Closed alvarouc closed 7 years ago
Sorry for the long wait here. I'm actually slightly against this, as I'd rather not silently ignore misspelled parameters (e.g. eror_score
) or parameters we don't yet support (e.g. verbose
). I've experienced bugs in the past that were difficult to diagnose because of behavior like this. The same could be accomplished in your code with a small branch:
if use_dask_grid_search:
est = DaskGridSearchCV(model, grid, **params)
else:
est = GridSearchCV(model, grid, n_jobs=4, **params)
Totally understand your concern. I tend to prefer portability but that its totally up to you.
n_jobs
is now supported in the recent release.
Ignoring n_jobs helps with compatibility while porting code from sklearn to dask-learn