def build_model():
input_layer = ak.Input()
cnn_layer = ak.ConvBlock()(input_layer)
cnn_layer2 = ak.ConvBlock()(cnn_layer)
dense_layer = ak.DenseBlock()(cnn_layer2)
dense_layer2 = ak.DenseBlock()(dense_layer)
output_layer = ak.ClassificationHead(num_classes=10)(dense_layer2)
automodel = ak.auto_model.AutoModel(input_layer, output_layer, max_trials=20, seed=123, project_name='autoML')
return automodel
def test_graph_save_load(tmp_path):
input1 = ak.Input()
input2 = ak.Input()
output1 = ak.DenseBlock()(input1)
output2 = ak.ConvBlock()(input2)
output = ak.Merge()([output1, output2])
output1 = ak.RegressionHead()(output)
output2 = ak.ClassificationHead()(output)
graph = graph_module.Graph(
inputs=[input1, input2],
outputs=[output1, output2],
override_hps=[
hp_module.Choice("dense_block_1/num_layers", [6], default=6)
],
)
path = os.path.join(tmp_path, "graph")
graph.save(path)
graph = graph_module.load_graph(path)
assert len(graph.inputs) == 2
assert len(graph.outputs) == 2
assert isinstance(graph.inputs[0].out_blocks[0], ak.DenseBlock)
assert isinstance(graph.inputs[1].out_blocks[0], ak.ConvBlock)
assert isinstance(graph.override_hps[0], hp_module.Choice)
def test_graph_can_init_with_one_missing_output():
input_node = ak.ImageInput()
output_node = ak.ConvBlock()(input_node)
output_node = ak.RegressionHead()(output_node)
ak.ClassificationHead()(output_node)
graph_module.Graph(input_node, output_node)
def test_graph_save_load(tmp_dir):
input1 = ak.Input()
input2 = ak.Input()
output1 = ak.DenseBlock()(input1)
output2 = ak.ConvBlock()(input2)
output = ak.Merge()([output1, output2])
output1 = ak.RegressionHead()(output)
output2 = ak.ClassificationHead()(output)
graph = graph_module.HyperGraph(inputs=[input1, input2],
outputs=[output1, output2],
override_hps=[
hp_module.Choice(
'dense_block_1/num_layers', [6],
default=6)
])
path = os.path.join(tmp_dir, 'graph')
graph.save(path)
config = graph.get_config()
graph = graph_module.HyperGraph.from_config(config)
graph.reload(path)
assert len(graph.inputs) == 2
assert len(graph.outputs) == 2
assert isinstance(graph.inputs[0].out_blocks[0], ak.DenseBlock)
assert isinstance(graph.inputs[1].out_blocks[0], ak.ConvBlock)
assert isinstance(graph.override_hps[0], hp_module.Choice)
def test_graph_compile_with_adadelta():
input_node = ak.ImageInput(shape=(32, 32, 3))
output_node = ak.ConvBlock()(input_node)
output_node = ak.RegressionHead(output_shape=(1, ))(output_node)
graph = graph_module.Graph(input_node, output_node)
hp = kerastuner.HyperParameters()
hp.values = {"optimizer": "adadelta"}
graph.build(hp)
def create_image_regressor(self):
input_node = ak.ImageInput()
output_node = ak.ConvBlock()(input_node)
output_node = ak.DenseBlock()(output_node)
output_node = ak.RegressionHead()(output_node)
reg = ak.AutoModel(inputs=input_node,
outputs=output_node,
max_trials=10)
return reg
def test_functional_api(tmp_dir):
# Prepare the data.
num_instances = 20
(image_x, train_y), (test_x, test_y) = mnist.load_data()
(text_x, train_y), (test_x, test_y) = common.imdb_raw()
(structured_data_x, train_y), (test_x, test_y) = common.dataframe_numpy()
image_x = image_x[:num_instances]
text_x = text_x[:num_instances]
structured_data_x = structured_data_x[:num_instances]
classification_y = common.generate_one_hot_labels(
num_instances=num_instances, num_classes=3)
regression_y = common.generate_data(num_instances=num_instances,
shape=(1, ))
# Build model and train.
image_input = ak.ImageInput()
output = ak.Normalization()(image_input)
output = ak.ImageAugmentation()(output)
outputs1 = ak.ResNetBlock(version='next')(image_input)
outputs2 = ak.XceptionBlock()(image_input)
image_output = ak.Merge()((outputs1, outputs2))
structured_data_input = ak.StructuredDataInput(
column_names=common.COLUMN_NAMES_FROM_CSV,
column_types=common.COLUMN_TYPES_FROM_CSV)
structured_data_output = ak.FeatureEngineering()(structured_data_input)
structured_data_output = ak.DenseBlock()(structured_data_output)
text_input = ak.TextInput()
outputs1 = ak.TextToIntSequence()(text_input)
outputs1 = ak.EmbeddingBlock()(outputs1)
outputs1 = ak.ConvBlock(separable=True)(outputs1)
outputs1 = ak.SpatialReduction()(outputs1)
outputs2 = ak.TextToNgramVector()(text_input)
outputs2 = ak.DenseBlock()(outputs2)
text_output = ak.Merge()((outputs1, outputs2))
merged_outputs = ak.Merge()(
(structured_data_output, image_output, text_output))
regression_outputs = ak.RegressionHead()(merged_outputs)
classification_outputs = ak.ClassificationHead()(merged_outputs)
automodel = ak.GraphAutoModel(
inputs=[image_input, text_input, structured_data_input],
directory=tmp_dir,
outputs=[regression_outputs, classification_outputs],
max_trials=2,
seed=common.SEED)
automodel.fit((image_x, text_x, structured_data_x),
(regression_y, classification_y),
validation_split=0.2,
epochs=2)
def functional_api():
(x_train, y_train), (x_test, y_test) = mnist.load_data()
input_node = ak.ImageInput()
output_node = input_node
output_node = ak.Normalization()(output_node)
output_node = ak.ConvBlock()(output_node)
output_node = ak.SpatialReduction()(output_node)
output_node = ak.DenseBlock()(output_node)
output_node = ak.ClassificationHead()(output_node)
clf = ak.GraphAutoModel(input_node, output_node, seed=5, max_trials=3)
clf.fit(x_train, y_train, validation_split=0.2)
return clf.evaluate(x_test, y_test)
def test_functional_api(tmp_path):
# Prepare the data.
num_instances = 80
(image_x, train_y), (test_x, test_y) = mnist.load_data()
(text_x, train_y), (test_x, test_y) = utils.imdb_raw()
(structured_data_x, train_y), (test_x, test_y) = utils.dataframe_numpy()
image_x = image_x[:num_instances]
text_x = text_x[:num_instances]
structured_data_x = structured_data_x[:num_instances]
classification_y = utils.generate_one_hot_labels(
num_instances=num_instances, num_classes=3)
regression_y = utils.generate_data(num_instances=num_instances,
shape=(1, ))
# Build model and train.
image_input = ak.ImageInput()
output = ak.Normalization()(image_input)
output = ak.ImageAugmentation()(output)
outputs1 = ak.ResNetBlock(version='next')(output)
outputs2 = ak.XceptionBlock()(output)
image_output = ak.Merge()((outputs1, outputs2))
structured_data_input = ak.StructuredDataInput()
structured_data_output = ak.CategoricalToNumerical()(structured_data_input)
structured_data_output = ak.DenseBlock()(structured_data_output)
text_input = ak.TextInput()
outputs1 = ak.TextToIntSequence()(text_input)
outputs1 = ak.Embedding()(outputs1)
outputs1 = ak.ConvBlock(separable=True)(outputs1)
outputs1 = ak.SpatialReduction()(outputs1)
outputs2 = ak.TextToNgramVector()(text_input)
outputs2 = ak.DenseBlock()(outputs2)
text_output = ak.Merge()((outputs1, outputs2))
merged_outputs = ak.Merge()(
(structured_data_output, image_output, text_output))
regression_outputs = ak.RegressionHead()(merged_outputs)
classification_outputs = ak.ClassificationHead()(merged_outputs)
automodel = ak.AutoModel(
inputs=[image_input, text_input, structured_data_input],
directory=tmp_path,
outputs=[regression_outputs, classification_outputs],
max_trials=2,
tuner=ak.Hyperband,
seed=utils.SEED)
automodel.fit((image_x, text_x, structured_data_x),
(regression_y, classification_y),
validation_split=0.2,
epochs=1)
def CreateSupergraph(output_dir, hp_tuner):
input_node = ak.Input()
conv2d_1 = ak.ConvBlock(num_blocks=1, num_layers=3,
max_pooling=True, dropout=0)(input_node)
dense_1 = ak.DenseBlock(dropout=0)(conv2d_1)
output_node = ak.ClassificationHead(num_classes=4,
metrics=['accuracy'])(dense_1)
automodel = ak.AutoModel(inputs=input_node, outputs=output_node,
max_trials=3, directory=output_dir, project_name="autoML",
tuner=hp_tuner, seed=123)
return automodel
def test_text_and_structured_data(tmp_path):
# Prepare the data.
num_instances = 80
(x_text, y_train), (x_test, y_test) = utils.imdb_raw()
x_structured_data = pd.read_csv(utils.TRAIN_CSV_PATH)
x_text = x_text[:num_instances]
x_structured_data = x_structured_data[:num_instances]
y_classification = utils.generate_one_hot_labels(
num_instances=num_instances, num_classes=3)
y_regression = utils.generate_data(num_instances=num_instances,
shape=(1, ))
# Build model and train.
structured_data_input = ak.StructuredDataInput()
structured_data_output = ak.CategoricalToNumerical()(structured_data_input)
structured_data_output = ak.DenseBlock()(structured_data_output)
text_input = ak.TextInput()
outputs1 = ak.TextToIntSequence()(text_input)
outputs1 = ak.Embedding()(outputs1)
outputs1 = ak.ConvBlock(separable=True)(outputs1)
outputs1 = ak.SpatialReduction()(outputs1)
outputs2 = ak.TextToNgramVector()(text_input)
outputs2 = ak.DenseBlock()(outputs2)
text_output = ak.Merge()((outputs1, outputs2))
merged_outputs = ak.Merge()((structured_data_output, text_output))
regression_outputs = ak.RegressionHead()(merged_outputs)
classification_outputs = ak.ClassificationHead()(merged_outputs)
automodel = ak.AutoModel(
inputs=[text_input, structured_data_input],
directory=tmp_path,
outputs=[regression_outputs, classification_outputs],
max_trials=2,
tuner=ak.Hyperband,
seed=utils.SEED,
)
automodel.fit(
(x_text, x_structured_data),
(y_regression, y_classification),
validation_split=0.2,
epochs=1,
)
def test_conv_block(tmp_dir):
x_train = np.random.rand(100, 3, 3, 3)
y_train = np.random.randint(10, size=100)
y_train = tf.keras.utils.to_categorical(y_train)
input_node = ak.Input()
output_node = input_node
output_node = ak.ConvBlock()(output_node)
output_node = ak.ClassificationHead()(output_node)
input_node.shape = (3, 3, 3)
output_node[0].shape = (10,)
graph = ak.GraphAutoModel(input_node, output_node, directory=tmp_dir)
model = graph.build(kerastuner.HyperParameters())
model.fit(x_train, y_train, epochs=1, batch_size=100, verbose=False)
result = model.predict(x_train)
assert result.shape == (100, 10)
def test_minist(self):
(x_train, y_train), (x_test, y_test) = mnist.load_data()
print(x_train.shape) # (60000, 28, 28)
print(y_train.shape) # (60000,)
print(y_train[:3]) # array([7, 2, 1], dtype=uint8)
input_node = ak.ImageInput()
output_node = ak.ConvBlock()(input_node)
output_node = ak.DenseBlock()(output_node)
output_node = ak.ClassificationHead()(output_node)
clf = ak.AutoModel(inputs=input_node,
outputs=output_node,
max_trials=10)
clf.fit(x_train, y_train)
model = clf.export_model()
print(type(model))
model.save('./auto_model.hd5')
def functional_api():
max_features = 20000
max_words = 400
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.imdb.load_data(
num_words=max_features, index_from=3)
x_train = tf.keras.preprocessing.sequence.pad_sequences(x_train,
maxlen=max_words)
x_test = tf.keras.preprocessing.sequence.pad_sequences(x_test,
maxlen=max_words)
print(x_train.dtype)
print(x_train[:10])
input_node = ak.Input()
output_node = input_node
output_node = ak.EmbeddingBlock(max_features=max_features)(output_node)
output_node = ak.ConvBlock()(output_node)
output_node = ak.SpatialReduction()(output_node)
output_node = ak.DenseBlock()(output_node)
output_node = ak.ClassificationHead()(output_node)
clf = ak.AutoModel(input_node, output_node, seed=5, max_trials=3)
clf.fit(x_train, y_train, validation_split=0.2)
return clf.evaluate(x_test, y_test)
def test_graph_save_load(tmp_path):
input1 = ak.Input()
input2 = ak.Input()
output1 = ak.DenseBlock()(input1)
output2 = ak.ConvBlock()(input2)
output = ak.Merge()([output1, output2])
output1 = ak.RegressionHead()(output)
output2 = ak.ClassificationHead()(output)
graph = graph_module.Graph(
inputs=[input1, input2],
outputs=[output1, output2],
)
path = os.path.join(tmp_path, "graph")
graph.save(path)
graph = graph_module.load_graph(path)
assert len(graph.inputs) == 2
assert len(graph.outputs) == 2
assert isinstance(graph.inputs[0].out_blocks[0], ak.DenseBlock)
assert isinstance(graph.inputs[1].out_blocks[0], ak.ConvBlock)
def test_conv_block(tmp_dir):
x_train = np.random.rand(100, 3, 3, 3)
y_train = np.random.randint(10, size=100)
y_train = tf.keras.utils.to_categorical(y_train)
input_node = ak.Input()
output_node = input_node
output_node = ak.ConvBlock()(output_node)
output_node = ak.ClassificationHead()(output_node)
input_node.shape = (3, 3, 3)
output_node[0].shape = (10,)
graph = ak.GraphAutoModel(input_node, output_node,
directory=tmp_dir,
max_trials=1)
graph.fit(x_train, y_train,
epochs=1,
batch_size=100,
verbose=False,
validation_split=0.2)
result = graph.predict(x_train)
assert result.shape == (100, 10)
id5 --> id6
id7(StructuredDataInput) --> id8(CategoricalToNumerical)
id8 --> id9(DenseBlock)
id6 --> id10(Merge)
id9 --> id10
id10 --> id11(Classification Head)
id10 --> id12(Regression Head)
</div>
"""
import autokeras as ak
input_node1 = ak.ImageInput()
output_node = ak.Normalization()(input_node1)
output_node = ak.ImageAugmentation()(output_node)
output_node1 = ak.ConvBlock()(output_node)
output_node2 = ak.ResNetBlock(version='v2')(output_node)
output_node1 = ak.Merge()([output_node1, output_node2])
input_node2 = ak.StructuredDataInput()
output_node = ak.CategoricalToNumerical()(input_node2)
output_node2 = ak.DenseBlock()(output_node)
output_node = ak.Merge()([output_node1, output_node2])
output_node1 = ak.ClassificationHead()(output_node)
output_node2 = ak.RegressionHead()(output_node)
auto_model = ak.AutoModel(inputs=[input_node1, input_node2],
outputs=[output_node1, output_node2],
overwrite=True,
max_trials=2)
[functional API](https://www.tensorflow.org/guide/keras/functional) of Keras. Basically, you are building a graph, whose edges are blocks and the nodes are intermediate outputs of blocks. To add an edge from `input_node` to `output_node` with `output_node = ak.[some_block]([block_args])(input_node)`. You can even also use more fine grained blocks to customize the search space even further. See the following example. """ import autokeras as ak input_node = ak.TextInput() output_node = ak.TextToIntSequence()(input_node) output_node = ak.Embedding()(output_node) # Use separable Conv layers in Keras. output_node = ak.ConvBlock(separable=True)(output_node) output_node = ak.ClassificationHead()(output_node) clf = ak.AutoModel(inputs=input_node, outputs=output_node, max_trials=1) clf.fit(x_train, y_train, epochs=2) """ ## Data Format The AutoKeras TextClassifier is quite flexible for the data format. For the text, the input data should be one-dimensional For the classification labels, AutoKeras accepts both plain labels, i.e. strings or integers, and one-hot encoded encoded labels, i.e. vectors of 0s and 1s. We also support using [tf.data.Dataset]( https://www.tensorflow.org/api_docs/python/tf/data/Dataset?version=stable) format for the training data. The labels have to be one-hot encoded for multi-class
y_train = np.zeros(num_instances).astype(np.float32)
y_train[0 : int(num_instances / 2)] = 1
x_test = np.random.rand(num_instances, num_features).astype(np.float32)
y_test = np.zeros(num_instances).astype(np.float32)
y_train[0 : int(num_instances / 2)] = 1
x_train = np.expand_dims(
x_train, axis=2
) # This step it's very important an CNN will only accept this data shape
print(x_train.shape)
print(y_train.shape)
# Prepare Automodel for search
input_node = ak.Input()
output_node = ak.ConvBlock()(input_node)
# output_node = ak.DenseBlock()(output_node) #optional
# output_node = ak.SpatialReduction()(output_node) #optional
output_node = ak.ClassificationHead(num_classes=2, multi_label=True)(output_node)
auto_model = ak.AutoModel(
inputs=input_node, outputs=output_node, overwrite=True, max_trials=1
)
# Search
auto_model.fit(x_train, y_train, epochs=1)
print(auto_model.evaluate(x_test, y_test))
# Export as a Keras Model