626_privacy/tensorflow_privacy
2
0
Fork 1
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity Actions

Clean-up pass to eliminate warnings: replacing deprecated endpoints with recommended versions and annotating test sizes.

Browse source 
PiperOrigin-RevId: 246901723
This commit is contained in:
Ilya Mironov 2019-05-06 14:50:04 -07:00 committed by A. Unique TensorFlower
parent 85280ab568
commit 9cece21d92
6 changed files with 80 additions and 50 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
privacy/dp_query/BUILD
View file

@ -13,17 +13,21 @@ py_library(
deps = [
":dp_query",
":normalized_query",
"//third_party/py/distutils",
"//third_party/py/tensorflow",
],
)
py_test(
name = "gaussian_query_test",
size = "small",
srcs = ["gaussian_query_test.py"],
deps = [
":gaussian_query",
":test_utils",
"//third_party/py/absl/testing:parameterized",
"//third_party/py/numpy",
"//third_party/py/six",
"//third_party/py/tensorflow",
],
)
@ -33,12 +37,14 @@ py_library(
srcs = ["no_privacy_query.py"],
deps = [
":dp_query",
"//third_party/py/distutils",
"//third_party/py/tensorflow",
],
)
py_test(
name = "no_privacy_query_test",
size = "small",
srcs = ["no_privacy_query_test.py"],
deps = [
":no_privacy_query",
@ -53,12 +59,14 @@ py_library(
srcs = ["normalized_query.py"],
deps = [
":dp_query",
"//third_party/py/distutils",
"//third_party/py/tensorflow",
],
)
py_test(
name = "normalized_query_test",
size = "small",
srcs = ["normalized_query_test.py"],
deps = [
":gaussian_query",
@ -73,18 +81,22 @@ py_library(
srcs = ["nested_query.py"],
deps = [
":dp_query",
"//third_party/py/distutils",
"//third_party/py/tensorflow",
],
)
py_test(
name = "nested_query_test",
size = "small",
srcs = ["nested_query_test.py"],
deps = [
":gaussian_query",
":nested_query",
":test_utils",
"//third_party/py/absl/testing:parameterized",
"//third_party/py/distutils",
"//third_party/py/numpy",
"//third_party/py/tensorflow",
],
)

10
privacy/optimizers/dp_optimizer.py
View file

@ -27,9 +27,9 @@ from privacy.dp_query import gaussian_query
def make_optimizer_class(cls):
"""Constructs a DP optimizer class from an existing one."""
if LooseVersion(tf.__version__) < LooseVersion('2.0.0'):
parent_code = tf.train.Optimizer.compute_gradients.__code__
parent_code = tf.compat.v1.train.Optimizer.compute_gradients.__code__
child_code = cls.compute_gradients.__code__
GATE_OP = tf.train.Optimizer.GATE_OP # pylint: disable=invalid-name
GATE_OP = tf.compat.v1.train.Optimizer.GATE_OP # pylint: disable=invalid-name
else:
parent_code = tf.optimizers.Optimizer._compute_gradients.__code__ # pylint: disable=protected-access
child_code = cls._compute_gradients.__code__ # pylint: disable=protected-access
@ -211,9 +211,9 @@ def make_gaussian_optimizer_class(cls):
# Compatibility with tf 1 and 2 APIs
try:
AdagradOptimizer = tf.train.AdagradOptimizer
AdamOptimizer = tf.train.AdamOptimizer
GradientDescentOptimizer = tf.train.GradientDescentOptimizer
AdagradOptimizer = tf.compat.v1.train.AdagradOptimizer
AdamOptimizer = tf.compat.v1.train.AdamOptimizer
GradientDescentOptimizer = tf.compat.v1.train.GradientDescentOptimizer
except: # pylint: disable=bare-except
AdagradOptimizer = tf.optimizers.Adagrad
AdamOptimizer = tf.optimizers.Adam

34
tutorials/lm_dpsgd_tutorial.py
View file

@ -36,6 +36,10 @@ from __future__ import division
from __future__ import print_function
import os
from absl import app
from absl import flags
import numpy as np
import tensorflow as tf
import tensorflow_datasets as tfds
@ -45,20 +49,22 @@ from privacy.analysis.rdp_accountant import compute_rdp
from privacy.analysis.rdp_accountant import get_privacy_spent
from privacy.optimizers import dp_optimizer
tf.flags.DEFINE_boolean('dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
tf.flags.DEFINE_float('learning_rate', 0.001, 'Learning rate for training')
tf.flags.DEFINE_float('noise_multiplier', 0.001,
'Ratio of the standard deviation to the clipping norm')
tf.flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
tf.flags.DEFINE_integer('batch_size', 256, 'Batch size')
tf.flags.DEFINE_integer('epochs', 60, 'Number of epochs')
tf.flags.DEFINE_integer('microbatches', 256, 'Number of microbatches '
'(must evenly divide batch_size)')
tf.flags.DEFINE_string('model_dir', None, 'Model directory')
tf.flags.DEFINE_string('data_dir', None, 'Directory containing the PTB data.')
flags.DEFINE_boolean(
'dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
flags.DEFINE_float('learning_rate', 0.001, 'Learning rate for training')
flags.DEFINE_float('noise_multiplier', 0.001,
'Ratio of the standard deviation to the clipping norm')
flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
flags.DEFINE_integer('batch_size', 256, 'Batch size')
flags.DEFINE_integer('epochs', 60, 'Number of epochs')
flags.DEFINE_integer(
'microbatches', 256, 'Number of microbatches '
'(must evenly divide batch_size)')
flags.DEFINE_string('model_dir', None, 'Model directory')
flags.DEFINE_string('data_dir', None, 'Directory containing the PTB data.')
FLAGS = tf.flags.FLAGS
FLAGS = flags.FLAGS
SEQ_LEN = 80
NB_TRAIN = 45000
@ -217,4 +223,4 @@ def main(unused_argv):
print('Trained with vanilla non-private SGD optimizer')
if __name__ == '__main__':
tf.app.run()
app.run(main)

35
tutorials/mnist_dpsgd_tutorial.py
View file

@ -18,6 +18,9 @@ from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from absl import app
from absl import flags
import numpy as np
import tensorflow as tf
@ -28,26 +31,28 @@ from privacy.optimizers import dp_optimizer
# Compatibility with tf 1 and 2 APIs
try:
GradientDescentOptimizer = tf.train.GradientDescentOptimizer
GradientDescentOptimizer = tf.compat.v1.train.GradientDescentOptimizer
except: # pylint: disable=bare-except
GradientDescentOptimizer = tf.optimizers.SGD # pylint: disable=invalid-name
tf.flags.DEFINE_boolean('dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
tf.flags.DEFINE_float('learning_rate', .15, 'Learning rate for training')
tf.flags.DEFINE_float('noise_multiplier', 1.1,
'Ratio of the standard deviation to the clipping norm')
tf.flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
tf.flags.DEFINE_integer('batch_size', 256, 'Batch size')
tf.flags.DEFINE_integer('epochs', 60, 'Number of epochs')
tf.flags.DEFINE_integer('microbatches', 256, 'Number of microbatches '
'(must evenly divide batch_size)')
tf.flags.DEFINE_string('model_dir', None, 'Model directory')
FLAGS = flags.FLAGS
FLAGS = tf.flags.FLAGS
flags.DEFINE_boolean(
'dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
flags.DEFINE_float('learning_rate', .15, 'Learning rate for training')
flags.DEFINE_float('noise_multiplier', 1.1,
'Ratio of the standard deviation to the clipping norm')
flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
flags.DEFINE_integer('batch_size', 256, 'Batch size')
flags.DEFINE_integer('epochs', 60, 'Number of epochs')
flags.DEFINE_integer(
'microbatches', 256, 'Number of microbatches '
'(must evenly divide batch_size)')
flags.DEFINE_string('model_dir', None, 'Model directory')
class EpsilonPrintingTrainingHook(tf.train.SessionRunHook):
class EpsilonPrintingTrainingHook(tf.estimator.SessionRunHook):
"""Training hook to print current value of epsilon after an epoch."""
def __init__(self, ledger):
@ -203,4 +208,4 @@ def main(unused_argv):
print('Test accuracy after %d epochs is: %.3f' % (epoch, test_accuracy))
if __name__ == '__main__':
tf.app.run()
app.run(main)

6
tutorials/mnist_dpsgd_tutorial_eager.py
View file

@ -18,7 +18,9 @@ from __future__ import print_function
from absl import app
from absl import flags
from distutils.version import LooseVersion
import numpy as np
import tensorflow as tf
@ -28,8 +30,8 @@ from privacy.dp_query.gaussian_query import GaussianAverageQuery
from privacy.optimizers.dp_optimizer import DPGradientDescentOptimizer
if LooseVersion(tf.__version__) < LooseVersion('2.0.0'):
GradientDescentOptimizer = tf.train.GradientDescentOptimizer
tf.enable_eager_execution()
GradientDescentOptimizer = tf.compat.v1.train.GradientDescentOptimizer
tf.compat.v1.enable_eager_execution()
else:
GradientDescentOptimizer = tf.optimizers.SGD # pylint: disable=invalid-name

33
tutorials/mnist_dpsgd_tutorial_keras.py
View file

@ -17,6 +17,9 @@ from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from absl import app
from absl import flags
import numpy as np
import tensorflow as tf
@ -27,23 +30,25 @@ from privacy.optimizers.dp_optimizer import DPGradientDescentOptimizer
# Compatibility with tf 1 and 2 APIs
try:
GradientDescentOptimizer = tf.train.GradientDescentOptimizer
GradientDescentOptimizer = tf.compat.v1.train.GradientDescentOptimizer
except: # pylint: disable=bare-except
GradientDescentOptimizer = tf.optimizers.SGD # pylint: disable=invalid-name
tf.flags.DEFINE_boolean('dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
tf.flags.DEFINE_float('learning_rate', 0.15, 'Learning rate for training')
tf.flags.DEFINE_float('noise_multiplier', 1.1,
'Ratio of the standard deviation to the clipping norm')
tf.flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
tf.flags.DEFINE_integer('batch_size', 250, 'Batch size')
tf.flags.DEFINE_integer('epochs', 60, 'Number of epochs')
tf.flags.DEFINE_integer('microbatches', 250, 'Number of microbatches '
'(must evenly divide batch_size)')
tf.flags.DEFINE_string('model_dir', None, 'Model directory')
flags.DEFINE_boolean(
'dpsgd', True, 'If True, train with DP-SGD. If False, '
'train with vanilla SGD.')
flags.DEFINE_float('learning_rate', 0.15, 'Learning rate for training')
flags.DEFINE_float('noise_multiplier', 1.1,
'Ratio of the standard deviation to the clipping norm')
flags.DEFINE_float('l2_norm_clip', 1.0, 'Clipping norm')
flags.DEFINE_integer('batch_size', 250, 'Batch size')
flags.DEFINE_integer('epochs', 60, 'Number of epochs')
flags.DEFINE_integer(
'microbatches', 250, 'Number of microbatches '
'(must evenly divide batch_size)')
flags.DEFINE_string('model_dir', None, 'Model directory')
FLAGS = tf.flags.FLAGS
FLAGS = flags.FLAGS
def compute_epsilon(steps):
@ -146,4 +151,4 @@ def main(unused_argv):
print('Trained with vanilla non-private SGD optimizer')
if __name__ == '__main__':
tf.app.run()
app.run(main)