import tensorflow as tf
from tensorflow.keras.layers import Dense, Flatten, Conv2D
from tensorflow.keras import Model
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
x_train.shape
# Add a channels dimension
# 多增加第4個 dimension:"channel"
x_train = x_train[..., tf.newaxis]
x_test = x_test[..., tf.newaxis]
x_train.shape
#== tf.data API 來建立 input pipeline ==#
# * tf.data
#1."from_tensor_slices" 建立一個 Dataset Object
# 在使用"from_tensor_slices"參數（x_train, y_train）是
# 包含很多筆類型一樣的data,像這個例子有 60000筆data。
# 不要跟"from_tensor"搞混了!
#2.將training data做shuffle打亂 和 將batch size設定為 32
train_dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train))
train_dataset = train_dataset.shuffle(10000).batch(32)
test_dataset = tf.data.Dataset.from_tensor_slices((x_test, y_test))
test_dataset = test_dataset.batch(32)
#== 使用 Keras 自訂一個 Model ==#
# * tf.Keras.Model
# * tf.Keras.layers
#1.建立一個 Class並繼承 tf.Keras.Model
#2.__init__裡面我們可以自己設計每一層 Layer長什麼樣子，
# 是哪種 Layer(Convolution, dense...)
#3.call定義了這個 Model的 forward pass
class MyModel(Model):
def __init__(self):
super(MyModel, self).__init__()
# Define your layers here.
self.conv1 = Conv2D(32, 3, activation='relu')
self.flatten = Flatten()
self.d1 = Dense(128, activation='relu')
self.d2 = Dense(10, activation='softmax')
def call(self, x):
# Define your forward pass here,
# using layers you previously defined (in __init__).
x = self.conv1(x)
x = self.flatten(x)
x = self.d1(x)
return self.d2(x)
model = MyModel()
#== 選擇 loss, optimizer和 metrics ==#
# * tf.keras.losses
# * tf.keras.optimizers
# * tf.keras.metrics
loss_object = tf.keras.losses.SparseCategoricalCrossentropy()
optimizer = tf.keras.optimizers.Adam()
train_loss = tf.keras.metrics.Mean(name='train_loss')
train_accuracy = tf.keras.metrics.SparseCategoricalAccuracy(name='train_accuracy')
test_loss = tf.keras.metrics.Mean(name='test_loss')
test_accuracy = tf.keras.metrics.SparseCategoricalAccuracy(name='test_accuracy')
#== train the model (定義)==#
# * tf.function
# * tf.GradientTape
@tf.function
def train_step(images, labels):
#
with tf.GradientTape() as tape:
# forward pass to get predictions
predictions = model(images)
# compute loss with the ground and our predictions
loss = loss_object(labels, predictions)
# compute gradient
gradients = tape.gradient(loss, model.trainable_variables)
# back propagation
optimizer.apply_gradients(zip(gradients, model.trainable_variables))
train_loss(loss)
train_accuracy(labels, predictions)
#== test the model (定義)==#
# * tf.function
@tf.function
def test_step(images, labels):
# forward pass to get predictions
predictions = model(images)
# compute loss with the ground and our predictions
t_loss = loss_object(labels, predictions)
test_loss(t_loss)
test_accuracy(labels, predictions)
#== 真正開始 train model ==#
# 定義EPOCH數(整個dataset 拿來train過一次稱一個epoch)
EPOCH = 5
for epoch in range(EPOCH):
for images, labels in train_dataset:
train_step(images, labels)
for test_images, test_labels in test_dataset:
test_step(test_images, test_labels)
template = 'Epoch {}, Loss: {}, Accuracy: {}, Test Loss: {}, Test Accuracy: {}\n'
print(template.format(epoch+1,
train_loss.result(),
train_accuracy.result()*100,
test_loss.result(),
test_accuracy.result()*100))
# Reset the metrics for the next epoch
train_loss.reset_states()
train_accuracy.reset_states()
test_loss.reset_states()
test_accuracy.reset_states()