import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import tensorflow as tf
from tensorflow import keras
from sklearn.preprocessing import StandardScaler
from sklearn.datasets import fetch_california_housing
import pprint
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler

print(sys.version_info)
for module in mpl,np,pd,sklearn,tf,keras:
    print(module.__name__,module.__version__)

housing = fetch_california_housing()
print(housing.DESCR)
print(housing.data)
print(housing.data.shape)
print(housing.target)
print(housing.target.shape)
pprint.pprint(housing.data)
pprint.pprint(housing.target)
print(housing.data[10])

x_train_all,x_test,y_train_all,y_test = train_test_split(housing.data,housing.target,random_state=7,test_size=0.25)
x_train,x_valid,y_train,y_valid = train_test_split(x_train_all,y_train_all,random_state=11,test_size=0.25)

print(x_train.shape,y_train.shape)
print(x_valid.shape,y_valid.shape)
print(x_test.shape,y_test.shape)

scaler = StandardScaler()
x_train_scaled = scaler.fit_transform(x_train)
x_valid_scaled = scaler.transform(x_valid)
x_test_scaled = scaler.transform(x_test)

def customized_mse(y_true,y_pre):
    return tf.reduce_mean(tf.square(y_pre - y_true))

# customized dense layer
class CustomizedDenseLayer(keras.layers.Layer):
    def __init__(self,units,activation=None,**kwargs):
        self.units = units
        self.activation = keras.layers.Activation(activation)
        super(CustomizedDenseLayer,self).__init__(**kwargs)

    def build(self, input_shape):
        '''构建所需要的参数'''
        # x * w + b input_shape:[None,a] w:[a,b] output_shape:[None,b] #add_weight获得一个变量
        self.kernel = self.add_weight(name='kernel',
                                      shape=(input_shape[1],self.units),
                                      initializer='uniform', #定义如何初始化参数矩阵,uniform:均匀分布的方法
                                      trainable=True)
        self.bias = self.add_weight(name='bias',
                                    shape=(self.units),
                                    initializer='zeros',
                                    trainable=True)
        super(CustomizedDenseLayer,self).build(input_shape)

    def call(self, x):
        '''完成正向计算'''
        return self.activation(tf.matmul(x,self.kernel) + self.bias)

model = keras.models.Sequential([
    CustomizedDenseLayer(30,activation='relu',
                       input_shape=x_train.shape[1:]),
    CustomizedDenseLayer(1)
])

model.summary()
sgd = keras.optimizers.SGD(lr=0.001)
model.compile(loss=customized_mse,optimizer=sgd,metrics=['mean_squared_error'])
callbacks = [keras.callbacks.EarlyStopping(patience=5,min_delta=1e-2,monitor='val_loss')]

history = model.fit(x_train_scaled,y_train,
                    validation_data=(x_valid_scaled,y_valid),
                    epochs=100,
                    callbacks=callbacks)

model.evaluate(x_test_scaled,y_test)

def plot_learning_curves(history):
    pd.DataFrame(history.history).plot(figsize=(8,5))
    plt.grid(True)
    plt.gca().set_ylim(0,1)
    plt.show()

plot_learning_curves(history)

老师，请问为什么会报错：TypeError: Failed to convert object of type <class 'tuple'> to Tensor. Contents: (Dimension(8), 30). Consider casting elements to a supported type.