. In addition, the .parquet format contains the data type and datetime information as saved. As we know, .csv files do not store that information.