Large ensemble model fitting guidence
Because I've decided that bigger is always better, I'm currently fitting an ensemble cluster expansion with 1666 parameters using a growing data set of ~ 7,000 structures. I sometimes hit the wall clock and lose all my progress because fitting takes so long. It would be handy if I could checkpoint ensemble model fitting.
Alternatively, I could fit the individual models and then combine them at the end. This approach could be nice for parallelism too.