model_ug_dmm = Doc2Vec.load('d2v_model_ug_dmm.doc2vec')
model_ug_dmm.delete_temporary_training_data(keep_doctags_vectors=True, keep_inference=True)

def get_w2v_ugdbowdmm(tweet, size):
    vec = np.zeros(size).reshape((1, size))
    count = 0.
    for word in tweet.split():
        try:
            vec += np.append(model_ug_dbow[word],model_ug_dmm[word]).reshape((1, size))
            count += 1.
        except KeyError:
            continue
    if count != 0:
        vec /= count
    return vec

train_vecs_w2v_dbowdmm = np.concatenate([get_w2v_ugdbowdmm(z, 200) for z in x_train])
validation_vecs_w2v_dbowdmm = np.concatenate([get_w2v_ugdbowdmm(z, 200) for z in x_validation])

clf = LogisticRegression()
clf.fit(train_vecs_w2v_dbowdmm, y_train)
clf.score(validation_vecs_w2v_dbowdmm, y_validation)