Not loading entire dataset into memory to use, is it an option?

[Rhiyo]

From reading the documentation it seems you have to convert a geodataframe object into the dask-geopandas object. If you would like to use a csv or other file, this would require loading the entire dataset into memory, correct? Is there a way to be more memory efficient when loading in datasets? I originally wanted to use dask for this benefit.

[Rhiyo]

I didn't catch that parquet part in the documentation, my bad.