Config option to not save dataset

I've noticed the `datasets` directory can be 40GB, which if I have separate runs for different scales, can take up 400GB+. Can we have an option in the config to not save the datasets? Isn't the runtime implication on regenerating the dataset small (I think minutes) compared to runs that take multiple hours?