TensorFlow进阶学习定制模型和训练算法

class CustomDense(tf.keras.layers.Layer):
    def __init__(self, units=10):
        super(CustomDense, self).__init__()
        self.units = units

    def build(self, input_shape):
        self.w = self.add_weight(shape=(input_shape[-1], self.units),
                                 initializer='random_normal',
                                 trainable=True)
        self.b = self.add_weight(shape=(self.units,),
                                 initializer='zeros',
                                 trainable=True)

    def call(self, inputs):
        return tf.matmul(inputs, self.w) + self.b

# 使用 CustomDense 层创建模型
model = tf.keras.Sequential([
    CustomDense(10),
    tf.keras.layers.Activation('relu'),
    tf.keras.layers.Dense(1)
])

class CustomModel(tf.keras.Model):
    def train_step(self, data):
        # 拆分数据
        x, y = data
        with tf.GradientTape() as tape:
            y_pred = self(x, training=True)  # 正向传播
            loss = self.compiled_loss(y, y_pred, regularization_losses=self.losses)
        # 计算梯度
        trainable_vars = self.trainable_variables
        gradients = tape.gradient(loss, trainable_vars)
        # 更新权重
        self.optimizer.apply_gradients(zip(gradients, trainable_vars))
        # 更新度量
        self.compiled_metrics.update_state(y, y_pred)
        return {m.name: m.result() for m in self.metrics}

model = CustomModel([
    CustomDense(10),
    tf.keras.layers.Activation('relu'),
    tf.keras.layers.Dense(1)
])
model.compile(optimizer='adam',
              loss='binary_crossentropy',
              metrics=['accuracy'])
history = model.fit(train_data, train_labels, epochs=10)