def setUp(self):
set_project_path(testenv.MOCK_PROJECT_PATH)
self.path_to_dataset = testenv.IRIS_DATASET
self.path_to_test_dir = testenv.TEST_PATH
self.aux_project_name = testenv.MOCK_PROJECT_NAME
self.path_to_auxproj = testenv.MOCK_PROJECT_PATH
self.project_default_name = testenv.DEFAULT_PROJECT_NAME
# Generate a project
self.p = Project(path=self.path_to_test_dir,
name=self.aux_project_name)
# Add a dataset
self.ds = Dataset.read_file(path=self.path_to_dataset, )
self.ds.save()
# Generate subdataset
self.sbds = SubDataset(self.ds, method="k_fold", by=5)
self.sbds.save()
# set apporach
self.approach = Approach(self.p,
"decision_tree",
self.sbds,
path=str(Path(testenv.TEST_PATH, "dt")))
shutil.copyfile(testenv.IRIS_APPROACH, str(self.approach.script_path))
self.approach.save()
# generate runs
import_from("test.dt.decision_tree", "DecisionTreeApproach")().run()
def setUp(self):
set_project_path(testenv.MOCK_PROJECT_PATH)
self.path_to_dataset = testenv.MOCK_DATASET
self.path_to_test_dir = testenv.TEST_PATH
self.aux_project_name = testenv.MOCK_PROJECT_NAME
self.path_to_auxproj = testenv.MOCK_PROJECT_PATH
self.project_default_name = testenv.DEFAULT_PROJECT_NAME
# Generate a project
self.p = Project(path=self.path_to_test_dir,
name=self.aux_project_name)
# Add a dataset
self.ds = Dataset.read_file(path=self.path_to_dataset)
self.ds.save()
# Generate subdataset
self.sbds = SubDataset(self.ds, method="k_fold", by=5)
self.sbds.save()
# set apporach
self.approach = Approach(self.p,
"logistic_regression",
self.sbds,
path=str(Path(testenv.TEST_PATH, "lr")))
shutil.copyfile(testenv.APPROACH_EXAMPLE,
str(self.approach.script_path))
self.approach.save()
# generate runs
import_from("test.lr.logistic_regression",
"LogisticRegressionApproach")().run()
def test_get_subdataset_runs(self):
runnable = import_from("test.lr.logistic_regression",
"LogisticRegressionApproach")
runs = RunGenerator.from_runnable_approach(runnable())
for run in runs:
run.save()
runs = Approach.load(self.approach.id).runs
self.assertTrue(len(runs) > 0)
for run in runs:
self.assertIsInstance(run, Run)
return runs
def run(approach_id, resume):
if not _is_running_in_project():
print("You must use driftai CLI inside an driftai project directory")
return
if not Approach.collection().exists(approach_id):
print("Approach with id {} does not exist.".format(approach_id))
return
sys.path.append(Project.load().path)
namespace = 'approaches.' + approach_id
cls_name = to_camel_case(approach_id) + "Approach"
approach_cls = import_from(namespace, cls_name)
approach_cls().run(resume=resume)
def test_generate_runs_from_subdataset(self):
# Trick to load runnable approach
LogisticRegressionApproach = import_from("test.lr.logistic_regression",
"LogisticRegressionApproach")
ra = LogisticRegressionApproach()
# Generate the runs
run_gens = RunGenerator.from_runnable_approach(ra)
# Write runs to database
ra.approach.runs = run_gens
ra.approach.update()
# Reload approach to test if runs were correctly stored
approach = Approach.load(ra.approach.id)
self.assertEqual(len(approach.runs), len(run_gens))
def load_from_data(data):
"""
Create datasource from serialized data
Parameters
----------
data: dict
Dictionary containing serialized datasource data
Returns
-------
Datasource
"""
datasource_class_ = import_from(data["module"], data["class_name"])
del data["class_name"]
del data["module"]
return datasource_class_(**data)
def _custom_datasource_class(dtype):
split = dtype.split('.')
return import_from('.'.join(split[:-1]), split[-1])