https://github.com/GPflow/GPflow
Tip revision: dc4b3634a8c71be058a5ad76e0c83db699479fe5 authored by dependabot[bot] on 20 March 2024, 15:48:08 UTC
Bump black from 20.8b1 to 24.3.0
Bump black from 20.8b1 to 24.3.0
Tip revision: dc4b363
test_inducing_variables.py
# Copyright 2020 The GPflow Contributors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from dataclasses import dataclass
from typing import Callable, Optional, Tuple
import numpy as np
import pytest
import tensorflow as tf
from check_shapes import get_shape
from check_shapes.error_contexts import ErrorContext, MessageBuilder
import gpflow
import gpflow.inducing_variables as giv
@dataclass(frozen=True)
class TestContext(ErrorContext):
message: str = "Fake test error context."
def print(self, builder: MessageBuilder) -> None:
builder.add_line(self.message)
def test_inducing_points_with_variable_shape() -> None:
N, M1, D, P = 50, 13, 3, 1
X, Y = np.random.randn(N, D), np.random.randn(N, P)
Z1 = np.random.randn(M1, D)
# use explicit tf.Variable with None shape:
iv = giv.InducingPoints(
tf.Variable(Z1, trainable=False, dtype=gpflow.default_float(), shape=(None, D))
)
# Note that we cannot have Z be trainable if we want to be able to change its shape;
# TensorFlow optimizers expect shape to be known at construction time.
m = gpflow.models.SGPR(data=(X, Y), kernel=gpflow.kernels.Matern32(), inducing_variable=iv)
# Check 1: that we can still optimize with None shape
opt = tf.optimizers.Adam()
@tf.function
def optimization_step() -> None:
opt.minimize(m.training_loss, m.trainable_variables)
optimization_step()
# Check 2: that we can successfully assign a new Z with different number of inducing points!
Z2 = np.random.randn(M1 + 1, D)
m.inducing_variable.Z.assign(Z2)
# Check 3: that we can also optimize with changed Z tensor
optimization_step()
@pytest.mark.parametrize(
"iv_factory,expected_shape",
[
(lambda t: giv.InducingPoints(t), (7, 3, 1)),
(lambda t: giv.Multiscale(t, t), (7, 3, 1)),
(lambda t: giv.InducingPatches(t), (7, 3, 1)),
(
lambda t: giv.FallbackSharedIndependentInducingVariables(giv.InducingPoints(t)),
(7, 3, None),
),
(
lambda t: giv.FallbackSeparateIndependentInducingVariables(
[
giv.InducingPoints(t),
giv.InducingPoints(t),
]
),
(7, 3, 2),
),
(
lambda t: giv.SharedIndependentInducingVariables(giv.InducingPoints(t)),
(7, 3, None),
),
(
lambda t: giv.SeparateIndependentInducingVariables(
[
giv.InducingPoints(t),
giv.InducingPoints(t),
]
),
(7, 3, 2),
),
],
)
@pytest.mark.parametrize("none_shape", [False, True])
def test_shape(
iv_factory: Callable[[tf.Tensor], giv.InducingVariables],
none_shape: bool,
expected_shape: Tuple[Optional[int], ...],
) -> None:
M = expected_shape[0]
ones = tf.ones((7, 3))
kwargs = {"shape": tf.TensorShape(None)} if none_shape else {}
iv = iv_factory(tf.Variable(ones, **kwargs))
if none_shape:
assert get_shape(iv, TestContext()) is None
else:
assert expected_shape == get_shape(iv, TestContext())
assert M == iv.num_inducing
assert M == len(iv)
@pytest.mark.parametrize("none_shape", [False, True])
def test_shape__inconsistent(none_shape: bool) -> None:
kwargs = {"shape": tf.TensorShape(None)} if none_shape else {}
iv = giv.SeparateIndependentInducingVariables(
[
giv.InducingPoints(tf.Variable(tf.ones((5, 3)), **kwargs)),
giv.InducingPoints(tf.Variable(tf.ones((7, 3)), **kwargs)),
]
)
assert get_shape(iv, TestContext()) is None
with pytest.raises(tf.errors.InvalidArgumentError):
iv.num_inducing
with pytest.raises(tf.errors.InvalidArgumentError):
len(iv)