Error when calling ModelAveraging in subsampling mode

[mnarayan]

Here is an example input where regularization is a scalar lam = 0.02566

    model = ModelAverage(
        estimator=estimator,
        n_trials=n_trials,
        init_method='spearman',
        penalization=penalization,
        subsample=.5,
        lam=lam,
        n_jobs=n_jobs)

Resulting in the following bug

/home/user/PYTHON/skggm/inverse_covariance/model_average.py in fit(self, X, y)
    382             # currently, dont estimate self.lam_ if penalty_name is different
    383             if self.penalty_name == 'lam':
--> 384                 self.lam_ += np.mean(new_estimator.lam_.flat)
    385
    386         # estimate support locations
AttributeError: 'float' object has no attribute 'flat'