#第一步:使用sequential封装网络层
model = keras.Sequential([
layers.Dense(20),activation= 'relu')
layers.Dense(10)])
#第二步:构建梯度记录环境
with tf.GradientTape() as tape:
out = model(输入x)
loss = tf.square(out - 标签)
#计算平均误差
loss = tf.reduce_sum(loss) / x.shape[0]
#第三步:自动计算所有参数梯度
grads = tape.gradient(loss,model.trainable_variables)
optimizer.apply_gradients(zip(grads, model.trainable_variables))