def test_image_block_augment_return_tensor():
block = blocks.ImageBlock(augment=True)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_structured_block_normalize_return_tensor():
block = blocks.StructuredDataBlock(normalize=True)
block.column_names = ["0", "1"]
block.column_types = {"0": analysers.NUMERICAL, "1": analysers.NUMERICAL}
outputs = block.build(keras_tuner.HyperParameters(),
tf.keras.Input(shape=(2, ), dtype=tf.string))
assert len(nest.flatten(outputs)) == 1
def test_spatial_build_return_tensor():
block = blocks.SpatialReduction()
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_dense_build_return_tensor():
block = blocks.DenseBlock(
num_units=keras_tuner.engine.hyperparameters.Choice(
"num_units", [10, 20]))
outputs = block.build(keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, ), dtype=tf.float32))
assert len(nest.flatten(outputs)) == 1
def test_resnet_pretrained_build_return_tensor():
block = blocks.ResNetBlock(pretrained=True)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_resnet_pretrained_with_one_channel_input():
block = blocks.ResNetBlock(pretrained=True)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(28, 28, 1), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_resnet_v1_return_tensor():
block = blocks.ResNetBlock(version="v1")
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_efficientnet_b0_return_tensor():
block = blocks.EfficientNetBlock(version="b0", pretrained=False)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_rnn_build_return_tensor():
block = blocks.RNNBlock(bidirectional=False)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 10), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_conv_build_with_dropout_return_tensor():
block = blocks.ConvBlock(dropout=0.5)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_conv_with_small_image_size_return_tensor():
block = blocks.ConvBlock()
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(10, 10, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_augment_build_with_vflip_only_return_tensor():
block = blocks.ImageAugmentation(vertical_flip=True, horizontal_flip=False)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_augment_build_with_zoom_factor_return_tensor():
block = blocks.ImageAugmentation(zoom_factor=0.1)
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_cat_to_num_build_return_tensor():
block = blocks.CategoricalToNumerical()
block.column_names = ["a"]
block.column_types = {"a": "num"}
outputs = block.build(keras_tuner.HyperParameters(),
tf.keras.Input(shape=(1, ), dtype=tf.string))
assert len(nest.flatten(outputs)) == 1
def test_temporal_global_avg_return_tensor():
block = blocks.TemporalReduction(reduction_type="global_avg")
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 10), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_xception_build_return_tensor():
block = blocks.XceptionBlock()
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 32, 2), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_clf_head_build_with_zero_dropout_return_tensor():
block = head_module.ClassificationHead(dropout=0, shape=(8, ))
outputs = block.build(
keras_tuner.HyperParameters(),
keras.Input(shape=(5, ), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_merge_single_input_return_tensor():
block = blocks.Merge()
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32,), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_image_block_xception_return_tensor():
block = blocks.ImageBlock(block_type="xception")
outputs = block.build(
keras_tuner.HyperParameters(),
keras.Input(shape=(32, 32, 3), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_multi_label_loss():
head = head_module.ClassificationHead(name="a",
multi_label=True,
num_classes=8,
shape=(8, ))
input_node = keras.Input(shape=(5, ))
output_node = head.build(keras_tuner.HyperParameters(), input_node)
model = keras.Model(input_node, output_node)
assert model.layers[-1].activation.__name__ == "sigmoid"
assert head.loss.name == "binary_crossentropy"
def test_txt_clf_init_hp0_equals_hp_of_a_model(tmp_path):
clf = ak.TextClassifier(directory=tmp_path)
clf.inputs[0].shape = (1, )
clf.outputs[0].in_blocks[0].shape = (10, )
init_hp = task_specific.TEXT_CLASSIFIER[0]
hp = keras_tuner.HyperParameters()
hp.values = copy.copy(init_hp)
clf.tuner.hypermodel.build(hp)
assert set(init_hp.keys()) == set(hp._hps.keys())
def test_graph_basics():
input_node = ak.Input(shape=(30, ))
output_node = input_node
output_node = ak.DenseBlock()(output_node)
output_node = ak.RegressionHead(shape=(1, ))(output_node)
model = graph_module.Graph(inputs=input_node, outputs=output_node).build(
keras_tuner.HyperParameters())
assert model.input_shape == (None, 30)
assert model.output_shape == (None, 1)
def _get_hyperparameters() -> kt.HyperParameters:
"""Returns hyperparameters for building a TFDF model."""
hp = kt.HyperParameters()
# Defines search space.
# For a complete list of configuration parameters of learners, please refer to
# https://github.com/google/yggdrasil-decision-forests/blob/main/documentation/learners.md
hp.Int('max_depth', 3, 8, default=6)
hp.Float('shrinkage', 0.01, 0.1, default=0.1)
hp.Boolean('use_hessian_gain', default=False)
return hp
def test_structured_block_search_normalize_return_tensor():
block = blocks.StructuredDataBlock(name="a")
block.column_names = ["0", "1"]
block.column_types = {"0": analysers.NUMERICAL, "1": analysers.NUMERICAL}
hp = keras_tuner.HyperParameters()
hp.values["a/" + blocks.wrapper.NORMALIZE] = True
outputs = block.build(hp, tf.keras.Input(shape=(2, ), dtype=tf.string))
assert len(nest.flatten(outputs)) == 1
def test_resnet_pretrained_error_with_two_channels():
block = blocks.ResNetBlock(pretrained=True)
with pytest.raises(ValueError) as info:
block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(224, 224, 2), dtype=tf.float32),
)
assert "When pretrained is set to True" in str(info.value)
def test_rnn_input_shape_one_dim_error():
block = blocks.RNNBlock()
with pytest.raises(ValueError) as info:
block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, ), dtype=tf.float32),
)
assert "Expect the input tensor of RNNBlock" in str(info.value)
def test_timeseries_build_return_tensor():
block = blocks.TimeseriesBlock()
block.column_names = ["0", "1"]
block.column_types = {"0": analysers.NUMERICAL, "1": analysers.NUMERICAL}
outputs = block.build(
keras_tuner.HyperParameters(),
tf.keras.Input(shape=(32, 2), dtype=tf.float32),
)
assert len(nest.flatten(outputs)) == 1
def test_reduction_2d_tensor_return_input_node():
block = blocks.TemporalReduction()
input_node = keras.Input(shape=(32,), dtype=tf.float32)
outputs = block.build(
keras_tuner.HyperParameters(),
input_node,
)
assert len(nest.flatten(outputs)) == 1
assert nest.flatten(outputs)[0] is input_node
def test_greedy_oracle_populate_space_with_no_hp(get_best_trials):
hp = keras_tuner.HyperParameters()
oracle = greedy.GreedyOracle(objective="val_loss", seed=utils.SEED)
trial = mock.Mock()
trial.hyperparameters = hp
get_best_trials.return_value = [trial]
oracle.update_space(hp)
values_a = oracle.populate_space("a")["values"]
assert len(values_a) == 0
def test_merge():
input_node1 = ak.Input(shape=(30, ))
input_node2 = ak.Input(shape=(40, ))
output_node1 = ak.DenseBlock()(input_node1)
output_node2 = ak.DenseBlock()(input_node2)
output_node = ak.Merge()([output_node1, output_node2])
output_node = ak.RegressionHead(shape=(1, ))(output_node)
model = graph_module.Graph(inputs=[input_node1, input_node2],
outputs=output_node).build(
keras_tuner.HyperParameters())
assert model.input_shape == [(None, 30), (None, 40)]
assert model.output_shape == (None, 1)
