GridSearchCV does not pick up default Client

[mrocklin]

from dask.distributed import Client
client = Client()

import numpy as np
from sklearn.linear_model import SGDClassifier

model = SGDClassifier(    
    max_iter=1.0, 
)

X = np.random.random((1000, 10))
y = np.random.randint(0, 10, size=(1000,))

from dask_ml.model_selection import GridSearchCV
params = {'alpha': np.logspace(-4, 1, num=20)}
grid = GridSearchCV(model, params, return_train_score=True)
_ = grid.fit(X, y)

assert client.cluster.scheduler.transition_log

My guess is that this changed in the 0.18.0 Dask update. The right function to use today is maybe

https://github.com/dask/dask/blob/master/dask/base.py#L841-L853