For those of you that downloaded the CHARTEVENTS.csv.gz file from my personal webpage, please check a new version with the right size that I've just uploaded. The file available in that site was less than 4 GBytes in size which does not correspond to the right actual size when decompressed: ~30 GBytes.
One more hint about this assignment: it would be convenient/useful to aggregate value items of patients (min, max and average values) and try to reduce the number of features by selecting the top most frequent items and the top less frequent items.