Current memory constraints.

[jhourani]

Currently, gp_baseline.py consumes a lot of memory. It runs on an 8GB machine, provided the PubChem dataset is not included. With PubChem, the memory usage can get above 30GB! Replacing the underlying dictionaries in parsed.py with a database (Python's DBM module) has shown to vastly improve memory usage over a subset of the data. The design would have to be modified slightly, so that the database is not being opened/closed with every iteration of the parser (potentially tens of millions).