test_suite.py

# ----------------------------------------------------------------------------
# Copyright (C) 2021 Deepchecks (https://www.deepchecks.com)
#
# This file is part of Deepchecks.
# Deepchecks is distributed under the terms of the GNU Affero General
# Public License (version 3 or later).
# You should have received a copy of the GNU Affero General Public License
# along with Deepchecks.  If not, see <http://www.gnu.org/licenses/>.
# ----------------------------------------------------------------------------
#
#
from collections import defaultdict

from hamcrest import assert_that, calling, contains_inanyorder, has_length, instance_of, is_, raises

from deepchecks.core import CheckResult, DatasetKind
from deepchecks.core.errors import DatasetValidationError, DeepchecksNotSupportedError, DeepchecksValueError
from deepchecks.vision.base_checks import SingleDatasetCheck, TrainTestCheck
from deepchecks.vision.datasets.classification import mnist_tensorflow
from deepchecks.vision.datasets.detection import coco_torch
from deepchecks.vision.suite import Suite
from deepchecks.vision.suites.default_suites import full_suite
from tests.common import get_expected_results_length, validate_suite_result


def test_suite_execution():
    coco_dataset = coco_torch.load_dataset(object_type='VisionData')
    executions = defaultdict(int)

    class DummyCheck(SingleDatasetCheck):
        def initialize_run(self, context, dataset_kind: DatasetKind):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context, dataset_kind: DatasetKind) -> CheckResult:
            executions["compute"] += 1
            return CheckResult(0)

    class DummyTrainTestCheck(TrainTestCheck):
        def initialize_run(self, context):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context) -> CheckResult:
            executions["compute"] += 1
            return CheckResult(1)

    suite = Suite("test",
                  DummyCheck(),
                  DummyTrainTestCheck())
    args = {'train_dataset': coco_dataset, 'test_dataset': coco_dataset}
    result = suite.run(**args)
    length = get_expected_results_length(suite, args)
    validate_suite_result(result, length)

    assert_that(result.results[0].value, is_(0))
    assert_that(result.results[2].value, is_(1))
    assert_that(executions, is_({'initialize_run': 3, 'update': 8, 'compute': 3}))


def test_suite_execution_with_initalize_exeption():
    coco_dataset = coco_torch.load_dataset(object_type='VisionData')
    executions = defaultdict(int)

    class DummyCheck(SingleDatasetCheck):
        def initialize_run(self, context, dataset_kind: DatasetKind):
            executions["initialize_run"] += 1
            raise DeepchecksValueError('bad init')

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context, dataset_kind: DatasetKind) -> CheckResult:
            executions["compute"] += 1
            return CheckResult(None)

    class DummyTrainTestCheck(TrainTestCheck):
        def initialize_run(self, context):
            executions["initialize_run"] += 1
            raise DeepchecksValueError('bad init')

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context) -> CheckResult:
            executions["compute"] += 1
            return CheckResult(None)

    suite = Suite("test",
                  DummyCheck(),
                  DummyTrainTestCheck())

    args = {'train_dataset': coco_dataset, 'test_dataset': coco_dataset}
    result = suite.run(**args)
    length = get_expected_results_length(suite, args)
    validate_suite_result(result, length)

    assert_that(result.results[0].exception, instance_of(DeepchecksValueError))
    assert_that(result.results[0].exception.message, is_('bad init'))

    assert_that(result.results[1].exception, instance_of(DeepchecksValueError))
    assert_that(result.results[1].exception.message, is_('bad init'))

    assert_that(executions, is_({'initialize_run': 3}))


def test_suite_execution_with_exception_on_compute():
    coco_dataset = coco_torch.load_dataset(object_type='VisionData')
    executions = defaultdict(int)

    class DummyCheck(SingleDatasetCheck):
        def initialize_run(self, context, dataset_kind: DatasetKind):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context, dataset_kind: DatasetKind) -> CheckResult:
            executions["compute"] += 1
            raise DeepchecksValueError('bad compute')
            return CheckResult(None)

    class DummyTrainTestCheck(TrainTestCheck):
        def initialize_run(self, context):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context) -> CheckResult:
            executions["compute"] += 1
            raise DeepchecksValueError('bad compute')

    suite = Suite("test",
                  DummyCheck(),
                  DummyTrainTestCheck())

    args = {'train_dataset': coco_dataset, 'test_dataset': coco_dataset}
    result = suite.run(**args)
    length = get_expected_results_length(suite, args)
    validate_suite_result(result, length)

    assert_that(result.results[0].exception, instance_of(DeepchecksValueError))
    assert_that(result.results[0].exception.message, is_('bad compute'))

    assert_that(result.results[1].exception, instance_of(DeepchecksValueError))
    assert_that(result.results[1].exception.message, is_('bad compute'))

    assert_that(executions, is_({'initialize_run': 3, 'update': 8, 'compute': 3}))


def test_suite_execution_with_missing_train():
    coco_dataset = coco_torch.load_dataset(object_type='VisionData')
    executions = defaultdict(int)

    class DummyTrainTestCheck(TrainTestCheck):
        def initialize_run(self, context):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context) -> CheckResult:
            executions["compute"] += 1
            raise DeepchecksValueError('bad compute')

    suite = Suite("test",
                  DummyTrainTestCheck())
    assert_that(calling(suite.run).with_args(test_dataset=coco_dataset),
                raises(DatasetValidationError,
                       "Can't initialize context with only test. if you have single dataset, initialize it as train"))

    assert_that(executions, is_({}))


def test_suite_execution_with_missing_test():
    coco_dataset = coco_torch.load_dataset(object_type='VisionData')
    executions = defaultdict(int)

    class DummyTrainTestCheck(TrainTestCheck):
        def initialize_run(self, context):
            executions["initialize_run"] += 1

        def update(self, context, batch, dataset_kind: DatasetKind):
            executions["update"] += 1

        def compute(self, context) -> CheckResult:
            executions["compute"] += 1
            return CheckResult(None)

    suite = Suite("test",
                  DummyTrainTestCheck())

    args = {'train_dataset': coco_dataset}
    result = suite.run(**args)
    length = get_expected_results_length(suite, args)
    validate_suite_result(result, length)

    assert_that(result.results[0].exception, instance_of(DeepchecksNotSupportedError))
    assert_that(result.results[0].exception.message,
                is_('Check is irrelevant if not supplied with both train and test datasets'))

    assert_that(executions, has_length(0))


def test_full_suite_execution_mnist(mnist_visiondata_train, mnist_visiondata_test,
                                    mnist_iterator_visiondata_train, mnist_iterator_visiondata_test):
    suite = full_suite(imaginery_kwarg='just to make sure all checks have kwargs in the init')
    arguments = (
        dict(train_dataset=mnist_visiondata_train, test_dataset=mnist_visiondata_test),
        dict(train_dataset=mnist_visiondata_train),
        dict(train_dataset=mnist_visiondata_train, with_display=False),
        dict(train_dataset=mnist_iterator_visiondata_train, test_dataset=mnist_iterator_visiondata_test,
             max_samples=100),
    )

    for args in arguments:
        result = suite.run(**args)
        length = get_expected_results_length(suite, args)
        validate_suite_result(result, length)


def test_full_suite_execution_mnist_tf():
    suite = full_suite(imaginery_kwarg='just to make sure all checks have kwargs in the init')
    mnist_visiondata_train = mnist_tensorflow.load_dataset(train=True)
    mnist_visiondata_test = mnist_tensorflow.load_dataset(train=False)
    arguments = (
        dict(train_dataset=mnist_visiondata_train, test_dataset=mnist_visiondata_test),
    )

    for args in arguments:
        result = suite.run(**args)
        length = get_expected_results_length(suite, args)
        validate_suite_result(result, length)


def test_full_suite_execution_coco_torch(coco_visiondata_train, coco_visiondata_test):
    suite = full_suite(imaginery_kwarg='just to make sure all checks have kwargs in the init')
    arguments = (
        dict(train_dataset=coco_visiondata_train, test_dataset=coco_visiondata_test),
        dict(train_dataset=coco_visiondata_train),
        dict(train_dataset=coco_visiondata_train, with_display=False),
    )

    for args in arguments:
        result = suite.run(**args)
        length = get_expected_results_length(suite, args)
        validate_suite_result(result, length)

# TODO: Again started to fail
# def test_full_suite_execution_coco_tf(tf_coco_visiondata_train, tf_coco_visiondata_test):
#      suite = full_suite(imaginery_kwarg='just to make sure all checks have kwargs in the init')
#      arguments = (
#          dict(train_dataset=tf_coco_visiondata_train, test_dataset=tf_coco_visiondata_test),
#      )

#      for args in arguments:
#          result = suite.run(**args)
#          length = get_expected_results_length(suite, args)
#          validate_suite_result(result, length)

def test_single_dataset(coco_visiondata_train, coco_visiondata_test):
    suite = full_suite()
    res_train = suite.run(coco_visiondata_train, coco_visiondata_test,
                          max_samples=100, with_display=False, run_single_dataset='Train')
    expected_train_headers = ['Class Performance', 'Confusion Matrix', 'Property Label Correlation Change',
                              'Heatmap Comparison', 'Image Dataset Drift', 'Image Property Drift',
                              'Image Property Outliers', 'Label Property Outliers', 'Mean Average Precision Report',
                              'Mean Average Recall Report', 'New Labels', 'Property Label Correlation',
                              'Simple Model Comparison', 'Label Drift', 'Prediction Drift',
                              'Weak Segments Performance']

    res_test = suite.run(coco_visiondata_train, coco_visiondata_test,
                         max_samples=100, with_display=False, run_single_dataset='Test')
    res_full = suite.run(coco_visiondata_train, coco_visiondata_test,
                         max_samples=100, with_display=False)
    res_names = sorted(x.get_header() for x in res_train.results)
    assert_that(res_names, contains_inanyorder(*expected_train_headers))
    assert_that(res_test.results, has_length(16))
    assert_that(res_full.results, has_length(23))