question2.py

# ======================================================================
# There are 5 questions in this exam with increasing difficulty from 1-5.
# Please note that the weight of the grade for the question is relative
# to its difficulty. So your Category 1 question will score significantly
# less than your Category 5 question.
#
# Don't use lambda layers in your model.
# You do not need them to solve the question.
# Lambda layers are not supported by the grading infrastructure.
#
# You must use the Submit and Test button to submit your model
# at least once in this category before you finally submit your exam,
# otherwise you will score zero for this category.
# ======================================================================
#
# Basic Datasets Question
#
# Create and train a classifier for the MNIST dataset.
# Note that the test will expect it to classify 10 classes and that the 
# input shape should be the native size of the MNIST dataset which is 
# 28x28 monochrome. Do not resize the data. Your input layer should accept
# (28,28) as the input shape only. If you amend this, the tests will fail.
#

import tensorflow as tf
from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint
from tensorflow.keras.layers import Dense, Dropout, Flatten
from tensorflow.keras.models import Sequential

physical_devices = tf.config.list_physical_devices('GPU')
try:
    tf.config.experimental.set_memory_growth(physical_devices[0], True)
except:
    pass


def solution_model():
    mnist = tf.keras.datasets.mnist
    (training_images, training_labels), (test_images, test_labels) = mnist.load_data()

    training_images = training_images / 255.0
    test_images = test_images / 255.0

    callbacks = [
        EarlyStopping(
            monitor='val_accuracy',
            min_delta=1e-4,
            patience=3,
            verbose=1
        ),
        ModelCheckpoint(
            filepath='mymodel.h5',
            monitor='val_accuracy',
            mode='max',
            save_best_only=True,
            save_weights_only=False,
            verbose=1
        )
    ]

    model = Sequential()
    model.add(Flatten())
    model.add(Dense(64, activation='relu'))
    model.add(Dropout(0.2))
    model.add(Dense(10, activation='softmax'))

    model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
    model.fit(
        training_images,
        training_labels,
        batch_size=128,
        epochs=20,
        verbose=1,
        validation_data=(test_images, test_labels),
        callbacks=callbacks
    )

    # YOUR CODE HERE
    return model


# Note that you'll need to save your model as a .h5 like this.
# When you press the Submit and Test button, your saved .h5 model will
# be sent to the testing infrastructure for scoring
# and the score will be returned to you.

if __name__ == '__main__':
    model = solution_model()
    # model.save("mymodel.bak.h5")