TensorFlow 2.0 识别MNIST手写数字

• 2019 年 12 月 23 日
• 笔记

TensorFlow 2.0 在 1.x版本上进行了大量改进，主要变化如下：

• 以Eager模式为默认的运行模式，不必构建Session
• 删除tf.contrib库，将其中的高阶API整合到tf.kears库下。
• 将1.x版本中大量重复重叠的API进行合并精简

下面是TF2.0 入门demo, 训练集是MNIST。代码略有更改:

"""TF2.0 官方demo，略有更改，加了Dropout"""from __future__ import absolute_import, division, print_function, unicode_literalsimport tensorflow as tffrom tensorflow.keras.layers import Dense, Flatten, Conv2D,Dropoutfrom tensorflow.keras import Model#加载并准备 MNIST 数据集。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# Add a channels dimensionx_train = x_train[..., tf.newaxis]x_test = x_test[..., tf.newaxis]#使用 tf.data 来将数据集切分为 batch 以及混淆数据集：train_ds = tf.data.Dataset.from_tensor_slices(    (x_train, y_train)).shuffle(10000).batch(100) #打乱样本顺序，训练batchsize设为100test_ds = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(32)#使用 Keras 模型子类化（model subclassing） API 构建 tf.keras 模型：  class MyModel(Model):  def __init__(self):    super(MyModel, self).__init__()    self.conv1 = Conv2D(filters=64, kernel_size=(3,3), activation='relu')    #self.dropout1 = Dropout(0.2)    self.flatten = Flatten() # 展平    self.d1 = Dense(256, activation='relu')#全连接层    self.dropout2 = Dropout(0.5)    self.d2 = Dense(10, activation='softmax')      def call(self, x):    x = self.conv1(x)    #x = self.dropout1(x)    x = self.flatten(x)    x = self.d1(x)    x = self.dropout2(x)    return self.d2(x)    model = MyModel()#为训练选择优化器与损失函数：loss_object = tf.keras.losses.SparseCategoricalCrossentropy()optimizer = tf.keras.optimizers.Adam()  #选择衡量指标来度量模型的损失值（loss）和准确率（accuracy）。这些指标在 epoch 上累积值，然后打印出整体结果。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')  #使用 tf.GradientTape 来训练模型：@tf.functiondef train_step(images, labels):  with tf.GradientTape() as tape:    predictions = model(images)    loss = loss_object(labels, predictions)  gradients = tape.gradient(loss, model.trainable_variables)  optimizer.apply_gradients(zip(gradients, model.trainable_variables))  train_loss(loss)  train_accuracy(labels, predictions)   #测试模型：@tf.functiondef test_step(images, labels):  predictions = model(images)  t_loss = loss_object(labels, predictions)  test_loss(t_loss)  test_accuracy(labels, predictions)    EPOCHS = 5for epoch in range(EPOCHS):  for images, labels in train_ds:    train_step(images, labels)  for test_images, test_labels in test_ds:    test_step(test_images, test_labels)  template = 'Epoch {}, Training Loss: {}, Training Accuracy: {}; Test Loss: {}, Test Accuracy: {}'  print (template.format(epoch+1,                         train_loss.result(),                         train_accuracy.result()*100,                         test_loss.result(),                         test_accuracy.result()*100))

5个Epoch后，测试集上的预测准确度达98.43%:

Epoch 1, Training Loss: 0.13766814768314362, Training Accuracy: 95.86332702636719; Test Loss: 0.05851453170180321, Test Accuracy: 98.0999984741211Epoch 2, Training Loss: 0.08793511241674423, Training Accuracy: 97.33499908447266; Test Loss: 0.05357005447149277, Test Accuracy: 98.25999450683594Epoch 3, Training Loss: 0.06439810246229172, Training Accuracy: 98.03555297851562; Test Loss: 0.05198133364319801, Test Accuracy: 98.32333374023438Epoch 4, Training Loss: 0.05080028995871544, Training Accuracy: 98.44750213623047; Test Loss: 0.05018126592040062, Test Accuracy: 98.39500427246094Epoch 5, Training Loss: 0.04206531494855881, Training Accuracy: 98.71266174316406; Test Loss: 0.05160188674926758, Test Accuracy: 98.43399810791016