def valid_expression_predictor_data():
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import RealDescriptor
shear_modulus = RealDescriptor('Property~Shear modulus',
lower_bound=0,
upper_bound=100,
units='GPa')
youngs_modulus = RealDescriptor('Property~Young\'s modulus',
lower_bound=0,
upper_bound=100,
units='GPa')
poissons_ratio = RealDescriptor('Property~Poisson\'s ratio',
lower_bound=-1,
upper_bound=0.5,
units='')
return dict(
module_type='PREDICTOR',
status='VALID',
status_info=[],
archived=False,
display_name='Expression predictor',
schema_id='f1601161-bb98-4fa9-bdd2-a2a673547532',
id=str(uuid.uuid4()),
config=dict(
type='AnalyticExpression',
name='Expression predictor',
description=
'Computes shear modulus from Youngs modulus and Poissons ratio',
expression='Y / (2 * (1 + v))',
output=shear_modulus.dump(),
aliases={
'Y': youngs_modulus.dump(),
'v': poissons_ratio.dump(),
}))
def valid_simple_ml_predictor_data():
"""Produce valid data used for tests."""
from citrine.informatics.data_sources import GemTableDataSource
from citrine.informatics.descriptors import RealDescriptor
x = RealDescriptor("x", 0, 100, "")
y = RealDescriptor("y", 0, 100, "")
z = RealDescriptor("z", 0, 100, "")
data_source = GemTableDataSource(
table_id=uuid.UUID('e5c51369-8e71-4ec6-b027-1f92bdc14762'),
table_version=2)
return dict(
module_type='PREDICTOR',
status='VALID',
status_info=[],
archived=False,
display_name='ML predictor',
schema_id='08d20e5f-e329-4de0-a90a-4b5e36b91703',
id=str(uuid.uuid4()),
config=dict(
type='Simple',
name='ML predictor',
description='Predicts z from input x and latent variable y',
inputs=[x.dump()],
outputs=[z.dump()],
latent_variables=[y.dump()],
training_data=[data_source.dump()]))
def valid_predictor_report_data():
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import RealDescriptor
x = RealDescriptor("x", 0, 1, "")
y = RealDescriptor("y", 0, 100, "")
z = RealDescriptor("z", 0, 101, "")
return dict(
id='7c2dda5d-675a-41b6-829c-e485163f0e43',
module_id='31c7f311-6f3d-4a93-9387-94cc877f170c',
status='OK',
create_time='2020-04-23T15:46:26Z',
update_time='2020-04-23T15:46:26Z',
report=dict(models=[
dict(name='GeneralLoloModel_1',
type='ML Model',
inputs=[x.key],
outputs=[y.key],
display_name='ML Model',
model_settings=[
dict(name='Algorithm',
value='Ensemble of non-linear estimators',
children=[
dict(name='Number of estimators',
value=64,
children=[]),
dict(name='Leaf model',
value='Mean',
children=[]),
dict(name='Use jackknife',
value=True,
children=[])
])
],
feature_importances=[
dict(response_key='y',
importances=dict(x=1.00),
top_features=5)
],
predictor_configuration_name="Predict y from x with ML"),
dict(name='GeneralLosslessModel_2',
type='Analytic Model',
inputs=[x.key, y.key],
outputs=[z.key],
display_name='GeneralLosslessModel_2',
model_settings=[
dict(name="Expression",
value="(z) <- (x + y)",
children=[])
],
feature_importances=[],
predictor_configuration_name="Expression for z",
predictor_configuration_uid=
"249bf32c-6f3d-4a93-9387-94cc877f170c")
],
descriptors=[x.dump(), y.dump(),
z.dump()]))
def valid_deprecated_expression_predictor_data():
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import RealDescriptor
shear_modulus = RealDescriptor('Property~Shear modulus',
lower_bound=0,
upper_bound=100,
units='GPa')
return dict(
module_type='PREDICTOR',
status='VALID',
status_info=[],
archived=False,
display_name='Expression predictor',
schema_id='866e72a6-0a01-4c5f-8c35-146eb2540166',
id=str(uuid.uuid4()),
config=dict(
type='Expression',
name='Expression predictor',
description=
'Computes shear modulus from Youngs modulus and Poissons ratio',
expression='Y / (2 * (1 + v))',
output=shear_modulus.dump(),
aliases={
'Y': "Property~Young's modulus",
'v': "Property~Poisson's ratio",
}))
def valid_mean_property_predictor_data():
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import FormulationDescriptor, RealDescriptor
from citrine.informatics.data_sources import GemTableDataSource
formulation_descriptor = FormulationDescriptor('simple mixture')
density = RealDescriptor(key='density',
lower_bound=0,
upper_bound=100,
units='g/cm^3')
return dict(module_type='PREDICTOR',
status='READY',
status_info=[],
archived=False,
display_name='Mean property predictor',
id=str(uuid.uuid4()),
config=dict(
type='MeanProperty',
name='Mean property predictor',
description='Computes mean ingredient properties',
input=formulation_descriptor.dump(),
properties=[density.dump()],
p=2,
training_data=[
GemTableDataSource(uuid.uuid4(), 0,
formulation_descriptor).dump()
],
impute_properties=True,
default_properties={'density': 1.0},
label='solvent'))
def invalid_predictor_data():
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import RealDescriptor
x = RealDescriptor("x", 0, 100, "")
y = RealDescriptor("y", 0, 100, "")
z = RealDescriptor("z", 0, 100, "")
return dict(module_type='PREDICTOR',
status='INVALID',
status_info=['Something is wrong', 'Very wrong'],
archived=True,
display_name='my predictor',
id=str(uuid.uuid4()),
config=dict(type='invalid',
name='my predictor',
description='does some things',
inputs=[x.dump(), y.dump()],
output=z.dump()))
def old_auto_ml_predictor_data(valid_gem_data_source_dict):
"""Produce valid data used for tests."""
from citrine.informatics.descriptors import RealDescriptor
x = RealDescriptor("x", 0, 100, "")
y = RealDescriptor("y", 0, 100, "")
z = RealDescriptor("z", 0, 100, "")
return dict(module_type='PREDICTOR',
status='READY',
status_info=[],
archived=False,
display_name='AutoML predictor',
id=str(uuid.uuid4()),
config=dict(type='AutoML',
name='AutoML predictor',
description='Predicts z from input x',
inputs=[x.dump()],
responses=[z.dump()],
training_data=[valid_gem_data_source_dict]))
def test_bad_predictor_report_build(valid_predictor_report_data):
"""Modify the predictor report to be non-ideal and check the behavior."""
too_many_descriptors = deepcopy(valid_predictor_report_data)
# Multiple descriptors with the same key
other_x = RealDescriptor("x", 0, 100, "")
too_many_descriptors['report']['descriptors'].append(other_x.dump())
with warnings.catch_warnings(record=True) as w:
Report.build(too_many_descriptors)
assert len(w) == 1
assert issubclass(w[-1].category, RuntimeWarning)
# A key that appears in inputs and/or outputs, but there is no corresponding descriptor.
# This is done twice for coverage, once to catch a missing input and once for a missing output.
too_few_descriptors = deepcopy(valid_predictor_report_data)
too_few_descriptors['report']['descriptors'].pop()
with pytest.raises(RuntimeError):
Report.build(too_few_descriptors)
too_few_descriptors['report']['descriptors'] = []
with pytest.raises(RuntimeError):
Report.build(too_few_descriptors)