Chapter 13 Save Your Models For Later With Serialization

Question 1

Keras separates the concerns of saving your model architecture and saving your model weights. Model weights are saved to … format. This is a grid format that is ideal for storing multi-dimensional arrays of numbers. The model structure can be described and saved (and loaded) using two different formats: … and…. P 97

Answer 1

HDF5, JSON, YAML

Question 2

Keras provides the ability to describe any model using JSON format with a …function. This can be saved to file and later loaded via the … function that will create a new model from the JSON specification. P 98

Answer 2

to_ json(), model_ from_json()

Question 3

The weights are saved directly from the model using the … function and later loaded using the symmetrical … function. P 98

Answer 3

save_ weights(), load_ weights()

Question 4

What does the below code do? P 98

import numpy
import pandas
from keras.models import Sequential
from keras.layers import Dense
from sklearn.datasets import load_breast_cancer, load_iris,load_diabetes
from sklearn.model_selection import GridSearchCV
from keras.models import model_from_json
seed = 7
numpy.random.seed(seed)
import tensorflow
tensorflow.random.set_seed(seed) #for reproducibility
dataset =  load_breast_cancer(as_frame=True)
X = dataset.data
Y = dataset.target
model = Sequential()
model.add(Dense(12, input_dim=30,activation= "relu" ))
model.add(Dense(8, activation= "relu" ))
model.add(Dense(1, activation= "sigmoid" ))
model.compile(loss= "binary_crossentropy" , optimizer= "adam" , metrics=[ "accuracy" ])
# Fit the model
model.fit(X, Y, epochs=150, batch_size=10, verbose=0)
# evaluate the model
scores = model.evaluate(X, Y, verbose=0)
print("%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))
# serialize model to JSON
model_json = model.to_json()
with open("model.json", "w") as json_file:
  json_file.write(model_json)
# serialize weights to HDF5
model.save_weights("model.h5")
print("Saved model to disk")
# later...
# load json and create model
with open("model.json" , "r" ) as json_file:
  loaded_model_json = json_file.read()
loaded_model = model_from_json(loaded_model_json)
# load weights into new model
loaded_model.load_weights("model.h5")
print("Loaded model from disk")
# evaluate loaded model on test data
loaded_model.compile(loss= "binary_crossentropy" , optimizer= "rmsprop" , metrics=[ "accuracy" ])
score = loaded_model.evaluate(X, Y, verbose=0)
print ("%s: %.2f%%" % (loaded_model.metrics_names[1], score[1]*100))

Answer 4

It saves the nn model and the weights in 2 separate files with json and h5 formats respectively. then loads the saved nn model and use it.

Note the use of .compile after loading an already saved model, before using it to evaluate the dataset

Question 5

What code should we use for having reproducible results in NNs? External

Answer 5

import tensorflow
tensorflow.random.set_seed(seed)