1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
|
# Copyright 2015 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.
# ==============================================================================
"""Adadelta for TensorFlow."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from tensorflow.python.framework import ops
from tensorflow.python.ops import math_ops
from tensorflow.python.training import optimizer
from tensorflow.python.training import training_ops
class AdadeltaOptimizer(optimizer.Optimizer):
"""Optimizer that implements the Adadelta algorithm.
See [M. D. Zeiler](http://arxiv.org/abs/1212.5701)
([pdf](http://arxiv.org/pdf/1212.5701v1.pdf))
"""
def __init__(self, learning_rate=0.001, rho=0.95, epsilon=1e-8,
use_locking=False, name="Adadelta"):
"""Construct a new Adadelta optimizer.
Args:
learning_rate: A `Tensor` or a floating point value. The learning rate.
rho: A `Tensor` or a floating point value. The decay rate.
epsilon: A `Tensor` or a floating point value. A constant epsilon used
to better conditioning the grad update.
use_locking: If `True` use locks for update operations.
name: Optional name prefix for the operations created when applying
gradients. Defaults to "Adadelta".
"""
super(AdadeltaOptimizer, self).__init__(use_locking, name)
self._lr = learning_rate
self._rho = rho
self._epsilon = epsilon
# Tensor versions of the constructor arguments, created in _prepare().
self._lr_t = None
self._rho_t = None
self._epsilon_t = None
def _create_slots(self, var_list):
for v in var_list:
self._zeros_slot(v, "accum", self._name)
self._zeros_slot(v, "accum_update", self._name)
def _prepare(self):
self._lr_t = ops.convert_to_tensor(self._lr, name="lr")
self._rho_t = ops.convert_to_tensor(self._rho, name="rho")
self._epsilon_t = ops.convert_to_tensor(self._epsilon, name="epsilon")
def _apply_dense(self, grad, var):
accum = self.get_slot(var, "accum")
accum_update = self.get_slot(var, "accum_update")
return training_ops.apply_adadelta(
var,
accum,
accum_update,
math_ops.cast(self._lr_t, var.dtype.base_dtype),
math_ops.cast(self._rho_t, var.dtype.base_dtype),
math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
grad,
use_locking=self._use_locking)
def _resource_apply_dense(self, grad, var):
accum = self.get_slot(var, "accum")
accum_update = self.get_slot(var, "accum_update")
return training_ops.resource_apply_adadelta(
var.handle,
accum.handle,
accum_update.handle,
math_ops.cast(self._lr_t, grad.dtype.base_dtype),
math_ops.cast(self._rho_t, grad.dtype.base_dtype),
math_ops.cast(self._epsilon_t, grad.dtype.base_dtype),
grad,
use_locking=self._use_locking)
def _apply_sparse(self, grad, var):
accum = self.get_slot(var, "accum")
accum_update = self.get_slot(var, "accum_update")
return training_ops.sparse_apply_adadelta(
var,
accum,
accum_update,
math_ops.cast(self._lr_t, var.dtype.base_dtype),
math_ops.cast(self._rho_t, var.dtype.base_dtype),
math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
grad.values,
grad.indices,
use_locking=self._use_locking)
def _resource_apply_sparse(self, grad, var, indices):
accum = self.get_slot(var, "accum")
accum_update = self.get_slot(var, "accum_update")
return training_ops.resource_sparse_apply_adadelta(
var.handle,
accum.handle,
accum_update.handle,
math_ops.cast(self._lr_t, grad.dtype),
math_ops.cast(self._rho_t, grad.dtype),
math_ops.cast(self._epsilon_t, grad.dtype),
grad,
indices,
use_locking=self._use_locking)
|
