mlr3
mlr3 copied to clipboard
Possibility to save / access models via callbacks during resample
In some cases, users might want to fit large models on large datasets using resample()
.
In this case here,
the person was unable to fit the models into RAM when specifying store_models = TRUE
, but he still wanted access to the models, or rather the impurity scores from the models.
I have to suggestions to make these things more flexible:
- Adding callbacks in
resample()
would allow to save these models to disk duringresample()
. - Can we somehow score some measures that depend on the model on the workers? So that measures that require the model can still be evaluated when
store_models = FALSE
?