머신러닝 기초 6 - 모델의 저장과 로딩

케라스로 MNIST 데이터셋 분류 파이썬 예제 저장하기

# save to model
import tensorflow as tf
import matplotlib.pyplot as plt

mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train/255.0, x_test / 255.0

plt.figure(figsize=(8,2)) # 8*2 inchs

for i in range(36):
    plt.subplot(3,12,i+1)
    plt.imshow(x_train[i], cmap="gray")
    plt.axis("off")
plt.show()

# define model
model = tf.keras.models.Sequential([
    tf.keras.layers.Flatten(input_shape=(28,28)),
    tf.keras.layers.Dense(10,input_dim=784, activation='softmax')
])

# model compile
model.compile(optimizer='adam',
             loss='sparse_categorical_crossentropy',
             metrics=["accuracy"])

model.summary()

# model fit
hist=model.fit(x_train, y_train,
              validation_data=(x_test, y_test),
              verbose=2, batch_size=100, epochs=15,
              use_multiprocessing=True)

# model evaluate
model.evaluate(x_test, y_test,
              verbose=2, batch_size=100, use_multiprocessing=True)

# save model to h5 file
file_name="softmax_mnist_model.h5"
model.save(file_name)
print(f"\nThis model has been saved to {file_name}.")

# Graph print
plt.figure(figsize=(8,4))
plt.subplot(1,2,1)
plt.plot(hist.history['loss'])
plt.title("Cost Graph")
plt.ylabel("cost")

plt.subplot(1,2,2)
plt.title("Accuracy Graph")
plt.ylabel("accuracy")
plt.plot(hist.history['accuracy'], 'b-', label="training accuracy")
plt.plot(hist.history['val_accuracy'], 'r:', label="validation accuracy")
plt.legend()
plt.tight_layout()
plt.show()

# use model
prediction = model.predict(x_test[:1, :])
prediction_class = tf.argmax(prediction, 1)
print(f"\nPrediction Result:\n{prediction}")
print("Predicted class: ", prediction_class.numpy())

plt.imshow(x_test[prediction_class[0]])
plt.grid(False)
plt.axis("off")
plt.show()

 

- 모델을 저장하는 부분

# save model to h5 file
file_name="softmax_mnist_model.h5"
model.save(file_name)
print(f"\nThis model has been saved to {file_name}.")

 

 

 

저장한 모델을 로드하여 사용하는 파이썬 예제

# load to model
import tensorflow as tf
import matplotlib.pyplot as plt

mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train/255.0, x_test / 255.0

plt.figure(figsize=(8,2)) # 8*2 inchs

for i in range(36):
    plt.subplot(3,12,i+1)
    plt.imshow(x_train[i], cmap="gray")
    plt.axis("off")
plt.show()

# load model
file_name="softmax_mnist_model.h5"
model = tf.keras.models.load_model(file_name)


# model compile
model.compile(optimizer='adam',
             loss='sparse_categorical_crossentropy',
             metrics=["accuracy"])

# model fit
hist=model.fit(x_train, y_train,
              validation_data=(x_test, y_test),
              verbose=2, batch_size=100, epochs=15,
              use_multiprocessing=True)

# model evaluate
model.evaluate(x_test, y_test,
              verbose=2, batch_size=100, use_multiprocessing=True)

# Graph print
plt.figure(figsize=(8,4))
plt.subplot(1,2,1)
plt.plot(hist.history['loss'])
plt.title("Cost Graph")
plt.ylabel("cost")

plt.subplot(1,2,2)
plt.title("Accuracy Graph")
plt.ylabel("accuracy")
plt.plot(hist.history['accuracy'], 'b-', label="training accuracy")
plt.plot(hist.history['val_accuracy'], 'r:', label="validation accuracy")
plt.legend()
plt.tight_layout()
plt.show()

# use model
prediction = model.predict(x_test[:1, :])
prediction_class = tf.argmax(prediction, 1)
print(f"\nPrediction Result:\n{prediction}")
print("Predicted class: ", prediction_class.numpy())

plt.imshow(x_test[prediction_class[0]])
plt.grid(False)
plt.axis("off")
plt.show()

 

- 모델을 로드 하는 부분

# load model
file_name="softmax_mnist_model.h5"
model = tf.keras.models.load_model(file_name)