cannot get answer with LGBMClassifier
jiahengqi opened this issue · comments
jiahengqi commented
I change xgb to lgb but can't get any return
it cost 1sec on GridSearchCV
grid=dict(num_leaves=[8,15,31],
n_estimators=[100, 200, 300])
for _ in trange(1):
model_lgb = GridSearchCV(
LGBMClassifier(),
grid, n_jobs=4, cv=3
)
model_lgb.fit(X,y)
but no return in 10 min with DistGridSearchCV
grid=dict(num_leaves=[8,15,31],
n_estimators=[100, 200, 300],
n_jobs=1)
for _ in trange(1):
model_lgb = DistGridSearchCV(
LGBMClassifier(),
grid, sc, cv=3,n_jobs=1
)
model_lgb.fit(X,y)
Evan Harris commented
Do you have LightGBM installed on all of the nodes of the cluster? Including the required bindings (https://pypi.org/project/glibc/)? This will all need to be installed using a node bootstrap.
We've never tested LightGBM with sk-dist
. It could work in theory but sk-dist
doesn't formally support it.
Evan Harris commented
We've added to the documentation around LightGBM: https://github.com/Ibotta/sk-dist#gradient-boosting