TensorFlow2代码解读(5)

import tensorflow as tf
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics

(x, y), (x_test, y_test) = datasets.fashion_mnist.load_data()
x = tf.convert_to_tensor(x, dtype=tf.float32) / 255.
y = tf.convert_to_tensor(y, dtype=tf.int32)
x_test = tf.convert_to_tensor(x_test, dtype=tf.float32) / 255.
y_test = tf.convert_to_tensor(y_test, dtype=tf.int32)

# 将28x28的图像转换成长度为784的向量
x = tf.reshape(x, [-1, 28*28])
x_test = tf.reshape(x_test, [-1, 28*28])

batch_size = 128

db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.shuffle(10000).batch(batch_size)

db_test = tf.data.Dataset.from_tensor_slices((x_test, y_test))
db_test = db_test.batch(batch_size)

optimizer = optimizers.Adam(lr=1e-3)

class MyDense(layers.Layer):
    def __init__(self,inp_dim,outp_dim):
        super(MyDense, self).__init__()
        self.kernel = self.add_variable('w',[inp_dim,outp_dim])
        self.bias = self.add_variable('b',[outp_dim])
        
    def call(self,inputs,training = None):
        out = inputs @ self.kernel + self.bias
        return out
    
class MyModel(tf.keras.Model):
    def __init__(self):
        super(MyModel,self).__init__()
        self.fc1 = MyDense(28*28,256)
        self.fc2 = MyDense(256,128)
        self.fc3 = MyDense(128,64)
        self.fc4 = MyDense(64,32)
        self.fc5 = MyDense(32,10)
        
    def call(self,inputs,training = None):
        x = self.fc1(inputs)
        x = tf.nn.relu(x)
        x = self.fc2(x)
        x = tf.nn.relu(x)
        x = self.fc3(x)
        x = tf.nn.relu(x)
        x = self.fc4(x)
        x = tf.nn.relu(x)
        x = self.fc5(x)
        
        return x
    
model = MyModel()
model.compile(optimizer=optimizer,
              loss=tf.losses.MSE,
              metrics=['accuracy'])

def main():
    model.fit(db, epochs=20, validation_data=db_test,validation_freq=2)

if __name__ == '__main__':
    main()

import tensorflow as tf
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics

导入 TensorFlow 和相关模块。

(x, y), (x_test, y_test) = datasets.fashion_mnist.load_data()

加载 Fashion MNIST 数据集并划分为训练集和测试集。

x = tf.convert_to_tensor(x, dtype=tf.float32) / 255.
y = tf.convert_to_tensor(y, dtype=tf.int32)
x_test = tf.convert_to_tensor(x_test, dtype=tf.float32) / 255.
y_test = tf.convert_to_tensor(y_test, dtype=tf.int32)

将输入特征和标签转换为 TensorFlow 张量，并归一化特征值。

x = tf.reshape(x, [-1, 28*28])
x_test = tf.reshape(x_test, [-1, 28*28])

将每个样本的图像数据从 28x28 的二维数组形式转换为长度为 784 的一维向量。

batch_size = 128

设置批处理大小。

db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.shuffle(10000).batch(batch_size)

db_test = tf.data.Dataset.from_tensor_slices((x_test, y_test))
db_test = db_test.batch(batch_size)

使用 TensorFlow 的 Dataset API 创建训练集和测试集的数据集对象，并进行洗牌和批处理。

optimizer = optimizers.Adam(lr=1e-3)

创建 Adam 优化器对象，用于训练模型。

class MyDense(layers.Layer):
    def __init__(self, inp_dim, outp_dim):
        super(MyDense, self).__init__()
        self.kernel = self.add_variable('w', [inp_dim, outp_dim])
        self.bias = self.add_variable('b', [outp_dim])
        
    def call(self, inputs, training=None):
        out = inputs @ self.kernel + self.bias
        return out

定义自定义的全连接层类 MyDense，继承于 TensorFlow 的 layers.Layer。其中包含了权重和偏置张量，并实现了前向传播的逻辑。

class MyModel(tf.keras.Model):
    def __init__(self):
        super(MyModel, self).__init__()
        self.fc1 = MyDense(28*28, 256)
        self.fc2 = MyDense(256, 128)
        self.fc3 = MyDense(128, 64)
        self.fc4 = MyDense(64, 32)
        self.fc5 = MyDense(32, 10)
        
    def call(self, inputs, training=None):
        x = self.fc1(inputs)
        x = tf.nn.relu(x)
        x = self.fc2(x)
        x = tf.nn.relu(x)
        x = self.fc3(x)
        x = tf.nn.relu(x)
        x = self.fc4(x)
        x = tf.nn.relu(x)
        x = self.fc5(x)
        
        return x

定义自定义的模型类 MyModel，继承于 TensorFlow 的 tf.keras.Model。其中包含了 5 个全连接层，并定义了模型的前向传播过程。

model = MyModel()

创建模型对象。

model.compile(optimizer=optimizer, loss=tf.losses.MSE, metrics=['accuracy'])

配置模型的优化器、损失函数和评估指标。

def main():
    model.fit(db, epochs=20, validation_data=db_test, validation_freq=2)

if __name__ == '__main__':
    main()


定义主函数 main，在该函数中使用训练集进行训练，并在每两个 epoch 后使用测试集进行验证。最后通过调用 main() 来运行整个程序。