I'm doing the EM clustering using 3 components on a dataset (x), that is just dataframe with 15 features.
from sklearn import mixture
import pandas as pd
em = mixture.GMM(n_components=3)
x['CLUSTER'] = pd.Series(em.weights_, index=x.index).astype(str)
In order to get "labels" you need to call
.weights are (one of many!) parameters of the fitted distribution, not point-wise labels.
x['CLUSTER'] = em.predict(x)