https://github.com/GPflow/GPflow
Tip revision: f4ce06708816199b1926b627322181b74d7a75eb authored by Alexander G. de G. Matthews on 30 August 2017, 11:28:47 UTC
Merge pull request #496 from GPflow/artemav/release-update
Merge pull request #496 from GPflow/artemav/release-update
Tip revision: f4ce067
test_model.py
# Copyright 2016 the GPflow authors.
#
# 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 __future__ import print_function
from __future__ import print_function
import gpflow
import tensorflow as tf
import numpy as np
import unittest
from testing.gpflow_testcase import GPflowTestCase
class TestOptimize(GPflowTestCase):
def setUp(self):
rng = np.random.RandomState(0)
class Quadratic(gpflow.model.Model):
def __init__(self):
gpflow.model.Model.__init__(self)
self.x = gpflow.param.Param(rng.randn(10))
def build_likelihood(self):
return -tf.reduce_sum(tf.square(self.x))
self.m = Quadratic()
def test_adam(self):
with self.test_session():
o = tf.train.AdamOptimizer()
self.m.optimize(o, maxiter=5000)
self.assertTrue(self.m.x.value.max() < 1e-2)
def test_lbfgsb(self):
with self.test_session():
self.m.optimize(disp=False)
self.assertTrue(self.m.x.value.max() < 1e-6)
def test_feval_counter(self):
with self.test_session():
self.m.compile()
self.m.num_fevals = 0
for _ in range(10):
self.m._objective(self.m.get_free_state())
self.assertTrue(self.m.num_fevals == 10)
class TestNeedsRecompile(GPflowTestCase):
def setUp(self):
with self.test_session():
self.m = gpflow.model.Model()
self.m.p = gpflow.param.Param(1.0)
def test_fix(self):
with self.test_session():
self.m._needs_recompile = False
self.m.p.fixed = True
self.assertTrue(self.m._needs_recompile)
def test_replace_param(self):
with self.test_session():
self.m._needs_recompile = False
new_p = gpflow.param.Param(3.0)
self.m.p = new_p
self.assertTrue(self.m._needs_recompile)
def test_set_prior(self):
with self.test_session():
self.m._needs_recompile = False
self.m.p.prior = gpflow.priors.Gaussian(0, 1)
self.assertTrue(self.m._needs_recompile)
def test_set_transform(self):
with self.test_session():
self.m._needs_recompile = False
self.m.p.transform = gpflow.transforms.Identity()
self.assertTrue(self.m._needs_recompile)
def test_replacement(self):
with self.test_session():
m = gpflow.model.Model()
m.p = gpflow.param.Parameterized()
m.p.p = gpflow.param.Param(1.0)
m._needs_recompile = False
# replace Parameterized
new_p = gpflow.param.Parameterized()
new_p.p = gpflow.param.Param(1.0)
m.p = new_p
self.assertTrue(m._needs_recompile is True)
class TestModelSessionGraphArguments(GPflowTestCase):
"""Tests for external graph and session passed to model."""
class Dummy(gpflow.model.Model):
"""Dummy class with naive build_likelihood function"""
def __init__(self):
gpflow.model.Model.__init__(self)
self.x = gpflow.param.Param(10)
def build_likelihood(self):
return tf.negative(tf.reduce_sum(tf.square(self.x)))
def test_session_graph_properties(self):
models = [TestModelSessionGraphArguments.Dummy()
for i in range(6)]
m1, m2, m3, m4, m5, m6 = models
session = tf.Session()
graph = tf.Graph()
m1.compile()
m2.compile(session=session)
m3.compile(graph=graph)
with graph.as_default():
m4.compile()
m5.compile(session=session, graph=graph)
with self.test_session() as sess_default:
m6.compile()
sessions = [m.session for m in models]
sess1, sess2, sess3, sess4, sess5, sess6 = sessions
sessions_set = set(map(str, sessions))
self.assertNotEqual(sess_default, tf.get_default_graph())
self.assertEqual(len(sessions_set), 5)
self.assertEqual(sess2, sess5)
self.assertEqual(sess1.graph, sess2.graph)
self.assertEqual(sess3.graph, sess4.graph)
self.assertEqual(sess2.graph, tf.get_default_graph())
self.assertEqual(sess3.graph, graph)
self.assertEqual(sess6.graph, sess_default.graph)
self.assertEqual(sess6, sess_default)
self.assertNotEqual(sess1.graph, sess3.graph)
m6.compile(graph=sess_default.graph)
self.assertEqual(sess6.graph, sess_default.graph)
self.assertEqual(sess6, sess_default)
[m.session.close() for m in models]
class KeyboardRaiser:
"""
This wraps a function and makes it raise a KeyboardInterrupt after some number of calls
"""
def __init__(self, iters_to_raise, f):
self.iters_to_raise, self.f = iters_to_raise, f
self.count = 0
def __call__(self, *a, **kw):
self.count += 1
if self.count >= self.iters_to_raise:
raise KeyboardInterrupt
return self.f(*a, **kw)
class TestKeyboardCatching(GPflowTestCase):
def setUp(self):
with self.test_session():
X = np.random.randn(1000, 3)
Y = np.random.randn(1000, 3)
Z = np.random.randn(100, 3)
self.m = gpflow.sgpr.SGPR(X, Y, Z=Z, kern=gpflow.kernels.RBF(3))
def test_optimize_np(self):
with self.test_session():
x0 = self.m.get_free_state()
self.m.compile()
self.m._objective = KeyboardRaiser(15, self.m._objective)
self.m.optimize(disp=0, maxiter=1000, ftol=0, gtol=0)
x1 = self.m.get_free_state()
self.assertFalse(np.allclose(x0, x1))
def test_optimize_tf(self):
with self.test_session():
x0 = self.m.get_free_state()
callback = KeyboardRaiser(5, lambda x: None)
o = tf.train.AdamOptimizer()
self.m.optimize(o, maxiter=10, callback=callback)
x1 = self.m.get_free_state()
self.assertFalse(np.allclose(x0, x1))
class TestLikelihoodAutoflow(GPflowTestCase):
def setUp(self):
with self.test_session():
X = np.random.randn(1000, 3)
Y = np.random.randn(1000, 3)
Z = np.random.randn(100, 3)
self.m = gpflow.sgpr.SGPR(X, Y, Z=Z, kern=gpflow.kernels.RBF(3))
def test_lik_and_prior(self):
with self.test_session():
l0 = self.m.compute_log_likelihood()
p0 = self.m.compute_log_prior()
self.m.kern.variance.prior = gpflow.priors.Gamma(1.4, 1.6)
l1 = self.m.compute_log_likelihood()
p1 = self.m.compute_log_prior()
self.assertTrue(p0 == 0.0)
self.assertFalse(p0 == p1)
self.assertTrue(l0 == l1)
class TestName(GPflowTestCase):
def test_name(self):
m = gpflow.model.Model(name='foo')
self.assertEqual(m.name, 'foo')
class TestNoRecompileThroughNewModelInstance(GPflowTestCase):
""" Regression tests for Bug #454 """
def setUp(self):
self.X = np.random.rand(10, 2)
self.Y = np.random.rand(10, 1)
def test_gpr(self):
with self.test_session():
m1 = gpflow.gpr.GPR(self.X, self.Y, gpflow.kernels.Matern32(2))
m1.compile()
m2 = gpflow.gpr.GPR(self.X, self.Y, gpflow.kernels.Matern32(2))
self.assertFalse(m1._needs_recompile)
def test_sgpr(self):
with self.test_session():
m1 = gpflow.sgpr.SGPR(self.X, self.Y, gpflow.kernels.Matern32(2), Z=self.X)
m1.compile()
m2 = gpflow.sgpr.SGPR(self.X, self.Y, gpflow.kernels.Matern32(2), Z=self.X)
self.assertFalse(m1._needs_recompile)
def test_gpmc(self):
with self.test_session():
m1 = gpflow.gpmc.GPMC(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT())
m1.compile()
m2 = gpflow.gpmc.GPMC(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT())
self.assertFalse(m1._needs_recompile)
def test_sgpmc(self):
with self.test_session():
m1 = gpflow.sgpmc.SGPMC(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT(),
Z=self.X)
m1.compile()
m2 = gpflow.sgpmc.SGPMC(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT(),
Z=self.X)
self.assertFalse(m1._needs_recompile)
def test_svgp(self):
with self.test_session():
m1 = gpflow.svgp.SVGP(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT(),
Z=self.X)
m1.compile()
m2 = gpflow.svgp.SVGP(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT(),
Z=self.X)
self.assertFalse(m1._needs_recompile)
def test_vgp(self):
with self.test_session():
m1 = gpflow.vgp.VGP(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT())
m1.compile()
m2 = gpflow.vgp.VGP(
self.X, self.Y,
gpflow.kernels.Matern32(2),
likelihood=gpflow.likelihoods.StudentT())
self.assertFalse(m1._needs_recompile)
if __name__ == "__main__":
unittest.main()