def test_sparse_variational_optimize_with_defaults() -> None:
x_observed = np.linspace(0, 100, 100).reshape((-1, 1))
y_observed = _3x_plus_gaussian_noise(x_observed)
data = x_observed, y_observed
dataset = Dataset(*data)
optimizer = create_optimizer(tf.optimizers.Adam(), dict(max_iter=20))
model = SparseVariational(_svgp(x_observed[:10]), dataset, optimizer=optimizer)
loss = model.model.training_loss(data)
model.optimize(dataset)
assert model.model.training_loss(data) < loss
def test_gaussian_process_regression_optimize(
optimizer: gpflow.optimizers.Scipy | tf.optimizers.Optimizer | None, gpr_interface_factory
) -> None:
data = _mock_data()
optimizer_wrapper = create_optimizer(optimizer, {})
model = gpr_interface_factory(*data, optimizer=optimizer_wrapper)
internal_model = model.model
loss = internal_model.training_loss()
model.optimize(Dataset(*data))
assert internal_model.training_loss() < loss
def test_sparse_variational_optimize(batcher, compile: bool) -> None:
x_observed = np.linspace(0, 100, 100).reshape((-1, 1))
y_observed = _3x_plus_gaussian_noise(x_observed)
data = x_observed, y_observed
dataset = Dataset(*data)
optimizer = create_optimizer(
tf.optimizers.Adam(),
dict(max_iter=20, batch_size=10, dataset_builder=batcher, compile=compile),
)
model = SparseVariational(_svgp(x_observed[:10]), dataset, optimizer=optimizer)
loss = model.model.training_loss(data)
model.optimize(dataset)
assert model.model.training_loss(data) < loss
def test_gaussian_process_regression_optimize(
optimizer: gpflow.optimizers.Scipy | tf.optimizers.Optimizer,
gpflow_interface_factory: ModelFactoryType,
compile: bool,
) -> None:
data = mock_data()
optimizer_wrapper = create_optimizer(optimizer, dict(compile=compile))
model, _ = gpflow_interface_factory(*data, optimizer=optimizer_wrapper)
internal_model = model.model
if isinstance(internal_model, SVGP):
args = {"data": data}
else:
args = {}
loss = internal_model.training_loss(**args)
model.optimize(Dataset(*data))
assert internal_model.training_loss(**args) < loss
def test_vgp_optimize_with_and_without_natgrads(
batcher: DatasetTransformer, compile: bool, use_natgrads: bool
) -> None:
x_observed = np.linspace(0, 100, 100).reshape((-1, 1))
y_observed = _3x_plus_gaussian_noise(x_observed)
data = x_observed, y_observed
dataset = Dataset(*data)
optimizer = create_optimizer(
tf.optimizers.Adam(),
dict(max_iter=10, batch_size=10, dataset_builder=batcher, compile=compile),
)
model = VariationalGaussianProcess(
vgp_model(x_observed[:10], y_observed[:10]), optimizer=optimizer, use_natgrads=use_natgrads
)
loss = model.model.training_loss()
model.optimize(dataset)
assert model.model.training_loss() < loss
def test_gaussian_process_regression_ref_optimize(
gpflow_interface_factory: ModelFactoryType,
) -> None:
x = tf.constant(np.arange(5).reshape(-1, 1), dtype=gpflow.default_float())
y = fnc_2sin_x_over_3(x)
model, _reference_model = gpflow_interface_factory(
x, y, optimizer=create_optimizer(gpflow.optimizers.Scipy(), {})
)
reference_model = _reference_model(x, y)
model.optimize(Dataset(x, y))
internal_model = model.model
if isinstance(internal_model, SVGP):
args = {"data": (x, y)}
else:
args = {}
reference_model.data = (
tf.Variable(
reference_model.data[0],
trainable=False,
shape=[None, *reference_model.data[0].shape[1:]],
),
tf.Variable(
reference_model.data[1],
trainable=False,
shape=[None, *reference_model.data[1].shape[1:]],
),
)
gpflow.optimizers.Scipy().minimize(
reference_model.training_loss_closure(**args, compile=False),
reference_model.trainable_variables,
)
npt.assert_allclose(
internal_model.training_loss(**args), reference_model.training_loss(**args), rtol=1e-6
)
def test_create_optimizer_tf_produces_correct_optimizer() -> None:
optim = create_optimizer(tf.optimizers.Adam(), {})
assert isinstance(optim, BatchOptimizer)
def test_create_optimizer_raises_on_none() -> None:
with pytest.raises(NotImplementedError):
create_optimizer(None, {})
def test_create_optimizer_scipy_produces_correct_optimizer() -> None:
optim = create_optimizer(gpflow.optimizers.Scipy(), {})
assert isinstance(optim,
Optimizer) and not isinstance(optim, BatchOptimizer)
