def test_keras_model_train():
model = Mock()
layer = Mock
criterion = Mock()
optimizer = Mock()
metrics = Mock()
sizes = [(1, 24, 24), (24, 10)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, criterion, optimizer, batch, epoch, metrics)
num_data = 30
data = np.array([np.random.rand(24, 24) for i in range(num_data)])
labels = np.array([np.zeros(10) for i in range(num_data)])
for l in labels:
l[np.random.randint(0, len(l))] = 1
kdpm.train(data, labels)
kdpm._model.fit.assert_called_once()
params = kdpm._model.fit.call_args_list[0][1]
assert np.array_equal(params['x'], data)
assert np.array_equal(params['y'], labels)
assert params['batch_size'] == batch
assert params['epochs'] == epoch
def test_get_model_params():
model = Mock()
layer = Mock
criterion = Mock()
optimizer = Mock()
metrics = Mock()
num_data = 30
sizes = [(1, 24, 24), (24, num_data)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, criterion, optimizer, batch, epoch, metrics)
params = np.random.rand(30)
kdpm._model.get_weights.return_value = params
parm = kdpm.get_model_params()
assert np.array_equal(params, parm)
def test_evaluate():
model = Mock()
layer = Mock
criterion = Mock()
optimizer = Mock()
metrics = Mock()
sizes = [(1, 24, 24), (24, 10)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, criterion, optimizer, batch, epoch, metrics)
num_data = 30
data = np.array([np.random.rand(24, 24) for i in range(num_data)])
labels = np.array([np.zeros(10) for i in range(num_data)])
for l in labels:
l[np.random.randint(0, len(l))] = 1
kdpm.evaluate(data, labels)
kdpm._model.evaluate.assert_called_once()
def test_predict():
model = Mock()
layer = Mock
criterion = Mock()
optimizer = Mock()
metrics = Mock()
sizes = [(1, 24, 24), (24, 10)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, criterion, optimizer, batch, epoch, metrics)
num_data = 30
data = np.array([np.random.rand(24, 24) for i in range(num_data)])
kdpm.predict(data)
kdpm._model.predict.assert_called_once_with(data, batch_size=batch)
def test_performance():
model = Mock()
layer = Mock
num_data = 30
sizes = [(1, 24, 24), (24, num_data)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
model.evaluate.return_value = [0, 0]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, batch, epoch)
kdpm._check_data = Mock()
kdpm._check_labels = Mock()
data = np.random.rand(25).reshape((5, 5))
labels = np.random.randint(0, 2, 5)
res = kdpm.performance(data, labels)
kdpm._check_data.assert_called_once_with(data)
kdpm._check_labels.assert_called_once_with(labels)
kdpm._model.evaluate.assert_called_once_with(data, labels, verbose=0)
assert res == 0
def test_set_weights():
model = Mock()
layer = Mock
num_data = 30
sizes = [(1, 24, 24), (24, num_data)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, batch, epoch)
params = np.random.rand(30)
kdpm.set_model_params(params)
kdpm._model.set_weights.assert_called_once_with(params)
def test_wrong_predict():
model = Mock()
layer = Mock
sizes = [(1, 24, 24), (24, 10)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpm = DeepLearningModel(model, batch, epoch)
num_data = 30
data = np.array([np.random.rand(16, 16) for i in range(num_data)])
with pytest.raises(AssertionError):
kdpm.predict(data)
def test_train_wrong_data():
model = Mock()
layer = Mock
criterion = Mock()
optimizer = Mock()
metrics = Mock()
sizes = [(30, 24, 24), (24, 10)]
l1 = layer()
l1.get_input_shape_at.return_value = sizes[0]
l2 = layer()
l2.get_output_shape_at.return_value = sizes[1]
model.layers = [l1, l2]
batch = 32
epoch = 2
kdpl = DeepLearningModel(model, criterion, optimizer, batch, epoch, metrics)
num_data = 30
data = np.array([np.random.rand(16, 16) for i in range(num_data)])
label = np.array([np.zeros(10) for i in range(num_data)])
for l in label:
l[np.random.randint(0, len(l))] = 1
with pytest.raises(AssertionError):
kdpl.train(data, label)
data = np.array([np.random.rand(24, 24) for i in range(num_data)])
label = np.array([np.zeros(8) for i in range(num_data)])
for l in label:
l[np.random.randint(0, len(l))] = 1
with pytest.raises(AssertionError):
kdpl.train(data, label)
def model_builder():
"""
Create a Tensorflow Model for image classification.
# Returns:
model: Instance of DeepLearningModel [DeepLearningModel](../model/#deeplearningmodel-class)).
"""
model = tf.keras.models.Sequential()
model.add(
tf.keras.layers.Conv2D(32,
kernel_size=(3, 3),
padding='same',
activation='relu',
strides=1,
input_shape=(28, 28, 1)))
model.add(
tf.keras.layers.MaxPooling2D(pool_size=2,
strides=2,
padding='valid'))
model.add(tf.keras.layers.Dropout(0.4))
model.add(
tf.keras.layers.Conv2D(32,
kernel_size=(3, 3),
padding='same',
activation='relu',
strides=1))
model.add(
tf.keras.layers.MaxPooling2D(pool_size=2,
strides=2,
padding='valid'))
model.add(tf.keras.layers.Dropout(0.3))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(128, activation='relu'))
model.add(tf.keras.layers.Dropout(0.1))
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
criterion = tf.keras.losses.CategoricalCrossentropy()
optimizer = tf.keras.optimizers.RMSprop()
metrics = [tf.keras.metrics.CategoricalAccuracy()]
return DeepLearningModel(model=model,
criterion=criterion,
optimizer=optimizer,
metrics=metrics)
def model_builder():
"""
Create a Tensorflow Model for image classification.
# Returns:
model: Instance of DeepLearningModel [DeepLearningModel](../model/#deeplearningmodel-class)).
"""
model = tf.keras.models.Sequential()
model.add(
tf.keras.layers.Conv2D(32,
kernel_size=(3, 3),
padding='same',
activation='relu',
strides=1,
input_shape=(28, 28, 1)))
model.add(
tf.keras.layers.MaxPooling2D(pool_size=2,
strides=2,
padding='valid'))
model.add(tf.keras.layers.Dropout(0.4))
model.add(
tf.keras.layers.Conv2D(32,
kernel_size=(3, 3),
padding='same',
activation='relu',
strides=1))
model.add(
tf.keras.layers.MaxPooling2D(pool_size=2,
strides=2,
padding='valid'))
model.add(tf.keras.layers.Dropout(0.3))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(128, activation='relu'))
model.add(tf.keras.layers.Dropout(0.1))
model.add(tf.keras.layers.Dense(64, activation='relu'))
model.add(tf.keras.layers.Dense(10, activation='softmax'))
model.compile(optimizer="rmsprop",
loss="categorical_crossentropy",
metrics=["accuracy"])
return DeepLearningModel(model)