I read a large parquet file (does not fit in memory), with dask.
df = dd.read_parquet(\'100-GBs-of-data.parquet\', engine=\'pyarrow\')
I