def test_example_runs_with_non_default_optimizer_kwargs(self):
"""Simple integration test for example with non default kwargs."""
run_example(
num_to_sample=1,
verbose=False,
testapp=self.testapp,
gp_next_points_kwargs={
'optimizer_info': {
'num_multistarts':
TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples':
TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters':
TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
}
},
gp_hyper_opt_kwargs={
'optimizer_info': {
'num_multistarts':
TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples':
TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters':
TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
}
},
gp_mean_var_kwargs={},
rest_port=1337,
)
def _build_json_payload(domain, covariance, historical_data):
"""Create a json_payload to POST to the /gp/hyper_opt endpoint with all needed info."""
hyper_dim = domain.dim + 1
dict_to_dump = {
'gp_historical_info': historical_data.json_payload(),
'covariance_info': covariance.get_json_serializable_info(),
'domain_info': domain.get_json_serializable_info(minimal=True),
'optimizer_info': {
'optimizer_type': GRADIENT_DESCENT_OPTIMIZER,
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
'hyperparameter_domain_info': {
'dim': hyper_dim,
'domain_type': 'tensor_product',
'domain_bounds': [],
},
}
for _ in range(hyper_dim):
dict_to_dump['hyperparameter_domain_info']['domain_bounds'].append({
'min': 0.1,
'max': 2.0,
})
return json.dumps(dict_to_dump)
def _build_json_payload(domain, covariance, historical_data):
"""Create a json_payload to POST to the /gp/hyper_opt endpoint with all needed info."""
hyper_dim = domain.dim + 1
dict_to_dump = {
'gp_historical_info': historical_data.json_payload(),
'covariance_info': covariance.get_json_serializable_info(),
'domain_info': domain.get_json_serializable_info(minimal=True),
'optimizer_info': {
'optimizer_type':
GRADIENT_DESCENT_OPTIMIZER,
'num_multistarts':
TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples':
TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters':
dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
'hyperparameter_domain_info': {
'dim': hyper_dim,
'domain_type': 'tensor_product',
'domain_bounds': [],
},
}
for _ in range(hyper_dim):
dict_to_dump['hyperparameter_domain_info']['domain_bounds'].append(
{
'min': 0.1,
'max': 2.0,
})
return json.dumps(dict_to_dump)
def test_example_runs(self):
"""Simple integration test for example."""
run_example(
verbose=False,
testapp=self.testapp,
optimizer_info={
'optimizer_type': GRADIENT_DESCENT_OPTIMIZER,
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
})
def test_example_runs(self):
"""Simple integration test for example."""
run_example(
num_points_to_sample=1,
verbose=False,
testapp=self.testapp,
optimizer_info={
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
},
rest_port=1337,
)
def test_example_runs_with_non_default_optimizer_kwargs(self):
"""Simple integration test for example with non default kwargs."""
run_example(
num_to_sample=1,
verbose=False,
testapp=self.testapp,
gp_next_points_kwargs={
'optimizer_info': {
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
}
},
gp_hyper_opt_kwargs={
'optimizer_info': {
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': TEST_GRADIENT_DESCENT_PARAMETERS._asdict(),
}
},
gp_mean_var_kwargs={},
rest_port=1337,
)
def test_optimizer_params_passed_through(self):
"""Test that the optimizer parameters get passed through to the endpoint."""
# TODO(GH-305): turn this into a unit test by going through OptimizableGpPrettyView
# and mocking out dependencies (instead of awkwardly constructing a more complex object).
test_case = self.gp_test_environments[0]
num_to_sample = 1
python_domain, python_gp = test_case
python_cov, historical_data = python_gp.get_core_data_copy()
# Test default test parameters get passed through
json_payload = json.loads(
self._build_json_payload(python_domain, python_cov,
historical_data, num_to_sample))
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpNextPointsPrettyView(request)
# get_params_from_request() requires this field is set. value is arbitrary for now.
# TODO(GH-305): mock out this and other members
view._route_name = GP_NEXT_POINTS_CONSTANT_LIAR_ROUTE_NAME
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(
params)
test_param_dict = TEST_GRADIENT_DESCENT_PARAMETERS._asdict()
test_param_dict['num_multistarts'] = TEST_OPTIMIZER_MULTISTARTS
assert optimizer_parameters._get_member_dict() == test_param_dict
# Test arbitrary parameters get passed through
for i, key in enumerate(test_param_dict.iterkeys()):
test_param_dict[key] /= 2
test_num_multistarts = test_param_dict.pop('num_multistarts')
json_payload['optimizer_info'][
'num_multistarts'] = test_num_multistarts
json_payload['optimizer_info'][
'optimizer_parameters'] = test_param_dict
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpNextPointsPrettyView(request)
# get_params_from_request() requires this field is set. value is arbitrary for now.
# TODO(GH-305): mock out this and other members
view._route_name = GP_NEXT_POINTS_CONSTANT_LIAR_ROUTE_NAME
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(
params)
test_param_dict['num_multistarts'] = test_num_multistarts
assert optimizer_parameters._get_member_dict() == test_param_dict
def _build_json_payload(self,
domain,
covariance,
historical_data,
num_to_sample,
lie_value=None,
lie_method=None,
l_bfgs_b=False):
"""Create a json_payload to POST to the /gp/next_points/* endpoint with all needed info."""
if l_bfgs_b:
dict_to_dump = {
'num_to_sample': num_to_sample,
'mc_iterations': TEST_EXPECTED_IMPROVEMENT_MC_ITERATIONS,
'gp_historical_info': historical_data.json_payload(),
'covariance_info': covariance.get_json_serializable_info(),
'domain_info': domain.get_json_serializable_info(),
'optimizer_info': {
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters':
dict(TEST_LBFGSB_PARAMETERS._asdict()),
},
'mvndst_parameters': {
'releps': 1.0,
'maxpts_per_dim': 200,
},
}
else:
dict_to_dump = {
'num_to_sample': num_to_sample,
'mc_iterations': TEST_EXPECTED_IMPROVEMENT_MC_ITERATIONS,
'gp_historical_info': historical_data.json_payload(),
'covariance_info': covariance.get_json_serializable_info(),
'domain_info': domain.get_json_serializable_info(),
'optimizer_info': {
'num_multistarts':
TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples':
TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters':
dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
}
if lie_value is not None:
dict_to_dump['lie_value'] = lie_value
if lie_method is not None:
dict_to_dump['lie_method'] = lie_method
return json.dumps(dict_to_dump)
def test_optimizer_params_passed_through(self):
"""Test that the optimizer parameters get passed through to the endpoint."""
# TODO(GH-305): turn this into a unit test by going through OptimizableGpPrettyView
# and mocking out dependencies (instead of awkwardly constructing a more complex object).
test_case = self.gp_test_environments[0]
num_to_sample = 1
python_domain, python_gp = test_case
python_cov, historical_data = python_gp.get_core_data_copy()
# Test default test parameters get passed through
json_payload = json.loads(self._build_json_payload(python_domain, python_cov, historical_data, num_to_sample))
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpNextPointsPrettyView(request)
# get_params_from_request() requires this field is set. value is arbitrary for now.
# TODO(GH-305): mock out this and other members
view._route_name = GP_NEXT_POINTS_CONSTANT_LIAR_ROUTE_NAME
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(params)
test_param_dict = TEST_GRADIENT_DESCENT_PARAMETERS._asdict()
test_param_dict["num_multistarts"] = TEST_OPTIMIZER_MULTISTARTS
assert optimizer_parameters._get_member_dict() == test_param_dict
# Test arbitrary parameters get passed through
for i, key in enumerate(test_param_dict.iterkeys()):
test_param_dict[key] /= 2
test_num_multistarts = test_param_dict.pop("num_multistarts")
json_payload["optimizer_info"]["num_multistarts"] = test_num_multistarts
json_payload["optimizer_info"]["optimizer_parameters"] = test_param_dict
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpNextPointsPrettyView(request)
# get_params_from_request() requires this field is set. value is arbitrary for now.
# TODO(GH-305): mock out this and other members
view._route_name = GP_NEXT_POINTS_CONSTANT_LIAR_ROUTE_NAME
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(params)
test_param_dict["num_multistarts"] = test_num_multistarts
assert optimizer_parameters._get_member_dict() == test_param_dict
def _build_json_payload(self, domain, covariance, historical_data):
"""Create a json_payload to POST to the /gp/hyper_opt endpoint with all needed info."""
hyper_dim = domain.dim + 1
dict_to_dump = {
"gp_historical_info": historical_data.json_payload(),
"covariance_info": covariance.get_json_serializable_info(),
"domain_info": domain.get_json_serializable_info(minimal=True),
"optimizer_info": {
"optimizer_type": GRADIENT_DESCENT_OPTIMIZER,
"num_multistarts": TEST_OPTIMIZER_MULTISTARTS,
"num_random_samples": TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
"optimizer_parameters": dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
"hyperparameter_domain_info": {"dim": hyper_dim, "domain_type": "tensor_product", "domain_bounds": []},
}
for _ in range(hyper_dim):
dict_to_dump["hyperparameter_domain_info"]["domain_bounds"].append({"min": 0.1, "max": 2.0})
return json.dumps(dict_to_dump)
def test_optimizer_params_passed_through(self):
"""Test that the optimizer parameters get passed through to the endpoint."""
test_case = self.gp_test_environments[0]
python_domain, python_gp = test_case
python_cov, historical_data = python_gp.get_core_data_copy()
# Test default test parameters get passed through
json_payload = json.loads(
self._build_json_payload(python_domain, python_cov,
historical_data))
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpHyperOptView(request)
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(
params)
test_param_dict = TEST_GRADIENT_DESCENT_PARAMETERS._asdict()
test_param_dict['num_multistarts'] = TEST_OPTIMIZER_MULTISTARTS
assert optimizer_parameters._get_member_dict() == test_param_dict
# Test arbitrary parameters get passed through
for i, key in enumerate(iter(test_param_dict.keys())):
test_param_dict[key] /= 2
test_num_multistarts = test_param_dict.pop('num_multistarts')
json_payload['optimizer_info'][
'num_multistarts'] = test_num_multistarts
json_payload['optimizer_info'][
'optimizer_parameters'] = test_param_dict
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpHyperOptView(request)
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(
params)
test_param_dict['num_multistarts'] = test_num_multistarts
assert optimizer_parameters._get_member_dict() == test_param_dict
def _build_json_payload(self, domain, covariance, historical_data, num_to_sample, lie_value=None, lie_method=None):
"""Create a json_payload to POST to the /gp/next_points/* endpoint with all needed info."""
dict_to_dump = {
'num_to_sample': num_to_sample,
'mc_iterations': TEST_EXPECTED_IMPROVEMENT_MC_ITERATIONS,
'gp_historical_info': historical_data.json_payload(),
'covariance_info': covariance.get_json_serializable_info(),
'domain_info': domain.get_json_serializable_info(),
'optimizer_info': {
'num_multistarts': TEST_OPTIMIZER_MULTISTARTS,
'num_random_samples': TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
'optimizer_parameters': dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
}
if lie_value is not None:
dict_to_dump['lie_value'] = lie_value
if lie_method is not None:
dict_to_dump['lie_method'] = lie_method
return json.dumps(dict_to_dump)
def _build_json_payload(
domain, covariance, historical_data, num_to_sample, lie_value=None, lie_method=None, l_bfgs_b=False
):
"""Create a json_payload to POST to the /gp/next_points/* endpoint with all needed info."""
if l_bfgs_b:
dict_to_dump = {
"num_to_sample": num_to_sample,
"mc_iterations": TEST_EXPECTED_IMPROVEMENT_MC_ITERATIONS,
"gp_historical_info": historical_data.json_payload(),
"covariance_info": covariance.get_json_serializable_info(),
"domain_info": domain.get_json_serializable_info(),
"optimizer_info": {
"num_multistarts": TEST_OPTIMIZER_MULTISTARTS,
"num_random_samples": TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
"optimizer_parameters": dict(TEST_LBFGSB_PARAMETERS._asdict()),
},
"mvndst_parameters": {"releps": 1.0, "maxpts_per_dim": 200},
}
else:
dict_to_dump = {
"num_to_sample": num_to_sample,
"mc_iterations": TEST_EXPECTED_IMPROVEMENT_MC_ITERATIONS,
"gp_historical_info": historical_data.json_payload(),
"covariance_info": covariance.get_json_serializable_info(),
"domain_info": domain.get_json_serializable_info(),
"optimizer_info": {
"num_multistarts": TEST_OPTIMIZER_MULTISTARTS,
"num_random_samples": TEST_OPTIMIZER_NUM_RANDOM_SAMPLES,
"optimizer_parameters": dict(TEST_GRADIENT_DESCENT_PARAMETERS._asdict()),
},
}
if lie_value is not None:
dict_to_dump["lie_value"] = lie_value
if lie_method is not None:
dict_to_dump["lie_method"] = lie_method
return json.dumps(dict_to_dump)
def test_optimizer_params_passed_through(self):
"""Test that the optimizer parameters get passed through to the endpoint."""
test_case = self.gp_test_environments[0]
python_domain, python_gp = test_case
python_cov, historical_data = python_gp.get_core_data_copy()
# Test default test parameters get passed through
json_payload = json.loads(self._build_json_payload(python_domain, python_cov, historical_data))
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpHyperOptView(request)
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(params)
test_param_dict = TEST_GRADIENT_DESCENT_PARAMETERS._asdict()
test_param_dict['num_multistarts'] = TEST_OPTIMIZER_MULTISTARTS
assert optimizer_parameters._get_member_dict() == test_param_dict
# Test arbitrary parameters get passed through
for i, key in enumerate(test_param_dict.iterkeys()):
test_param_dict[key] /= 2
test_num_multistarts = test_param_dict.pop('num_multistarts')
json_payload['optimizer_info']['num_multistarts'] = test_num_multistarts
json_payload['optimizer_info']['optimizer_parameters'] = test_param_dict
request = pyramid.testing.DummyRequest(post=json_payload)
request.json_body = json_payload
view = GpHyperOptView(request)
params = view.get_params_from_request()
_, optimizer_parameters, num_random_samples = _make_optimizer_parameters_from_params(params)
test_param_dict['num_multistarts'] = test_num_multistarts
assert optimizer_parameters._get_member_dict() == test_param_dict
