How to build a neural network without using keras compile method

I have the following neural network:

normalizer = preprocessing.Normalization()
normalizer.adapt(np.array(trainX))

batch_size=32
learning_rate=1e-3

model = tf.keras.Sequential([
      normalizer,
      layers.Dense(128, activation='elu', kernel_regularizer=regularizers.l2(0.01)),
      layers.Dropout(0.5),
      layers.Dense(128, activation='elu', kernel_regularizer=regularizers.l2(0.01)),
      layers.Dropout(0.5),
      layers.Dense(2),
      layers.Softmax()])

model.compile(optimizer = keras.optimizers.Adam(learning_rate=learning_rate),
              loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits = False),
              metrics = ['accuracy'])

fitted_model = model.fit(trainX, trainY, epochs=50, verbose=0, batch_size=batch_size)

I would like to know how to build this neural network without using the compile function

Also what would I need to change if I want to run it on gpu instead of cpu