def test_max_similarity():
"""Tests that the maximum similarity is properly returned."""
submission = SubmissionFactory.build(
similarity=0,
mt_similarity=0,
)
assert submission.max_similarity == 0
submission = SubmissionFactory.build(
similarity=0.5,
mt_similarity=0.6,
)
assert submission.max_similarity == 0.6
submission = SubmissionFactory.build(
similarity=0.5,
mt_similarity=None,
)
assert submission.max_similarity == 0.5
submission = SubmissionFactory.build(
similarity=None,
mt_similarity=None,
)
assert submission.max_similarity == 0
def test_needs_scorelog():
"""Tests if the submission needs to be logged or not."""
# Changing the STATE from UNTRANSLATED won't record any logs
submission = SubmissionFactory.build(
field=SubmissionFields.STATE,
type=SubmissionTypes.NORMAL,
old_value=UNTRANSLATED,
new_value=TRANSLATED,
)
assert not submission.needs_scorelog()
# Changing other fields (or even STATE, in a different direction) should
# need to record a score log
submission = SubmissionFactory.build(
field=SubmissionFields.STATE,
type=SubmissionTypes.NORMAL,
old_value=TRANSLATED,
new_value=UNTRANSLATED,
)
assert submission.needs_scorelog()
submission = SubmissionFactory.build(
field=SubmissionFields.TARGET,
type=SubmissionTypes.SUGG_ADD,
old_value=u'',
new_value=u'',
)
assert submission.needs_scorelog()
def test_create_evaluation_is_idempotent(self):
s = SubmissionFactory(
challenge=self.method.challenge,
algorithm_image=self.algorithm_image,
)
s.create_evaluation()
assert AlgorithmEvaluation.objects.count() == 2
def test_submission_conversion(capsys, submission_file, settings):
# Override the celery settings
settings.task_eager_propagates = (True,)
settings.task_always_eager = (True,)
challenge = ChallengeFactory()
test_set = challenge.imageset_set.get(phase=ImageSet.TESTING)
with mute_signals(post_save):
submission = SubmissionFactory(
file__from_path=submission_file, challenge=challenge
)
call_command("convertsubmissions", challenge.short_name)
_, err = capsys.readouterr()
assert err == ""
annotation_set = AnnotationSet.objects.all()[0]
assert annotation_set.submission == submission
assert annotation_set.base == test_set
images = annotation_set.images.all()
assert len(images) == 1
assert images[0].name == "image10x10x10.mhd"
with mute_signals(post_save):
submission = SubmissionFactory(
file__from_path=Path(__file__).parent.parent
/ "evaluation_tests"
/ "resources"
/ "submission.csv",
challenge=challenge,
)
call_command("convertsubmissions", challenge.short_name)
_, err = capsys.readouterr()
assert err == ""
annotation_set = AnnotationSet.objects.all()[1]
assert annotation_set.submission == submission
assert annotation_set.base == test_set
labels = annotation_set.labels
assert len(labels) == 10
assert labels[0]["class"] == 0
def test_submission_time_limit(client, TwoChallengeSets):
SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
def get_submission_view():
return get_view_for_user(
viewname="evaluation:submission-create",
challenge=TwoChallengeSets.ChallengeSet1.challenge,
client=client,
user=TwoChallengeSets.ChallengeSet1.participant,
)
assert "make 9 more" in get_submission_view().rendered_content
s = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
s.created = timezone.now() - timedelta(hours=23)
s.save()
assert "make 8 more" in get_submission_view().rendered_content
s = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
s.created = timezone.now() - timedelta(hours=25)
s.save()
assert "make 8 more" in get_submission_view().rendered_content
def test_invoice_get_rates_inconsistent_scorelog_rates(member, store0):
USER_RATE_ONE = 0.5
USER_RATE_TWO = 0.2
# Set some rate
member.rate = USER_RATE_ONE
member.review_rate = USER_RATE_ONE
member.save()
month = timezone.make_aware(timezone.datetime(2014, 4, 1))
submission_kwargs = {
"store": store0,
"unit": store0.units[0],
"field": SubmissionFields.TARGET,
"type": SubmissionTypes.NORMAL,
"old_value": "foo",
"new_value": "bar",
"submitter": member,
"translation_project": store0.translation_project,
"creation_time": month,
}
scorelog_kwargs = {
"wordcount": 1,
"similarity": 0,
"action_code": TranslationActionCodes.NEW,
"creation_time": month,
"user": member,
"submission": SubmissionFactory(**submission_kwargs),
}
ScoreLogFactory(**scorelog_kwargs)
# Alter rates, producing an inconsistent state when recording the ScoreLog
member.rate = USER_RATE_TWO
member.review_rate = USER_RATE_TWO
member.save()
submission_kwargs["unit"] = store0.units[1]
scorelog_kwargs["submission"] = SubmissionFactory(**submission_kwargs)
ScoreLogFactory(**scorelog_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
with pytest.raises(ValueError) as e:
invoice.get_rates()
assert "Multiple rate values recorded for user %s" % (
member.username, ) in str(e.value)
def test_invoice_get_rates_inconsistent_scorelog_rates(member, store0):
USER_RATE_ONE = 0.5
USER_RATE_TWO = 0.2
# Set some rate
member.rate = USER_RATE_ONE
member.review_rate = USER_RATE_ONE
member.save()
month = timezone.datetime(2014, 04, 01)
submission_kwargs = {
'store': store0,
'unit': store0.units[0],
'field': SubmissionFields.TARGET,
'type': SubmissionTypes.NORMAL,
'old_value': 'foo',
'new_value': 'bar',
'submitter': member,
'translation_project': store0.translation_project,
'creation_time': month,
}
scorelog_kwargs = {
'wordcount': 1,
'similarity': 0,
'action_code': TranslationActionCodes.NEW,
'creation_time': month,
'user': member,
'submission': SubmissionFactory(**submission_kwargs),
}
ScoreLogFactory(**scorelog_kwargs)
# Alter rates, producing an inconsistent state when recording the ScoreLog
member.rate = USER_RATE_TWO
member.review_rate = USER_RATE_TWO
member.save()
submission_kwargs['unit'] = store0.units[1]
scorelog_kwargs['submission'] = SubmissionFactory(**submission_kwargs)
ScoreLogFactory(**scorelog_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
with pytest.raises(ValueError) as e:
invoice.get_rates()
assert ('Multiple rate values recorded for user %s' % (member.username, )
in e.value.message)
def test_submission_evaluation(
client, evaluation_image, submission_file, settings
):
# Override the celery settings
settings.task_eager_propagates = (True,)
settings.task_always_eager = (True,)
settings.broker_url = ("memory://",)
settings.backend = "memory"
# Upload a submission and create a job
dockerclient = docker.DockerClient(
base_url=settings.CONTAINER_EXEC_DOCKER_BASE_URL
)
eval_container, sha256 = evaluation_image
method = MethodFactory(
image__from_path=eval_container, image_sha256=sha256, ready=True
)
# We should not be able to download methods
response = client.get(method.image.url)
assert response.status_code == 404
num_containers_before = len(dockerclient.containers.list())
num_volumes_before = len(dockerclient.volumes.list())
# This will create a job, and we'll wait for it to be executed
submission = SubmissionFactory(
file__from_path=submission_file, challenge=method.challenge
)
# The evaluation method should clean up after itself
assert len(dockerclient.volumes.list()) == num_volumes_before
assert len(dockerclient.containers.list()) == num_containers_before
# The evaluation method should return the correct answer
assert len(submission.job_set.all()) == 1
assert submission.job_set.all()[0].result.metrics["acc"] == 0.5
# Try with a csv file
submission = SubmissionFactory(
file__from_path=Path(__file__).parent / "resources" / "submission.csv",
challenge=method.challenge,
)
assert len(submission.job_set.all()) == 1
assert submission.job_set.all()[0].result.metrics["acc"] == 0.5
def test_submission_download(client, two_challenge_sets):
"""Only the challenge admin should be able to download submissions."""
submission = SubmissionFactory(
challenge=two_challenge_sets.challenge_set_1.challenge,
creator=two_challenge_sets.challenge_set_1.participant,
)
tests = [
# (
# image response + annotation response not test ground truth,
# user
# )
(403, None),
(403, two_challenge_sets.challenge_set_1.non_participant),
(403, two_challenge_sets.challenge_set_1.participant),
(403, two_challenge_sets.challenge_set_1.participant1),
(302, two_challenge_sets.challenge_set_1.creator),
(302, two_challenge_sets.challenge_set_1.admin),
(403, two_challenge_sets.challenge_set_2.non_participant),
(403, two_challenge_sets.challenge_set_2.participant),
(403, two_challenge_sets.challenge_set_2.participant1),
(403, two_challenge_sets.challenge_set_2.creator),
(403, two_challenge_sets.challenge_set_2.admin),
(302, two_challenge_sets.admin12),
(403, two_challenge_sets.participant12),
(302, two_challenge_sets.admin1participant2),
]
for test in tests:
response = get_view_for_user(
url=submission.file.url, client=client, user=test[1]
)
assert response.status_code == test[0]
def test_is_similarity_taken_from_mt(similarity, mt_similarity):
submission = SubmissionFactory.build(similarity=similarity,
mt_similarity=mt_similarity)
score_log = ScoreLogFactory.build(submission=submission)
if submission.similarity < submission.mt_similarity:
assert score_log.is_similarity_taken_from_mt()
else:
assert not score_log.is_similarity_taken_from_mt()
def test_submission_evaluation(client, evaluation_image, submission_file,
settings):
# Override the celery settings
settings.task_eager_propagates = (True, )
settings.task_always_eager = (True, )
# Upload a submission and create an evaluation
dockerclient = docker.DockerClient(
base_url=settings.COMPONENTS_DOCKER_BASE_URL)
eval_container, sha256 = evaluation_image
method = MethodFactory(image__from_path=eval_container,
image_sha256=sha256,
ready=True)
# We should not be able to download methods
with pytest.raises(NotImplementedError):
_ = method.image.url
num_containers_before = len(dockerclient.containers.list())
num_volumes_before = len(dockerclient.volumes.list())
# This will create an evaluation, and we'll wait for it to be executed
submission = SubmissionFactory(predictions_file__from_path=submission_file,
challenge=method.challenge)
# The evaluation method should clean up after itself
assert len(dockerclient.volumes.list()) == num_volumes_before
assert len(dockerclient.containers.list()) == num_containers_before
# The evaluation method should return the correct answer
assert len(submission.evaluation_set.all()) == 1
assert (submission.evaluation_set.first().outputs.get(
interface__slug="metrics-json-file").value["acc"] == 0.5)
# Try with a csv file
submission = SubmissionFactory(
predictions_file__from_path=Path(__file__).parent / "resources" /
"submission.csv",
challenge=method.challenge,
)
assert len(submission.evaluation_set.all()) == 1
assert (submission.evaluation_set.first().outputs.get(
interface__slug="metrics-json-file").value["acc"] == 0.5)
def test_create_job_for_submission(self):
submission: Submission = SubmissionFactory()
self.assertIsNone(submission.algorithm_job)
create_algorithm_job_for_submission(submission)
self.assertIsNotNone(submission.algorithm_job)
self.assertIsNone(submission.evaluation_job)
create_evaluation_job_for_submission(submission)
self.assertIsNotNone(submission.evaluation_job)
def test_submission_evaluation(client, evaluation_image, submission_file):
# Upload a submission and create a job
dockerclient = docker.DockerClient(base_url=settings.DOCKER_BASE_URL)
user = UserFactory()
submission = SubmissionFactory(file__from_path=submission_file,
creator=user)
eval_container, sha256 = evaluation_image
method = MethodFactory(image__from_path=eval_container,
image_sha256=sha256,
ready=True)
# We should not be able to download methods
response = client.get(method.image.url)
assert response.status_code == 403
job = JobFactory(submission=submission, method=method)
num_containers_before = len(dockerclient.containers.list())
num_volumes_before = len(dockerclient.volumes.list())
res = evaluate_submission(job=job)
# The evaluation method should return the correct answer
assert res["acc"] == 0.5
# The evaluation method should clean up after itself
assert len(dockerclient.volumes.list()) == num_volumes_before
assert len(dockerclient.containers.list()) == num_containers_before
# Try with a csv file
submission = SubmissionFactory(
file__from_path=Path(__file__).parent / 'resources' / 'submission.csv',
creator=user,
)
job = JobFactory(submission=submission, method=method)
res = evaluate_submission(job=job)
assert res["acc"] == 0.5
def test_submission_detail(client, TwoChallengeSets):
submission = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
validate_admin_only_view(
viewname="evaluation:submission-detail",
two_challenge_set=TwoChallengeSets,
reverse_kwargs={"pk": submission.pk},
client=client,
)
def test_evaluation_detail(client, eval_challenge_set):
submission = SubmissionFactory(
challenge=eval_challenge_set.challenge_set.challenge,
creator=eval_challenge_set.challenge_set.participant,
)
e = EvaluationFactory(submission=submission)
validate_open_view(
viewname="evaluation:detail",
challenge_set=eval_challenge_set.challenge_set,
reverse_kwargs={"pk": e.pk},
client=client,
)
def test_algorithm_submission_creates_one_job_per_test_set_image(self):
SubmissionFactory(
challenge=self.method.challenge,
algorithm_image=self.algorithm_image,
)
assert AlgorithmEvaluation.objects.count() == 2
assert [
inpt.image
for ae in AlgorithmEvaluation.objects.all()
for inpt in ae.inputs.all()
] == self.images[:2]
def test_result_detail(client, EvalChallengeSet):
submission = SubmissionFactory(
challenge=EvalChallengeSet.ChallengeSet.challenge,
creator=EvalChallengeSet.ChallengeSet.participant,
)
job = JobFactory(submission=submission)
result = ResultFactory(job=job)
validate_open_view(
viewname="evaluation:result-detail",
challenge_set=EvalChallengeSet.ChallengeSet,
reverse_kwargs={"pk": result.pk},
client=client,
)
def test_unsuccessful_jobs_fail_evaluation(self):
submission = SubmissionFactory()
evaluation = EvaluationFactory(submission=submission)
AlgorithmEvaluationFactory(status=AlgorithmEvaluation.SUCCESS,
submission=submission)
AlgorithmEvaluationFactory(status=AlgorithmEvaluation.FAILURE,
submission=submission)
set_evaluation_inputs(evaluation_pk=evaluation.pk)
evaluation.refresh_from_db()
assert evaluation.status == evaluation.FAILURE
assert (evaluation.output ==
"The algorithm failed to execute on 1 images.")
def test_invoice_get_rates_inconsistent_paidtask_rates(member, task_type,
task_type_name,
user_rate_attr_name,
store0):
USER_RATE = 0.5
PAID_TASK_RATE = 0.2
# Set some user rate
setattr(member, user_rate_attr_name, USER_RATE)
member.save()
month = timezone.make_aware(timezone.datetime(2014, 4, 1))
submission_kwargs = {
"store": store0,
"unit": store0.units[0],
"field": SubmissionFields.TARGET,
"type": SubmissionTypes.NORMAL,
"old_value": "foo",
"new_value": "bar",
"submitter": member,
"translation_project": store0.translation_project,
"creation_time": month,
}
scorelog_kwargs = {
"wordcount": 1,
"similarity": 0,
"action_code": TranslationActionCodes.NEW,
"creation_time": month,
"user": member,
"submission": SubmissionFactory(**submission_kwargs),
}
paid_task_kwargs = {
"rate": PAID_TASK_RATE, # Note how this doesn't match user's rate
"datetime": month,
"user": member,
"task_type": task_type,
}
ScoreLogFactory(**scorelog_kwargs)
PaidTaskFactory(**paid_task_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
with pytest.raises(ValueError) as e:
invoice.get_rates()
assert "Multiple %s rate values for user %s" % (
task_type_name,
member.username,
) in str(e.value)
def test_invoice_get_rates_inconsistent_paidtask_rates(member, task_type,
task_type_name,
user_rate_attr_name,
store0):
USER_RATE = 0.5
PAID_TASK_RATE = 0.2
# Set some user rate
setattr(member, user_rate_attr_name, USER_RATE)
member.save()
month = timezone.datetime(2014, 04, 01)
submission_kwargs = {
'store': store0,
'unit': store0.units[0],
'field': SubmissionFields.TARGET,
'type': SubmissionTypes.NORMAL,
'old_value': 'foo',
'new_value': 'bar',
'submitter': member,
'translation_project': store0.translation_project,
'creation_time': month,
}
scorelog_kwargs = {
'wordcount': 1,
'similarity': 0,
'action_code': TranslationActionCodes.NEW,
'creation_time': month,
'user': member,
'submission': SubmissionFactory(**submission_kwargs),
}
paid_task_kwargs = {
'rate': PAID_TASK_RATE, # Note how this doesn't match user's rate
'datetime': month,
'user': member,
'task_type': task_type,
}
ScoreLogFactory(**scorelog_kwargs)
PaidTaskFactory(**paid_task_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
with pytest.raises(ValueError) as e:
invoice.get_rates()
assert ('Multiple %s rate values for user %s' %
(task_type_name, member.username) in e.value.message)
def test_set_evaluation_inputs(self):
submission = SubmissionFactory()
evaluation = EvaluationFactory(submission=submission)
algorithms = AlgorithmEvaluationFactory.create_batch(
2, status=AlgorithmEvaluation.SUCCESS, submission=submission)
civs = ComponentInterfaceValueFactory.create_batch(2)
for alg, civ in zip(algorithms, civs):
alg.outputs.set([civ])
set_evaluation_inputs(evaluation_pk=evaluation.pk)
evaluation.refresh_from_db()
assert evaluation.status == evaluation.PENDING
assert evaluation.output == ""
assert list(evaluation.inputs.all()) == civs
def test_job_detail(client, TwoChallengeSets):
method = MethodFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.admin,
ready=True,
)
submission = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
job = JobFactory(method=method, submission=submission)
validate_admin_only_view(
viewname="evaluation:job-detail",
two_challenge_set=TwoChallengeSets,
reverse_kwargs={"pk": job.pk},
client=client,
)
def test_record_submission(member, submission_type, store0):
unit = store0.units.first()
submission_params = {
"store": store0,
"unit": unit,
"field": SubmissionFields.TARGET,
"type": submission_type,
"old_value": unit.target,
"new_value": "New target",
"similarity": 0,
"mt_similarity": 0,
"submitter": member,
"translation_project": store0.translation_project,
"creation_time": timezone.now(),
}
sub = SubmissionFactory(**submission_params)
assert ScoreLog.objects.filter(submission=sub).count() == 1
def test_invoice_get_rates_scorelog_rates(member, task_type, task_type_name,
user_rate_attr_name, store0):
"""Tests that `Invoice.get_rates()` returns the rates set for users in their
`ScoreLog` entries.
"""
USER_RATE_ONE = 0.5
USER_RATE_TWO = 0.2
# Set some user rate
setattr(member, user_rate_attr_name, USER_RATE_ONE)
member.save()
month = timezone.datetime(2014, 04, 01)
submission_kwargs = {
'store': store0,
'unit': store0.units[0],
'field': SubmissionFields.TARGET,
'type': SubmissionTypes.NORMAL,
'old_value': 'foo',
'new_value': 'bar',
'submitter': member,
'translation_project': store0.translation_project,
'creation_time': month,
}
scorelog_kwargs = {
'wordcount': 1,
'similarity': 0,
'action_code': TranslationActionCodes.NEW,
'creation_time': month,
'user': member,
'submission': SubmissionFactory(**submission_kwargs),
}
ScoreLogFactory(**scorelog_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
# Set user rate to something else to ensure we get the recorded rates
setattr(member, user_rate_attr_name, USER_RATE_TWO)
member.save()
rate, review_rate, hourly_rate = invoice.get_rates()
assert locals()[user_rate_attr_name] == USER_RATE_ONE
def test_invoice_get_rates_scorelog_rates(member, task_type, task_type_name,
user_rate_attr_name, store0):
"""Tests that `Invoice.get_rates()` returns the rates set for users in their
`ScoreLog` entries.
"""
USER_RATE_ONE = 0.5
USER_RATE_TWO = 0.2
# Set some user rate
setattr(member, user_rate_attr_name, USER_RATE_ONE)
member.save()
month = timezone.make_aware(timezone.datetime(2014, 4, 1))
submission_kwargs = {
"store": store0,
"unit": store0.units[0],
"field": SubmissionFields.TARGET,
"type": SubmissionTypes.NORMAL,
"old_value": "foo",
"new_value": "bar",
"submitter": member,
"translation_project": store0.translation_project,
"creation_time": month,
}
scorelog_kwargs = {
"wordcount": 1,
"similarity": 0,
"action_code": TranslationActionCodes.NEW,
"creation_time": month,
"user": member,
"submission": SubmissionFactory(**submission_kwargs),
}
ScoreLogFactory(**scorelog_kwargs)
invoice = Invoice(member, FAKE_CONFIG, month=month)
# Set user rate to something else to ensure we get the recorded rates
setattr(member, user_rate_attr_name, USER_RATE_TWO)
member.save()
rate, review_rate, hourly_rate = invoice.get_rates()
assert locals()[user_rate_attr_name] == USER_RATE_ONE
def test_submission_time_limit(client, TwoChallengeSets):
SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
def get_submission_view():
return get_view_for_user(
viewname="evaluation:submission-create",
challenge=TwoChallengeSets.ChallengeSet1.challenge,
client=client,
user=TwoChallengeSets.ChallengeSet1.participant,
)
assert "make 9 more" in get_submission_view().rendered_content
s = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
s.created = timezone.now() - timedelta(hours=23)
s.save()
assert "make 8 more" in get_submission_view().rendered_content
s = SubmissionFactory(
challenge=TwoChallengeSets.ChallengeSet1.challenge,
creator=TwoChallengeSets.ChallengeSet1.participant,
)
s.created = timezone.now() - timedelta(hours=25)
s.save()
assert "make 8 more" in get_submission_view().rendered_content
def submission_and_job(*, challenge, creator):
""" Creates a submission and a job for that submission """
s = SubmissionFactory(challenge=challenge, creator=creator)
j = JobFactory(submission=s)
return s, j
def submission_and_evaluation(*, challenge, creator):
"""Creates a submission and an evaluation for that submission."""
s = SubmissionFactory(challenge=challenge, creator=creator)
e = EvaluationFactory(submission=s)
return s, e
