Setting the initial state of an RNN represented as a Keras sequential model

How do I set the initial state of the recurrent neural network rnn constructed below?

from tensorflow.keras.layers import Dense, SimpleRNN
from tensorflow.keras.models import Sequential

rnn = Sequential([SimpleRNN(3), Dense(1)])

I'd like to specify the initial state of the first layer before fitting the model with model.fit.

from Setting the initial state of an RNN represented as a Keras sequential model