It simply does not work.
The internal algorithm for grid search runs a combination of parameters and expects the results.
But when breedR.bin = submit, it only gets a job ID.
Doing the grid search manually is the only current way to go.
But is inefficient, specially for large grids, as it implies transferring several times the same info (notably, the dataset) back and forward.
And we are obliged to store in memory as many model fits as parameter combinations, resulting in objects of several hundreds Mb.
It simply does not work. The internal algorithm for grid search runs a combination of parameters and expects the results. But when
breedR.bin = submit, it only gets a job ID.Doing the grid search manually is the only current way to go. But is inefficient, specially for large grids, as it implies transferring several times the same info (notably, the dataset) back and forward. And we are obliged to store in memory as many model fits as parameter combinations, resulting in objects of several hundreds Mb.