using_model_checkpoints

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from keras import regularizers, optimizers
from keras.layers import Dense, Dropout, Input
from keras.models import Sequential, load_model
from keras.callbacks import ModelCheckpoint
from keras.datasets import boston_housing
from keras.losses import MeanSquaredError

#Load Boston Housing Dataset
(X_train, y_train), (X_test, y_test) = boston_housing.load_data()

#Build model
model = Sequential()
model.add(Input(shape=(13,)))
model.add(Dense(32, activation='tanh', 
                kernel_regularizer = regularizers.l1_l2(l1=1e-5, l2=1e-4)))
model.add(Dense(1,activation='relu'))

mse = MeanSquaredError()
adam = optimizers.Adam(learning_rate=.1, decay=1e-3)

model.compile(optimizer = adam, loss = mse, metrics = None)
print(model.summary())

#create callback
filepath = 'my_best_model.hdf5'
checkpoint = ModelCheckpoint(filepath=filepath, 
                             monitor='val_loss',
                             verbose=1, 
                             save_best_only=True,
                             mode='min')
callbacks = [checkpoint]

#fit the model
history = model.fit(pd.DataFrame(X_train).apply(np.asarray),
                  y_train,
                  batch_size=10,
                  epochs=100,
                  validation_split=0.2,
                  callbacks=callbacks)

#plot the training history
plt.plot(history.history['loss'], label='Training Loss')
plt.plot(history.history['val_loss'], label='Validation Loss')
plt.legend()
plt.xlabel('Epochs')
plt.ylabel('Mean Squared Error')
plt.savefig('model_training_history')
plt.show()

#Load and evaluate the best model version
model = load_model(filepath)
yhat = model.predict(X_test)
print('Model MSE on test data = ', mse(y_test, yhat).numpy())