def test_JobConfig():
dfX, y, auto_ml_config = get_automl_config()
job_config = JobConfig()
job_config.guess_cv(auto_ml_config)
assert job_config.cv is not None
job_config.guess_scoring(auto_ml_config)
assert isinstance(job_config.scoring, list)
def initialize(self):
""" method to initialize auto_ml_config and job_config """
##################################
### ** load data and target ** ###
##################################
if self.dfX is None or self.y is None:
temp = self.loader()
if len(temp) == 2:
self.dfX, self.y = temp
self.groups = None
else:
self.dfX, self.y, self.groups = temp
###########################################
### ** create database configuration ** ###
###########################################
if self.auto_ml_config is None:
self.auto_ml_config = AutoMlConfig(dfX=self.dfX,
y=self.y,
groups=self.groups,
name=self.name)
self.auto_ml_config.guess_everything()
#####################################
### ** create job configuation ** ###
#####################################
if self.job_config is None:
self.job_config = JobConfig()
self.job_config.guess_cv(auto_ml_config=self.auto_ml_config,
n_splits=10)
self.job_config.guess_scoring(auto_ml_config=self.auto_ml_config)
###################################
### ** create data persister ** ###
###################################
if self.data_persister is None:
self.data_persister = FolderDataPersister(
base_folder=self.base_folder)
#################################
### ** apply custom config ** ###
#################################
if self.set_configs is not None:
self.set_configs(self)
def test_JobConfig(num_only):
dfX, y, auto_ml_config = get_automl_config(num_only)
job_config = JobConfig()
job_config.guess_cv(auto_ml_config)
assert job_config.cv is not None
assert hasattr(job_config.cv, "split")
job_config.guess_scoring(auto_ml_config)
assert isinstance(job_config.scoring, list)
assert hasattr(job_config, "allow_approx_cv")
assert hasattr(job_config, "start_with_default")
assert hasattr(job_config, "do_blocks_search")
assert isinstance(job_config.allow_approx_cv, bool)
assert isinstance(job_config.start_with_default, bool)
assert isinstance(job_config.do_blocks_search, bool)
with pytest.raises(ValueError):
job_config.cv = "this is not a cv"
def test_JobConfig_additional_scoring_function():
job_config = JobConfig()
assert job_config.additional_scoring_function is None
def f(x):
return x + 1
job_config.additional_scoring_function = f
assert job_config.additional_scoring_function is not None
assert job_config.additional_scoring_function(1) == 2
with pytest.raises(TypeError):
job_config.additional_scoring_function = 10 # no a function
def f(x):
return x + 1
job_config.additional_scoring_function = f
assert job_config.additional_scoring_function is not None
assert job_config.additional_scoring_function(1) == 2
with pytest.raises(TypeError):
job_config.additional_scoring_function = 10 # no a function
def test_create_everything_sequentially(num_only, tmpdir):
# DataPersister
data_persister = FolderDataPersister(base_folder=tmpdir)
# Data
dfX, y = loader(num_only)
# Auto Ml Config
auto_ml_config = AutoMlConfig(dfX, y)
auto_ml_config.guess_everything()
assert auto_ml_config
# Job Config
job_config = JobConfig()
job_config.guess_scoring(auto_ml_config)
job_config.guess_cv(auto_ml_config)
assert job_config
# Result Reader
result_reader = AutoMlResultReader(data_persister)
assert result_reader # just verify the object was created
# Auto ml guider
auto_ml_guider = AutoMlModelGuider(result_reader=result_reader,
job_config=job_config,
metric_transformation="default",
avg_metric=True)
assert auto_ml_guider
# Job Controller
job_controller = MlJobManager(
auto_ml_config=auto_ml_config,
job_config=job_config,
auto_ml_guider=auto_ml_guider,
data_persister=data_persister,
seed=None,
)
assert job_controller
# Job Runner
job_runner = MlJobRunner(
dfX=dfX,
y=y,
groups=None,
auto_ml_config=auto_ml_config,
job_config=job_config,
data_persister=data_persister,
seed=None,
)
assert job_runner
### Do one iteration of the job_controller
for i, (temp_job_id,
temp_job_param) in enumerate(job_controller.iterate()):
if i > 0:
break # I need to complete a loop, so I need the break to be AFTER second yield
job_id = temp_job_id
job_param = temp_job_param
assert isinstance(job_id, str)
assert isinstance(job_param, dict)
### retriveve job by worker
for worker_job_id, worker_job_param in job_runner.iterate():
break
assert isinstance(worker_job_id, str)
assert isinstance(worker_job_param, dict)
assert worker_job_id == job_id
assert worker_job_param == job_param