Currently the end user has to decide if an experiment should be ran in memory mode or ondisk mode or trust that the default parameter for determining a large vs. small experiment is sufficient for their experiment. A simple caching mechanism in samples.py would eliminate the need for any hard coding here.
Currently the end user has to decide if an experiment should be ran in memory mode or ondisk mode or trust that the default parameter for determining a large vs. small experiment is sufficient for their experiment. A simple caching mechanism in samples.py would eliminate the need for any hard coding here.