def test_run_experiment_lr_with_object_no_configdir():
"""Test rsmtool using a Configuration object and no specified configdir."""
source = 'lr-object-no-path'
experiment_id = 'lr_object_no_path'
# set up a temporary directory since
# we will be using getcwd
old_file_dict = {'train': 'data/files/train.csv',
'test': 'data/files/test.csv',
'features': 'data/experiments/lr-object-no-path/features.csv'}
temp_dir = tempfile.TemporaryDirectory(prefix=getcwd())
new_file_dict = copy_data_files(temp_dir.name,
old_file_dict,
rsmtool_test_dir)
config_dict = {"train_file": new_file_dict['train'],
"id_column": "ID",
"use_scaled_predictions": True,
"test_label_column": "score",
"train_label_column": "score",
"test_file": new_file_dict['test'],
"trim_max": 6,
"features": new_file_dict['features'],
"trim_min": 1,
"model": "LinearRegression",
"experiment_id": "lr_object_no_path",
"description": "Using all features with an LinearRegression model."}
config_obj = Configuration(config_dict)
check_run_experiment(source,
experiment_id,
config_obj_or_dict=config_obj)
def test_run_experiment_lr_with_object_and_configdir():
"""Test rsmtool using a Configuration object and specified configdir."""
source = 'lr-object'
experiment_id = 'lr_object'
configdir = join(rsmtool_test_dir,
'data',
'experiments',
source)
config_dict = {"train_file": "../../files/train.csv",
"id_column": "ID",
"use_scaled_predictions": True,
"test_label_column": "score",
"train_label_column": "score",
"test_file": "../../files/test.csv",
"trim_max": 6,
"features": "features.csv",
"trim_min": 1,
"model": "LinearRegression",
"experiment_id": "lr_object",
"description": "Using all features with an LinearRegression model."}
config_obj = Configuration(config_dict, configdir=configdir)
check_run_experiment(source,
experiment_id,
config_obj_or_dict=config_obj)
def test_run_experiment_lr_with_dictionary():
# Passing a dictionary as input.
source = 'lr-dictionary'
experiment_id = 'lr_dictionary'
old_file_dict = {'train': 'data/files/train.csv',
'test': 'data/files/test.csv',
'features': 'data/experiments/lr-dictionary/features.csv'}
temp_dir = tempfile.TemporaryDirectory(prefix=getcwd())
new_file_dict = copy_data_files(temp_dir.name,
old_file_dict,
rsmtool_test_dir)
config_dict = {"train_file": new_file_dict['train'],
"id_column": "ID",
"use_scaled_predictions": True,
"test_label_column": "score",
"train_label_column": "score",
"test_file": new_file_dict['test'],
"trim_max": 6,
"features": new_file_dict['features'],
"trim_min": 1,
"model": "LinearRegression",
"experiment_id": "lr_dictionary",
"description": "Using all features with an LinearRegression model."}
check_run_experiment(source,
experiment_id,
config_obj_or_dict=config_dict)
def test_run_experiment_parameterized(*args, **kwargs):
if TEST_DIR:
kwargs['given_test_dir'] = TEST_DIR
# suppress known convergence warnings for LinearSVR-based experiments
# TODO: once SKLL hyperparameters can be passed, replace this code
if args[0].startswith('linearsvr'):
kwargs['suppress_warnings_for'] = [ConvergenceWarning]
check_run_experiment(*args, **kwargs)
def test_run_experiment_lr_with_object_and_filepath():
"""Test for a rare use case where an old Configuration object is passed."""
source = 'lr-object'
experiment_id = 'lr_object'
config_file = join(rsmtool_test_dir,
'data',
'experiments',
source,
'{}.json'.format(experiment_id))
config_dict = {"train_file": "../../files/train.csv",
"id_column": "ID",
"use_scaled_predictions": True,
"test_label_column": "score",
"train_label_column": "score",
"test_file": "../../files/test.csv",
"trim_max": 6,
"features": "features.csv",
"trim_min": 1,
"model": "LinearRegression",
"experiment_id": "lr_object",
"description": "Using all features with an LinearRegression model."}
config_obj = Configuration(config_dict)
# we catch the deprecation warning triggered by this line
with warnings.catch_warnings():
warnings.filterwarnings('ignore', category=DeprecationWarning)
config_obj.filepath = config_file
# we have to explicitly remove configdir attribute
# since it will always be assigned a value by the current code
del(config_obj.configdir)
check_run_experiment(source,
experiment_id,
config_obj_or_dict=config_obj)
def test_run_experiment_parameterized(*args, **kwargs):
if TEST_DIR:
kwargs['given_test_dir'] = TEST_DIR
check_run_experiment(*args, **kwargs)