ruby/tensorflow_privacy
1
0
Fork 0
forked from 626_privacy/tensorflow_privacy
Code Pull requests Activity

Add support for microbatching in the tf.keras.layers.LayerNormalization fast square norm function.

Browse source 
PiperOrigin-RevId: 565050132
This commit is contained in:
A. Unique TensorFlower 2023-09-13 07:56:03 -07:00
parent bcc0d4927e
commit c7db4fa8cb
3 changed files with 19 additions and 12 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

5
tensorflow_privacy/privacy/fast_gradient_clipping/registry_functions/BUILD
View file

@ -102,7 +102,10 @@ py_library(
name = "layer_normalization", name = "layer_normalization",
srcs = ["layer_normalization.py"], srcs = ["layer_normalization.py"],
srcs_version = "PY3", srcs_version = "PY3",
deps = ["//tensorflow_privacy/privacy/fast_gradient_clipping:type_aliases"], deps = [
"//tensorflow_privacy/privacy/fast_gradient_clipping:common_manip_utils",
"//tensorflow_privacy/privacy/fast_gradient_clipping:type_aliases",
],
) )
py_test( py_test(

22
tensorflow_privacy/privacy/fast_gradient_clipping/registry_functions/layer_normalization.py
View file

@ -15,18 +15,13 @@
from typing import Any, Mapping, Tuple, Union from typing import Any, Mapping, Tuple, Union
import tensorflow as tf import tensorflow as tf
from tensorflow_privacy.privacy.fast_gradient_clipping import common_manip_utils
from tensorflow_privacy.privacy.fast_gradient_clipping import type_aliases from tensorflow_privacy.privacy.fast_gradient_clipping import type_aliases
# ============================================================================== # ==============================================================================
# Supported Keras layers # Supported Keras layers
# ============================================================================== # ==============================================================================
def _sqr_norm_fn(grads):
stacked_grads = tf.stack(grads, axis=-1)
reduction_axes = tf.range(1, tf.rank(stacked_grads))
return tf.reduce_sum(tf.square(stacked_grads), axis=reduction_axes)
def layer_normalization_computation( def layer_normalization_computation(
layer_instance: tf.keras.layers.LayerNormalization, layer_instance: tf.keras.layers.LayerNormalization,
input_args: Tuple[Any, ...], input_args: Tuple[Any, ...],
@ -51,9 +46,6 @@ def layer_normalization_computation(
See `dense_layer_computation()` in `dense.py`. See `dense_layer_computation()` in `dense.py`.
""" """
del input_kwargs # Unused in layer normaliztion calls. del input_kwargs # Unused in layer normaliztion calls.
if num_microbatches is not None:
raise NotImplementedError("Microbatching is not currently supported.")
# To make sure the watched variables (beta, gamma) generate per-example # To make sure the watched variables (beta, gamma) generate per-example
# gradients, we need to convert trainable variables from shape [S] to # gradients, we need to convert trainable variables from shape [S] to
# [batch_size, S] via duplication to `tf.shape(inputs)` via broadcasting. # [batch_size, S] via duplication to `tf.shape(inputs)` via broadcasting.
@ -86,4 +78,14 @@ def layer_normalization_computation(
layer_instance.gamma = orig_gamma layer_instance.gamma = orig_gamma
layer_instance.beta = orig_beta layer_instance.beta = orig_beta
return base_vars, outputs, _sqr_norm_fn def sqr_norm_fn(grads):
stacked_grads = tf.stack(grads, axis=-1)
if num_microbatches is not None:
stacked_grads = common_manip_utils.maybe_add_microbatch_axis(
grads, num_microbatches
)
stacked_grads = tf.reduce_sum(stacked_grads, axis=1)
reduction_axes = tf.range(1, tf.rank(stacked_grads))
return tf.reduce_sum(tf.square(stacked_grads), axis=reduction_axes)
return base_vars, outputs, sqr_norm_fn

4
tensorflow_privacy/privacy/fast_gradient_clipping/registry_functions/layer_normalization_test.py
View file

@ -87,6 +87,7 @@ class GradNormTest(tf.test.TestCase, parameterized.TestCase):
layer_name=list(get_layer_norm_layer_generators().keys()), layer_name=list(get_layer_norm_layer_generators().keys()),
parameter_tuple=get_layer_norm_parameter_tuples(), parameter_tuple=get_layer_norm_parameter_tuples(),
layer_registry_name=list(get_layer_norm_registries().keys()), layer_registry_name=list(get_layer_norm_registries().keys()),
num_microbatches=[None, 2],
is_eager=[True, False], is_eager=[True, False],
) )
def test_gradient_norms_on_various_models( def test_gradient_norms_on_various_models(
@ -95,6 +96,7 @@ class GradNormTest(tf.test.TestCase, parameterized.TestCase):
layer_name, layer_name,
parameter_tuple, parameter_tuple,
layer_registry_name, layer_registry_name,
num_microbatches,
is_eager, is_eager,
): ):
# Parse inputs to generate test data. # Parse inputs to generate test data.
@ -121,7 +123,7 @@ class GradNormTest(tf.test.TestCase, parameterized.TestCase):
return common_test_utils.get_computed_and_true_norms_from_model( return common_test_utils.get_computed_and_true_norms_from_model(
model=model, model=model,
per_example_loss_fn=None, per_example_loss_fn=None,
num_microbatches=None, num_microbatches=num_microbatches,
x_batch=[x_batch, x_batch] if model_name == 'tower2' else x_batch, x_batch=[x_batch, x_batch] if model_name == 'tower2' else x_batch,
weight_batch=None, weight_batch=None,
registry=get_layer_norm_registries()[layer_registry_name], registry=get_layer_norm_registries()[layer_registry_name],