How to save TextDataLoaders object before starting training

How to save TextDataLoaders object before training so that It can be loaded directly as TextDataLoaders.from_df() takes too much time in preprocessing?