# coding=utf-8
# Copyright 2019 The Interval Bound Propagation 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.
"""Tests for bounds."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from absl.testing import parameterized
import interval_bound_propagation as ibp
import numpy as np
import sonnet as snt
import tensorflow.compat.v1 as tf
class IntervalBoundsTest(parameterized.TestCase, tf.test.TestCase):
def testFCIntervalBounds(self):
m = snt.Linear(1, initializers={
'w': tf.constant_initializer(1.),
'b': tf.constant_initializer(2.),
})
z = tf.constant([[1, 2, 3]], dtype=tf.float32)
m(z) # Connect to create weights.
m = ibp.LinearFCWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
l, u = sess.run([output_bounds.lower, output_bounds.upper])
l = l.item()
u = u.item()
self.assertAlmostEqual(5., l)
self.assertAlmostEqual(11., u)
def testConv1dIntervalBounds(self):
m = snt.Conv1D(
output_channels=1,
kernel_shape=2,
padding='VALID',
stride=1,
use_bias=True,
initializers={
'w': tf.constant_initializer(1.),
'b': tf.constant_initializer(2.),
})
z = tf.constant([3, 4], dtype=tf.float32)
z = tf.reshape(z, [1, 2, 1])
m(z) # Connect to create weights.
m = ibp.LinearConv1dWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
l, u = sess.run([output_bounds.lower, output_bounds.upper])
l = l.item()
u = u.item()
self.assertAlmostEqual(7., l)
self.assertAlmostEqual(11., u)
def testConv2dIntervalBounds(self):
m = snt.Conv2D(
output_channels=1,
kernel_shape=(2, 2),
padding='VALID',
stride=1,
use_bias=True,
initializers={
'w': tf.constant_initializer(1.),
'b': tf.constant_initializer(2.),
})
z = tf.constant([1, 2, 3, 4], dtype=tf.float32)
z = tf.reshape(z, [1, 2, 2, 1])
m(z) # Connect to create weights.
m = ibp.LinearConv2dWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
l, u = sess.run([output_bounds.lower, output_bounds.upper])
l = l.item()
u = u.item()
self.assertAlmostEqual(8., l)
self.assertAlmostEqual(16., u)
def testReluIntervalBounds(self):
m = tf.nn.relu
z = tf.constant([[-2, 3]], dtype=tf.float32)
m = ibp.IncreasingMonotonicWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds)
with self.test_session() as sess:
l, u = sess.run([output_bounds.lower, output_bounds.upper])
self.assertAlmostEqual([[0., 2.]], l.tolist())
self.assertAlmostEqual([[0., 4.]], u.tolist())
def testMulIntervalBounds(self):
m = tf.multiply
z = tf.constant([[-2, 3, 0]], dtype=tf.float32)
m = ibp.PiecewiseMonotonicWrapper(m, (0,))
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds, input_bounds)
with self.test_session() as sess:
l, u = sess.run([output_bounds.lower, output_bounds.upper])
self.assertAlmostEqual([[1., 4., -1.]], l.tolist())
self.assertAlmostEqual([[9., 16., 1.]], u.tolist())
def testSubIntervalBounds(self):
m = tf.subtract
z = tf.constant([[-2, 3, 0]], dtype=tf.float32)
m = ibp.PiecewiseMonotonicWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds, input_bounds)
with self.test_session() as sess:
l, u = sess.run([output_bounds.lower, output_bounds.upper])
self.assertAlmostEqual([[-2., -2., -2.]], l.tolist())
self.assertAlmostEqual([[2., 2., 2.]], u.tolist())
@parameterized.named_parameters(
('DefaultAxis', -1, [[[1., 0.5, 0.5], [1., 0.5, 0.5]],
[[1. / 3, 0., 0.], [1. / 3, 0., 0.]]]),
('NonDefaultAxis', 0, [[[1., 1., 1.], [1., 1., 1.]],
[[0., 0., 0.], [0., 0., 0.]]]))
def testSoftmaxIntervalBounds(self, axis, expected_outputs):
z = tf.constant([[1., -10., -10.], [1., -10., -10.]])
input_bounds = ibp.IntervalBounds(z - 1.0, z + 10.0)
softmax_fn = lambda x: tf.nn.softmax(x, axis=axis)
softmax_fn = ibp.VerifiableModelWrapper(softmax_fn)
softmax_fn(z)
output_bounds = softmax_fn.propagate_bounds(input_bounds)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
l, u = sess.run([output_bounds.lower, output_bounds.upper])
self.assertTrue(np.all(np.abs(expected_outputs[0] - u) < 1e-3))
self.assertTrue(np.all(np.abs(expected_outputs[1] - l) < 1e-3))
def testBatchNormIntervalBounds(self):
z = tf.constant([[1, 2, 3]], dtype=tf.float32)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
g = tf.reshape(tf.range(-1, 2, dtype=tf.float32), [1, 3])
b = tf.reshape(tf.range(3, dtype=tf.float32), [1, 3])
batch_norm = ibp.BatchNorm(scale=True, offset=True, eps=0., initializers={
'gamma': lambda *args, **kwargs: g,
'beta': lambda *args, **kwargs: b,
'moving_mean': tf.constant_initializer(1.),
'moving_variance': tf.constant_initializer(4.),
})
batch_norm(z, is_training=False)
batch_norm = ibp.BatchNormWrapper(batch_norm)
# Test propagation.
output_bounds = batch_norm.propagate_bounds(input_bounds)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
l, u = sess.run([output_bounds.lower, output_bounds.upper])
self.assertAlmostEqual([[-.5, 1., 2.5]], l.tolist())
self.assertAlmostEqual([[.5, 1., 3.5]], u.tolist())
def testCaching(self):
m = snt.Linear(1, initializers={
'w': tf.constant_initializer(1.),
'b': tf.constant_initializer(2.),
})
z = tf.placeholder(shape=(1, 3), dtype=tf.float32)
m(z) # Connect to create weights.
m = ibp.LinearFCWrapper(m)
input_bounds = ibp.IntervalBounds(z - 1., z + 1.)
output_bounds = m.propagate_bounds(input_bounds)
input_bounds.enable_caching()
output_bounds.enable_caching()
update_all_caches_op = tf.group([input_bounds.update_cache_op,
output_bounds.update_cache_op])
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
# Initialise the caches based on the model inputs.
sess.run(update_all_caches_op, feed_dict={z: [[1., 2., 3.]]})
l, u = sess.run([output_bounds.lower, output_bounds.upper])
l = l.item()
u = u.item()
self.assertAlmostEqual(5., l)
self.assertAlmostEqual(11., u)
# Update the cache based on a different set of inputs.
sess.run([output_bounds.update_cache_op], feed_dict={z: [[2., 3., 7.]]})
# We only updated the output bounds' cache.
# This asserts that the computation depends on the underlying
# input bounds tensor, not on cached version of it.
# (Thus it doesn't matter what order the caches are updated.)
l, u = sess.run([output_bounds.lower, output_bounds.upper])
l = l.item()
u = u.item()
self.assertAlmostEqual(11., l)
self.assertAlmostEqual(17., u)
if __name__ == '__main__':
tf.test.main()
