GridSearchCV not working with h2o4gpu.RandomForestClassifier()

[Nezafati]

Please do not use issues to ask questions. For questions use either StackOverflow (h2o4gpu tag) or Gitter.

GitHub issues are used for:

• Bug reporting.
• Feature/enhancement requests and discussion.

Before submitting a new issue please make sure a similar issue is not already present in the tracker.

Please make sure to fill out the form below as best as you can. This will help both you and us.

---

Environment (for bugs)

• OS platform, distribution and version: Linux Ubuntu 16.04
• Installed from (source or binary): binary
• Version: h2o4gpu 0.2.0
• Python version (optional): 3.6
• CUDA/cuDNN version: cuda-9.0
• GPU model (optional): Geforce Titan (Maxwell)
• CPU model: Intel Xeon
• RAM available: 16GB

Description

Error message when I try .fit() method for sklearn.model_selection.GridSearch():

Code to reproduce:

import h2o4gpu from sklearn.datasets import load_breast_cancer from sklearn.model_selection import GridSearchCV

dat = load_breast_cancer() X = dat.data y = dat.target

clf_gpu = h2o4gpu.RandomForestClassifier() parameters_rf = {'n_estimators': [100, 200]} search_gpu = GridSearchCV(estimator=clf_gpu, param_grid=parameters_rf, cv=10) search_gpu.fit(X,y)

---

TypeError Traceback (most recent call last) in () 2 parameters_rf = {'n_estimators': [100, 200]} 3 search_gpu = GridSearchCV(estimator=clf_gpu, param_grid=parameters_rf, cv=10) ----> 4 search_gpu.fit(X,y)

~/anaconda2/envs/py36/lib/python3.6/site-packages/sklearn/model_selection/_search.py in fit(self, X, y, groups, **fit_params) 623 n_candidates * n_splits)) 624 --> 625 base_estimator = clone(self.estimator) 626 pre_dispatch = self.pre_dispatch 627

~/anaconda2/envs/py36/lib/python3.6/site-packages/sklearn/base.py in clone(estimator, safe) 58 % (repr(estimator), type(estimator))) 59 klass = estimator.class ---> 60 new_object_params = estimator.get_params(deep=False) 61 for name, param in six.iteritems(new_object_params): 62 new_object_params[name] = clone(param, safe=False)

TypeError: get_params() got an unexpected keyword argument 'deep'

[mdymczyk]

@Nezafati thanks for submitting the issue! Looks like our RandomForestClassifier isn't implementing fully the http://scikit-learn.org/stable/modules/generated/sklearn.base.BaseEstimator.html interface. Will look into it (looks like more of our estimators have the same issue).

You can try writing a wrapper for the time being, something like (untested):

import h2o4gpu

class RandomForestClassifierWrapper(h2o4gpu.RandomForestClassifier):
    def get_params(self, deep=True):
        return super().get_params()

And try using that wrapper instead.

[o-d-t]

Hi,

I am suffering from the same issue, using GridSearchCV with h2o4gpu.Ridge().

I wrote the following wrapper to the top of my code:

class RidgeWrapper(h2o4gpu.Ridge):               
    def get_params(self,deep=True):
        return super().get_params()

and initialize my model using:

    premodel=h2o4gpu.Ridge(fit_intercept=False,tol=0.0001)
    model=RidgeWrapper(premodel)

which results with the same error. Can you please help me with a temporary solution for this issue?

[kellinpelrine]

I was able to get it to work using the suggested wrapper for random forest like this:

class RandomForestClassifierWrapper(h2o4gpu.RandomForestClassifier):
    def get_params(self, deep=True):
        return super().get_params()

rf = RandomForestClassifierWrapper(class_weight=train_proportions)
rf_random = RandomizedSearchCV(estimator = rf, param_distributions = random_grid, n_iter = 100, cv = 3, n_jobs = 1)

Thank you!