Not able to generate a random array of categorical labels

I have a random state with a fixed seed I am using to make my predictive results replicable:

rng = np.random.RandomState(101)
len(finalTestSentences) = 500

I am trying to use this seed to generate an array of random categorical variables using a unique list of possibilities. Here's an example (I used set on my training labels to generate the unique labels I have):

unique_train_labels = set([u'bla', u'blee', u'blu', u'ma'])

I then do:

categorical_random = np.array[rng.choice(unique_train_labels, len(finalTestSentences))]

But I am getting:

categorical_random = np.array[rng.choice(unique_train_labels, len(finalTestSentences))]
  File "mtrand.pyx", line 1072, in mtrand.RandomState.choice (numpy/random/mtrand/mtrand.c:11545)
ValueError: a must be 1-dimensional or an integer

Note, this was from this workaround here

What is going wrong and how can I achieve my aim?