생성된 네트워크를 활용하여 모델을 훈련시켜 봅시다(RNNCell과 동일한 코드이지만 결과를 확인하려고 또 한 번 실행합니다).
코드 7-16 모델 훈련
import time
units = 64
epochs = 4
t0 = time.time()
model = LSTM_Build(units)
model.compile(optimizer=tf.keras.optimizers.Adam(0.001),
loss=tf.losses.BinaryCrossentropy(),
metrics=['accuracy'],
experimental_run_tf_function=False)
model.fit(train_data, epochs=epochs, validation_data=test_data, validation_freq=2)
다음은 모델을 훈련시킨 결과입니다.
Epoch 1/4
195/195 [==============================] - 24s 124ms/step - loss: 0.4885 - accuracy: 0.7488
Epoch 2/4
195/195 [==============================] - 37s 192ms/step - loss: 0.3153 - accuracy: 0.8706 - val_loss: 0.3548 - val_accuracy: 0.8423
Epoch 3/4
195/195 [==============================] - 23s 116ms/step - loss: 0.2572 - accuracy: 0.8968
Epoch 4/4
195/195 [==============================] - 35s 179ms/step - loss: 0.2148 - accuracy: 0.9183 - val_loss: 0.3987 - val_accuracy: 0.8340
<tensorflow.python.keras.callbacks.History at 0x21fbac26c88>