def save_predictions(db_file: Union[str, Path],
model_version_id: int,
gbd_round_id: int,
out_dir: Path,
locations: Optional[List[int]] = None,
sexes: Optional[List[int]] = None,
sample: bool = False,
predictions: Optional[pd.DataFrame] = None) -> None:
"""
Save the fit from this dismod database for a specific location and sex to be
uploaded later on.
"""
LOG.info("Extracting results from DisMod SQLite Database.")
da = DismodExtractor(path=db_file)
predictions = da.format_predictions_for_ihme(locations=locations,
sexes=sexes,
gbd_round_id=gbd_round_id,
samples=sample,
predictions=predictions)
LOG.info(f"Saving the results to {out_dir}.")
rh = ResultsHandler()
rh.save_draw_files(df=predictions,
directory=out_dir,
add_summaries=True,
model_version_id=model_version_id)
def main():
"""
Takes a dismod database that has had predict run on it and converts the predictions
into the format needed for the IHME Epi Databases. Also uploads inputs to tier 3 which
allows us to view those inputs in EpiViz.
"""
args = get_args()
logging.basicConfig(level=LEVELS[args.loglevel])
context = Context(model_version_id=args.model_version_id)
inputs, alchemy, settings = context.read_inputs()
if not inputs.csmr.raw.empty:
LOG.info("Uploading CSMR to t3")
inputs.csmr.attach_to_model_version_in_db(
model_version_id=args.model_version_id,
conn_def=context.model_connection)
LOG.info("Extracting results from DisMod SQLite Database.")
dismod_file = context.db_file(location_id=args.parent_location_id,
sex_id=args.sex_id,
make=False)
da = DismodExtractor(path=dismod_file)
predictions = da.format_predictions_for_ihme()
LOG.info("Saving the results.")
rh = ResultsHandler(model_version_id=args.model_version_id)
rh.save_draw_files(df=predictions, directory=context.draw_dir)
rh.upload_summaries(directory=context.draw_dir,
conn_def=context.model_connection)
def test_format_fit(ihme, dismod, df):
d = DismodExtractor(path=Path('temp.db'))
pred = d.format_predictions_for_ihme(gbd_round_id=6, samples=False)
# This prediction data frame is longer for each demographic group than the prediction
# dataframe in `test_get_predictions` because there is an extra measure.
assert len(pred) == 528
assert all(pred.columns == [
'location_id', 'year_id', 'age_group_id', 'sex_id', 'measure_id', 'mean'
])
def test_format_prior(mi, settings, dismod):
d = DismodExtractor(path=NAME)
pred = d.format_predictions_for_ihme(locations=[72],
sexes=[2],
gbd_round_id=6,
samples=True)
assert all(pred.columns == [
'location_id', 'year_id', 'age_group_id', 'sex_id', 'measure_id',
'draw_0', 'draw_1'
])
def test_default_gather_child_draws(mi, settings, dismod):
de = DismodExtractor(NAME)
draws = de.gather_draws_for_prior_grid(location_id=72,
sex_id=2,
rates=['iota', 'chi', 'pini'])
for rate in ['iota', 'chi', 'pini']:
assert rate in draws
assert draws[rate]['value'].shape[-1] == 2
assert 'dage' not in draws[rate].keys()
assert 'dtime' not in draws[rate].keys()
def test_get_predictions(ihme, dismod, df):
d = DismodExtractor(path=Path('temp.db'))
pred = d._extract_raw_predictions()
assert len(pred) == 484
assert all(pred.columns == [
'predict_id', 'sample_index', 'avgint_id', 'avg_integrand',
'integrand_id', 'node_id', 'weight_id', 'subgroup_id', 'age_lower',
'age_upper', 'time_lower', 'time_upper', 'c_age_group_id',
'c_location_id', 'c_sex_id', 'c_year_id', 'x_0', 'x_1', 'x_2',
'integrand_name', 'minimum_meas_cv', 'rate'
])
def test_format_for_ihme(ihme, dismod):
d = DismodExtractor(path=Path('temp.db'))
pred = d.format_predictions_for_ihme()
assert len(pred) == 36
assert all(pred.columns == [
'location_id', 'age_group_id', 'year_id', 'sex_id', 'measure_id',
'mean', 'upper', 'lower'
])
assert all(pred.location_id == np.tile(list(range(70, 73)), 12))
assert all(pred.sex_id == 2)
assert all(pred.age_group_id == 2)
assert all(pred.year_id == 1990)
def test_gather_child_draws(mi, settings, dismod):
de = DismodExtractor(NAME)
draws = de.gather_draws_for_prior_grid(location_id=72,
sex_id=2,
rates=['iota', 'chi', 'pini'],
value=True,
dage=True,
dtime=True)
for rate in ['iota', 'chi', 'pini']:
assert rate in draws
assert draws[rate]['value'].shape[-1] == 2
# pini will not have any dage or dtime draws because it only has one age and time
for rate in ['iota', 'chi']:
assert draws[rate]['dage'].shape[-1] == 2
assert draws[rate]['dtime'].shape[-1] == 2
def test_get_predictions(ihme, dismod):
d = DismodExtractor(path=Path('temp.db'))
pred = d.get_predictions()
assert len(pred) == 33
assert all(pred.columns == [
'predict_id', 'sample_index', 'avgint_id', 'avg_integrand',
'integrand_id', 'node_id', 'weight_id', 'subgroup_id', 'age_lower',
'age_upper', 'time_lower', 'time_upper', 'c_age_group_id',
'c_location_id', 'c_sex_id', 'c_year_id', 'x_0', 'x_1', 'x_2',
'integrand_name', 'minimum_meas_cv', 'rate'
])
assert all(pred.age_lower == 0.0)
assert all(pred.age_upper == 0.01917808)
assert all(pred.integrand_id == np.repeat(list(range(0, 11)), 3))
assert all(pred.time_lower == 1990.)
assert all(pred.time_upper == 1991.)
assert all(pred.c_age_group_id == 2)
assert all(pred.c_sex_id == 2)
def get_prior(path: Union[str, Path],
location_id: int,
sex_id: int,
rates: List[str],
samples: bool = True) -> Dict[str, Dict[str, np.ndarray]]:
"""
Gets priors from a path to a database for a given location ID and sex ID.
"""
child_prior = DismodExtractor(path=path).gather_draws_for_prior_grid(
location_id=location_id, sex_id=sex_id, rates=rates, samples=samples)
return child_prior
def main(args=None):
"""
Creates a dismod database using the saved inputs and the file
structure specified in the context.
Then runs an optional set of commands on the database passed
in the --commands argument.
Also passes an optional argument --options as a dictionary to
the dismod database to fill/modify the options table.
"""
args = get_args(args=args)
logging.basicConfig(level=LEVELS[args.loglevel])
if args.test_dir:
context = Context(model_version_id=args.model_version_id,
configure_application=False,
root_directory=args.test_dir)
else:
context = Context(model_version_id=args.model_version_id)
inputs, alchemy, settings = context.read_inputs()
# If we want to override the rate priors with posteriors from a previous
# database, pass them in here.
if args.prior_parent or args.prior_sex:
if not (args.prior_parent and args.prior_sex):
raise RuntimeError("Need to pass both prior parent and sex or neither.")
child_prior = DismodExtractor(path=context.db_file(
location_id=args.prior_parent,
sex_id=args.prior_sex
)).gather_draws_for_prior_grid(
location_id=args.parent_location_id,
sex_id=args.sex_id,
rates=[r.rate for r in settings.rate]
)
else:
child_prior = None
df = DismodFiller(
path=context.db_file(location_id=args.parent_location_id, sex_id=args.sex_id),
settings_configuration=settings,
measurement_inputs=inputs,
grid_alchemy=alchemy,
parent_location_id=args.parent_location_id,
sex_id=args.sex_id,
child_prior=child_prior
)
df.fill_for_parent_child(**args.options)
run_dismod_commands(dm_file=df.path.absolute(), commands=args.commands)
def test_empty_database():
with pytest.raises(DismodExtractorError):
DismodExtractor('temp-2.db')
