christegho
commented
4 years ago This is my implementation for tf.1.13
# Copyright 2018 The TensorFlow Authors. 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.
# ==============================================================================
"""Adam for TensorFlow."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
from tensorflow.python.eager import context
from tensorflow.python.framework import ops
from tensorflow.python.ops import control_flow_ops
from tensorflow.python.ops import math_ops
from tensorflow.python.ops import resource_variable_ops
from tensorflow.python.ops import state_ops
from tensorflow.python.training import optimizer
from tensorflow.python.training import training_ops
class AdamOptimizer(optimizer.Optimizer):
"""Optimizer that implements the Adam algorithm.
See [Kingma et al., 2014](http://arxiv.org/abs/1412.6980)
([pdf](http://arxiv.org/pdf/1412.6980.pdf)).
"""
def __init__(self, learning_rate=0.001, beta1=0.9, beta2=0.999, epsilon=1e-8,
use_locking=False, name="Adam", lr_dropout=0.0):
"""Construct a new Adam optimizer.
Initialization:
$$m_0 := 0 \text{(Initialize initial 1st moment vector)}$$
$$v_0 := 0 \text{(Initialize initial 2nd moment vector)}$$
$$t := 0 \text{(Initialize timestep)}$$
The update rule for `variable` with gradient `g` uses an optimization
described at the end of section2 of the paper:
$$t := t + 1$$
$$lr_t := \text{learning\_rate} * \sqrt{1 - beta_2^t} / (1 - beta_1^t)$$
$$m_t := beta_1 * m_{t-1} + (1 - beta_1) * g$$
$$v_t := beta_2 * v_{t-1} + (1 - beta_2) * g * g$$
$$variable := variable - lr_t * m_t / (\sqrt{v_t} + \epsilon)$$
The default value of 1e-8 for epsilon might not be a good default in
general. For example, when training an Inception network on ImageNet a
current good choice is 1.0 or 0.1. Note that since AdamOptimizer uses the
formulation just before Section 2.1 of the Kingma and Ba paper rather than
the formulation in Algorithm 1, the "epsilon" referred to here is "epsilon
hat" in the paper.
The sparse implementation of this algorithm (used when the gradient is an
IndexedSlices object, typically because of `tf.gather` or an embedding
lookup in the forward pass) does apply momentum to variable slices even if
they were not used in the forward pass (meaning they have a gradient equal
to zero). Momentum decay (beta1) is also applied to the entire momentum
accumulator. This means that the sparse behavior is equivalent to the dense
behavior (in contrast to some momentum implementations which ignore momentum
unless a variable slice was actually used).
Args:
learning_rate: A Tensor or a floating point value. The learning rate.
beta1: A float value or a constant float tensor.
The exponential decay rate for the 1st moment estimates.
beta2: A float value or a constant float tensor.
The exponential decay rate for the 2nd moment estimates.
epsilon: A small constant for numerical stability. This epsilon is
"epsilon hat" in the Kingma and Ba paper (in the formula just before
Section 2.1), not the epsilon in Algorithm 1 of the paper.
use_locking: If True use locks for update operations.
name: Optional name for the operations created when applying gradients.
Defaults to "Adam".
@compatibility(eager)
When eager execution is enabled, `learning_rate`, `beta1`, `beta2`, and
`epsilon` can each be a callable that takes no arguments and returns the
actual value to use. This can be useful for changing these values across
different invocations of optimizer functions.
@end_compatibility
"""
super(AdamOptimizer, self).__init__(use_locking, name)
self._lr = learning_rate
self._beta1 = beta1
self._beta2 = beta2
self._epsilon = epsilon
self._lr_dropout = lr_dropout
# Tensor versions of the constructor arguments, created in _prepare().
self._lr_t = None
self._beta1_t = None
self._beta2_t = None
self._epsilon_t = None
# Created in SparseApply if needed.
self._updated_lr = None
def _get_beta_accumulators(self):
with ops.init_scope():
if context.executing_eagerly():
graph = None
else:
graph = ops.get_default_graph()
return (self._get_non_slot_variable("beta1_power", graph=graph),
self._get_non_slot_variable("beta2_power", graph=graph))
def _create_slots(self, var_list):
# Create the beta1 and beta2 accumulators on the same device as the first
# variable. Sort the var_list to make sure this device is consistent across
# workers (these need to go on the same PS, otherwise some updates are
# silently ignored).
first_var = min(var_list, key=lambda x: x.name)
self._create_non_slot_variable(initial_value=self._beta1,
name="beta1_power",
colocate_with=first_var)
self._create_non_slot_variable(initial_value=self._beta2,
name="beta2_power",
colocate_with=first_var)
# Create slots for the first and second moments.
for v in var_list:
self._zeros_slot(v, "m", self._name)
self._zeros_slot(v, "v", self._name)
def _prepare(self):
lr = self._call_if_callable(self._lr)
beta1 = self._call_if_callable(self._beta1)
beta2 = self._call_if_callable(self._beta2)
epsilon = self._call_if_callable(self._epsilon)
self._lr_t = ops.convert_to_tensor(lr, name="learning_rate")
self._beta1_t = ops.convert_to_tensor(beta1, name="beta1")
self._beta2_t = ops.convert_to_tensor(beta2, name="beta2")
self._epsilon_t = ops.convert_to_tensor(epsilon, name="epsilon")
def _apply_dense(self, grad, var):
m = self.get_slot(var, "m")
v = self.get_slot(var, "v")
var_dtype = var.dtype.base_dtype
beta1_power, beta2_power = self._get_beta_accumulators()
beta_1_power = math_ops.cast(beta1_power, var_dtype)
beta_2_power = math_ops.cast(beta2_power, var_dtype)
beta_1_t = math_ops.cast(self._beta1_t, var_dtype)
beta_2_t = math_ops.cast(self._beta2_t, var_dtype)
epsilon_t = math_ops.cast(self._epsilon_t, var_dtype)
m_t = m.assign(
beta_1_t * m + (1.0 - beta_1_t) * grad,
use_locking=self._use_locking)
m_corr_t = m_t / (1.0 - beta_1_power)
v_t = v.assign(
beta_2_t * v + (1.0 - beta_2_t) * tf.square(grad),
use_locking=self._use_locking)
v_corr_t = tf.sqrt(v_t / (1.0 - beta_2_power))
distrib = tf.contrib.distributions.Bernoulli(probs=[self._lr_dropout])
sample = math_ops.cast(distrib.sample(grad.get_shape()), var.dtype.base_dtype)
lr_d = math_ops.cast(self._lr_t, var.dtype.base_dtype)*sample
lr_d = tf.squeeze(lr_d)
var_update = state_ops.assign_sub(var, lr_d*m_t/(v_corr_t+epsilon_t))
return control_flow_ops.group(*[var_update, m_t])
def _resource_apply_dense(self, grad, var):
raise NotImplementedError
def _apply_sparse_shared(self, grad, var, indices, scatter_add):
raise NotImplementedError
def _apply_sparse(self, grad, var):
raise NotImplementedError
def _resource_scatter_add(self, x, i, v):
with ops.control_dependencies(
[resource_variable_ops.resource_scatter_add(
x.handle, i, v)]):
return x.value()
def _resource_apply_sparse(self, grad, var, indices):
return self._apply_sparse_shared(
grad, var, indices, self._resource_scatter_add)
def _finish(self, update_ops, name_scope):
# Update the power accumulators.
with ops.control_dependencies(update_ops):
beta1_power, beta2_power = self._get_beta_accumulators()
with ops.colocate_with(beta1_power):
update_beta1 = beta1_power.assign(
beta1_power * self._beta1_t, use_locking=self._use_locking)
update_beta2 = beta2_power.assign(
beta2_power * self._beta2_t, use_locking=self._use_locking)
return control_flow_ops.group(*update_ops + [update_beta1, update_beta2],
name=name_scope)
Any Tensorflow implementation for the Adam optimizer?