def test_GPKG(self):
"""Tests GPKG instantiation."""
exp = get_branin_experiment(with_batch=True)
with self.assertRaises(ValueError):
get_GPKG(experiment=exp, data=exp.fetch_data())
exp.trials[0].run()
gpkg = get_GPKG(experiment=exp, data=exp.fetch_data())
self.assertIsInstance(gpkg, TorchModelBridge)
# test transform_configs with winsorization
configs = {
"Winsorize": {
"winsorization_lower": 0.1,
"winsorization_upper": 0.1
}
}
gpkg_win = get_GPKG(experiment=exp,
data=exp.fetch_data(),
transform_configs=configs)
self.assertIsInstance(gpkg_win, TorchModelBridge)
self.assertEqual(gpkg_win._transform_configs, configs)
# test multi-fidelity optimization
exp.parameters["x2"] = RangeParameter(
name="x2",
parameter_type=exp.parameters["x2"].parameter_type,
lower=-5.0,
upper=10.0,
is_fidelity=True,
target_value=10.0,
)
gpkg_mf = get_GPKG(experiment=exp, data=exp.fetch_data())
self.assertIsInstance(gpkg_mf, TorchModelBridge)
def test_GPKG(self):
"""Tests GPKG instantiation."""
exp = get_branin_experiment(with_batch=True)
with self.assertRaises(ValueError):
get_GPKG(experiment=exp, data=exp.fetch_data())
exp.trials[0].run()
gpkg = get_GPKG(experiment=exp, data=exp.fetch_data())
self.assertIsInstance(gpkg, TorchModelBridge)
gpkg_win = get_GPKG(experiment=exp,
data=exp.fetch_data(),
winsorization_limits=[0.1, 0.1])
self.assertIsInstance(gpkg_win, TorchModelBridge)
configs_expected = {
"Winsorize": {
"winsorization_lower": 0.1,
"winsorization_upper": 0.1
}
}
self.assertEqual(gpkg_win._transform_configs, configs_expected)
def test_GPKG(self):
"""Tests GPKG instantiation."""
exp = get_branin_experiment(with_batch=True)
with self.assertRaises(ValueError):
get_GPKG(experiment=exp, data=exp.fetch_data())
exp.trials[0].run().mark_completed()
gpkg = get_GPKG(experiment=exp, data=exp.fetch_data())
self.assertIsInstance(gpkg, TorchModelBridge)
# Check that .gen returns without failure
gr = gpkg.gen(n=1)
self.assertEqual(len(gr.arms), 1)
# test transform_configs with winsorization
configs = {
"Winsorize": {
"winsorization_config": WinsorizationConfig(
lower_quantile_margin=0.1,
upper_quantile_margin=0.1,
)
}
}
gpkg_win = get_GPKG(
experiment=exp, data=exp.fetch_data(), transform_configs=configs
)
self.assertIsInstance(gpkg_win, TorchModelBridge)
self.assertEqual(gpkg_win._transform_configs, configs)
# test multi-fidelity optimization
exp.parameters["x2"] = RangeParameter(
name="x2",
parameter_type=exp.parameters["x2"].parameter_type,
lower=-5.0,
upper=10.0,
is_fidelity=True,
target_value=10.0,
)
gpkg_mf = get_GPKG(experiment=exp, data=exp.fetch_data())
self.assertIsInstance(gpkg_mf, TorchModelBridge)