Am i using randomForest Classifier with gridsearch wrong & is xgboost supported?

[zahs123]

What is your question?
I am trying to get a quicker gridsearch then using the sklearn version. I have tried the following as a experument:

X = np.random.normal(size=(10,4)).astype(np.float32)
y = np.asarray([0,1]*5, dtype=np.int32)
rf_params = {'n_estimators': [10,12] }
cuml_model = RandomForestClassifier()
grid = dcv.GridSearchCV(cuml_model, rf_params,scoring='accuracy' )
%time grid.fit(X,y) 

the above works,

but when i use the cuml version of gridsearch:

clf = cuRFC( max_depth=16,max_features=1.0,)
grid = dcv.GridSearchCV(clf, rf_params, scoring='accuracy')
grid.fit(X,y)

i get the error:
AttributeError: 'NoneType' object has no attribute 'fit'

why am i getting this error? Am i doing something wrong

[teju85]

The issue reported regarding RF seems to be the same as in #4193 .

@RAMitchell , @trivialfis and/or @hcho3 do you folks know if we can do cuml.model_selection.GridSearchCV (which is basically sklearn.model_selection.GridSearchCV) with xgboost?

[trivialfis]

With xgboost and gradient boosted model, yes. With xgboost-dask, no. I haven't tried xgboost random forest, but should be no too.