def test_optimizer_selector(self):
pipeline = Pipeline([NetworkSelector(), OptimizerSelector()])
net_selector = pipeline[NetworkSelector.get_name()]
net_selector.add_network("mlpnet", MlpNet)
net_selector.add_network("shapedmlpnet", ShapedMlpNet)
net_selector.add_final_activation('none', nn.Sequential())
opt_selector = pipeline[OptimizerSelector.get_name()]
opt_selector.add_optimizer("adam", AdamOptimizer)
opt_selector.add_optimizer("sgd", SgdOptimizer)
pipeline_config = pipeline.get_pipeline_config()
pipeline_config["random_seed"] = 42
hyper_config = pipeline.get_hyperparameter_search_space(
).sample_configuration()
pipeline.fit_pipeline(hyperparameter_config=hyper_config,
pipeline_config=pipeline_config,
X=torch.rand(3, 3),
Y=torch.rand(3, 2),
embedding=nn.Sequential())
sampled_optimizer = opt_selector.fit_output['optimizer']
self.assertIn(type(sampled_optimizer), [optim.Adam, optim.SGD])
def test_network_selector(self):
pipeline = Pipeline([
NetworkSelector()
])
selector = pipeline[NetworkSelector.get_name()]
selector.add_network("mlpnet", MlpNet)
selector.add_network("shapedmlpnet", ShapedMlpNet)
selector.add_final_activation('none', nn.Sequential())
pipeline_config = pipeline.get_pipeline_config()
pipeline_config["random_seed"] = 42
hyper_config = pipeline.get_hyperparameter_search_space().sample_configuration()
pipeline.fit_pipeline(hyperparameter_config=hyper_config, pipeline_config=pipeline_config,
X=torch.rand(3,3), Y=torch.rand(3, 2), embedding=nn.Sequential())
sampled_network = pipeline[selector.get_name()].fit_output['network']
self.assertIn(type(sampled_network), [MlpNet, ShapedMlpNet])
def test_loss_selector(self):
pipeline = Pipeline([LossModuleSelector()])
selector = pipeline[LossModuleSelector.get_name()]
selector.add_loss_module("L1", nn.L1Loss)
selector.add_loss_module("cross_entropy", nn.CrossEntropyLoss,
LossWeightStrategyWeighted(), True)
pipeline_config = pipeline.get_pipeline_config(
loss_modules=["L1", "cross_entropy"])
pipeline_hyperparameter_config = pipeline.get_hyperparameter_search_space(
**pipeline_config).sample_configuration()
pipeline_hyperparameter_config["LossModuleSelector:loss_module"] = "L1"
pipeline.fit_pipeline(
hyperparameter_config=pipeline_hyperparameter_config,
train_indices=np.array([0, 1, 2]),
X=np.random.rand(3, 3),
Y=np.random.rand(3, 2),
pipeline_config=pipeline_config,
tmp=None)
selected_loss = pipeline[
selector.get_name()].fit_output['loss_function']
self.assertEqual(type(selected_loss.function), nn.L1Loss)
pipeline_hyperparameter_config[
"LossModuleSelector:loss_module"] = "cross_entropy"
pipeline.fit_pipeline(
hyperparameter_config=pipeline_hyperparameter_config,
train_indices=np.array([0, 1, 2]),
X=np.random.rand(3, 3),
Y=np.array([[1, 0], [0, 1], [1, 0]]),
pipeline_config=pipeline_config,
tmp=None)
selected_loss = pipeline[
selector.get_name()].fit_output['loss_function']
self.assertEqual(type(selected_loss.function), nn.CrossEntropyLoss)
self.assertEqual(
selected_loss(torch.tensor([[0.0, 10000.0]]),
torch.tensor([[0, 1]])), 0)
def test_lr_scheduler_selector(self):
pipeline = Pipeline([
NetworkSelector(),
OptimizerSelector(),
LearningrateSchedulerSelector(),
])
net_selector = pipeline[NetworkSelector.get_name()]
net_selector.add_network("mlpnet", MlpNet)
net_selector.add_network("shapedmlpnet", ShapedMlpNet)
net_selector.add_final_activation('none', nn.Sequential())
opt_selector = pipeline[OptimizerSelector.get_name()]
opt_selector.add_optimizer("adam", AdamOptimizer)
opt_selector.add_optimizer("sgd", SgdOptimizer)
lr_scheduler_selector = pipeline[
LearningrateSchedulerSelector.get_name()]
lr_scheduler_selector.add_lr_scheduler("step", SchedulerStepLR)
lr_scheduler_selector.add_lr_scheduler("exp", SchedulerExponentialLR)
pipeline_config = pipeline.get_pipeline_config()
pipeline_config["random_seed"] = 42
hyper_config = pipeline.get_hyperparameter_search_space(
).sample_configuration()
pipeline.fit_pipeline(hyperparameter_config=hyper_config,
pipeline_config=pipeline_config,
X=torch.rand(3, 3),
Y=torch.rand(3, 2),
embedding=nn.Sequential(),
training_techniques=[],
train_indices=np.array([0, 1, 2]))
sampled_lr_scheduler = pipeline[
lr_scheduler_selector.get_name()].fit_output[
'training_techniques'][0].training_components['lr_scheduler']
self.assertIn(type(sampled_lr_scheduler),
[lr_scheduler.ExponentialLR, lr_scheduler.StepLR])
def test_cross_validation(self):
class ResultNode(PipelineNode):
def fit(self, X, Y, train_indices, valid_indices):
return { 'loss': np.sum(X[valid_indices]), 'info': {'a': np.sum(X[train_indices]), 'b': np.sum(X[valid_indices])} }
pipeline = Pipeline([
CrossValidation([
ResultNode()
])
])
pipeline["CrossValidation"].add_cross_validator("k_fold", KFold, lambda x: x.reshape((-1 ,)))
pipeline["CrossValidation"].add_cross_validator("stratified_k_fold", StratifiedKFold, lambda x: x.reshape((-1 ,)))
x_train = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
y_train = np.array([[1], [0], [1]])
# test cv_splits
pipeline_config = pipeline.get_pipeline_config(cross_validator="k_fold", cross_validator_args={"n_splits": 3})
pipeline_config_space = pipeline.get_hyperparameter_search_space(**pipeline_config)
dataset_info = DataSetInfo()
dataset_info.categorical_features = [None] * 3
dataset_info.x_shape = x_train.shape
dataset_info.y_shape = y_train.shape
pipeline_config["random_seed"] = 42
cv_result = pipeline.fit_pipeline(hyperparameter_config=pipeline_config_space, pipeline_config=pipeline_config,
X_train=x_train, Y_train=y_train, X_valid=None, Y_valid=None,
budget=5, budget_type=BudgetTypeEpochs, one_hot_encoder=None,
optimize_start_time=time.time(), refit=False, dataset_info=dataset_info, rescore=False)
self.assertEqual(cv_result['loss'], 15)
self.assertDictEqual(cv_result['info'], {'a': 30, 'b': 15})
# test validation split
pipeline_config = pipeline.get_pipeline_config(validation_split=0.3)
pipeline_config_space = pipeline.get_hyperparameter_search_space(**pipeline_config)
pipeline_config['random_seed'] = 42
dataset_info = DataSetInfo()
dataset_info.categorical_features = [None] * 3
dataset_info.x_shape = x_train.shape
dataset_info.y_shape = y_train.shape
cv_result = pipeline.fit_pipeline(hyperparameter_config=pipeline_config_space, pipeline_config=pipeline_config,
X_train=x_train, Y_train=y_train, X_valid=None, Y_valid=None,
budget=5, budget_type=BudgetTypeEpochs, one_hot_encoder=None,
optimize_start_time=time.time(), refit=False, dataset_info=dataset_info, rescore=False)
self.assertEqual(cv_result['loss'], 24)
self.assertDictEqual(cv_result['info'], {'a': 21, 'b': 24})
# test stratified cv split
x_valid = x_train
y_valid = y_train
x_train = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12], [13, 14, 15], [16, 17, 18]])
y_train = np.array([[1], [1], [0], [0], [1], [0]])
pipeline_config = pipeline.get_pipeline_config(cross_validator="stratified_k_fold", cross_validator_args={"n_splits": 3})
pipeline_config_space = pipeline.get_hyperparameter_search_space(**pipeline_config)
pipeline_config['random_seed'] = 42
dataset_info = DataSetInfo()
dataset_info.categorical_features = [None] * 3
dataset_info.x_shape = x_train.shape
dataset_info.y_shape = y_train.shape
cv_result = pipeline.fit_pipeline(hyperparameter_config=pipeline_config_space, pipeline_config=pipeline_config,
X_train=x_train, Y_train=y_train, X_valid=None, Y_valid=None,
budget=5, budget_type=BudgetTypeEpochs, one_hot_encoder=None,
optimize_start_time=time.time(), refit=False, dataset_info=dataset_info, rescore=False)
self.assertEqual(cv_result['loss'], 57)
self.assertDictEqual(cv_result['info'], {'a': 114, 'b': 57})
pipeline_config = pipeline.get_pipeline_config()
pipeline_config_space = pipeline.get_hyperparameter_search_space(**pipeline_config)
pipeline_config['random_seed'] = 42
dataset_info = DataSetInfo()
dataset_info.categorical_features = [None] * 3
dataset_info.x_shape = x_train.shape
dataset_info.y_shape = y_train.shape
cv_result = pipeline.fit_pipeline(hyperparameter_config=pipeline_config_space, pipeline_config=pipeline_config,
X_train=x_train, Y_train=y_train, X_valid=x_valid, Y_valid=y_valid,
budget=5, budget_type=BudgetTypeEpochs, one_hot_encoder=None,
optimize_start_time=time.time(), refit=False, dataset_info=dataset_info, rescore=False)
self.assertEqual(cv_result['loss'], 45)
self.assertDictEqual(cv_result['info'], {'a': 171, 'b': 45})
def test_cross_validation(self):
class ResultNode(PipelineNode):
def fit(self, X_train, X_valid):
return {
'loss': np.sum(X_valid),
'info': {
'a': np.sum(X_train),
'b': np.sum(X_valid)
}
}
pipeline = Pipeline([CrossValidation([ResultNode()])])
x_train = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])
y_train = np.array([[1], [0], [1]])
# test cv_splits
pipeline_config = pipeline.get_pipeline_config(cv_splits=3)
pipeline_config_space = pipeline.get_hyperparameter_search_space(
**pipeline_config)
pipeline_config['categorical_features'] = None
cv_result = pipeline.fit_pipeline(
hyperparameter_config=pipeline_config_space,
pipeline_config=pipeline_config,
X_train=x_train,
Y_train=y_train,
X_valid=None,
Y_valid=None,
budget=5,
budget_type=BudgetTypeEpochs,
one_hot_encoder=None,
optimize_start_time=time.time())
self.assertEqual(cv_result['loss'], 15)
self.assertDictEqual(cv_result['info'], {'a': 30, 'b': 15})
# test validation split
pipeline_config = pipeline.get_pipeline_config(validation_split=0.3)
pipeline_config_space = pipeline.get_hyperparameter_search_space(
**pipeline_config)
pipeline_config['categorical_features'] = None
cv_result = pipeline.fit_pipeline(
hyperparameter_config=pipeline_config_space,
pipeline_config=pipeline_config,
X_train=x_train,
Y_train=y_train,
X_valid=None,
Y_valid=None,
budget=5,
budget_type=BudgetTypeEpochs,
one_hot_encoder=None,
optimize_start_time=time.time())
self.assertEqual(cv_result['loss'], 24)
self.assertDictEqual(cv_result['info'], {'a': 21, 'b': 24})
# test stratified cv split
x_train = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12],
[13, 14, 15], [16, 17, 18]])
y_train = np.array([[1], [1], [0], [0], [1], [0]])
pipeline_config = pipeline.get_pipeline_config(
cv_splits=3, use_stratified_cv_split=True)
pipeline_config_space = pipeline.get_hyperparameter_search_space(
**pipeline_config)
pipeline_config['categorical_features'] = None
cv_result = pipeline.fit_pipeline(
hyperparameter_config=pipeline_config_space,
pipeline_config=pipeline_config,
X_train=x_train,
Y_train=y_train,
X_valid=None,
Y_valid=None,
budget=5,
budget_type=BudgetTypeEpochs,
one_hot_encoder=None,
optimize_start_time=time.time())
self.assertEqual(cv_result['loss'], 57)
self.assertDictEqual(cv_result['info'], {'a': 114, 'b': 57})