1.导入依赖包

import tensorflow as tf

from tensorflow import keras

from tensorflow.keras.layers import Flatten, Dense

import numpy as np

import time

2.加载数据集

(train_x, train_y), (test_x, test_y) = keras.datasets.mnist.load_data()

3.图片归一化

train_x, test_x = train_x / 255.0, test_x / 255.0

4.创建模型

class Mnist(tf.keras.Model):

    def __init__(self):

        super(Mnist, self).__init__()

        self.dense1 = Flatten(input_shape=(28, 28))

        self.dense2 = Dense(100, activation='relu')

        self.dense3 = Dense(10, activation='softmax')

    def call(self, inputs):

        x = self.dense1(inputs)

        x = self.dense2(x)

        x = self.dense3(x)

        return x

model = Mnist()

5.指定优化函数、损失函数、评价指标

model.compile(optimizer='adam',

              loss=tf.keras.losses.SparseCategoricalCrossentropy(),

              metrics=['accuracy'])

6.记录模型开始训练时间

start_time = time.time()

7.训练模型

model.fit(train_x, train_y, batch_size=64, epochs=10, verbose=1, validation_freq=1)

Epoch 1/10
938/938 [==============================] - 2s 2ms/step - loss: 0.3147 - accuracy: 0.9126
Epoch 2/10
938/938 [==============================] - 2s 2ms/step - loss: 0.1486 - accuracy: 0.9571
Epoch 3/10
938/938 [==============================] - 2s 2ms/step - loss: 0.1059 - accuracy: 0.9692
Epoch 4/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0796 - accuracy: 0.9772
Epoch 5/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0632 - accuracy: 0.9812
Epoch 6/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0521 - accuracy: 0.9849
Epoch 7/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0422 - accuracy: 0.9872
Epoch 8/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0357 - accuracy: 0.9895
Epoch 9/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0292 - accuracy: 0.9918
Epoch 10/10
938/938 [==============================] - 2s 2ms/step - loss: 0.0254 - accuracy: 0.9929

<tensorflow.python.keras.callbacks.History at 0x1f69dd58400>

8.记录模型训练完成时间

end_time = time.time()

9.用测试集验证模型效果

test_loss, test_acc = model.evaluate(test_x, test_y, verbose=1)

predict = model.predict(test_x)

print('time:', end_time - start_time)

print(test_acc)

print('预测值：', np.argmax(predict[0]))

print('真实值：', test_y[0])

313/313 [==============================] - 1s 2ms/step - loss: 0.0771 - accuracy: 0.9776
time: 20.537350177764893
0.9775999784469604
预测值： 7
真实值： 7