How to build model with 3D array label with TensorFlow - tensorflow

Question

I have data predictor in 2D array like this below
array([[ 0, 0, 0, ..., 10, 6, 1],
[ 0, 0, 0, ..., 12, 6, 1],
[ 0, 0, 0, ..., 8, 6, 1],
...,
[ 0, 0, 0, ..., 54, 30, 60],
[ 0, 0, 0, ..., 1472, 5, 348],
[ 0, 0, 0, ..., 58, 45, 60]])
and label data with 3D shape like this below
[array([[0., 0., 1., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
...,
[0., 1., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.]], dtype=float32),
array([[0., 0., 1., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
...,
...
...,
[0., 1., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.],
[0., 0., 0., ..., 0., 0., 0.]], dtype=float32),
...]
I would build the model with TensorFlow
model = Sequential()
model.add(Embedding(total_words, 10, input_length=max_sequence_len))
model.add(LSTM(100))
model.add(Dropout(0.1))
model.add(Dense(total_words_label, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam')
But, I get error in latest Dense
ValueError: Shapes (None, 10, 4991) and (None, 4991) are incompatible
It looks like from latest Dense shape because I have 3D label. How the way I can adjust it? I'm learning about TensorFlow.
And, how to convert the output to the 3D label when I have predict it?
--- update (adding the process) --
the example data
input output/label
Hartford Avenue, Bellingham, MA, 2019 {'address': 'Hartford Avenue', 'city': 'Bellingham', 'state': 'MA','zip': '2019'}
Oak Street, Brockton, MA, 2301 {'address': 'Oak Street', 'city': 'Brockton', 'state': 'MA', 'zip': '2301'}
the input is as predictor and the output/label is as label
The process of creating label
tokenizer_label = Tokenizer()
l = list(all['input'])
tokenizer_label.fit_on_texts(l)
total_words_label = len(tokenizer_label.word_index) + 1
labels = []
for i in range(len(l)):
labels.append(keras_utils.to_categorical(tokenizer_label.texts_to_sequences(l[i].split()), num_classes=total_words_label))