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
|
# 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.
# ==============================================================================
"""Adagrad for TensorFlow."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from tensorflow.python.framework import constant_op
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 AdagradOptimizer(optimizer.Optimizer):
"""Optimizer that implements the Adagrad algorithm.
See this [paper](http://www.jmlr.org/papers/volume12/duchi11a/duchi11a.pdf)
or this
[intro](http://cs.stanford.edu/~ppasupat/a9online/uploads/proximal_notes.pdf).
"""
def __init__(self, learning_rate, initial_accumulator_value=0.1,
use_locking=False, name="Adagrad"):
"""Construct a new Adagrad optimizer.
Args:
learning_rate: A `Tensor` or a floating point value. The learning rate.
initial_accumulator_value: A floating point value.
Starting value for the accumulators, must be positive.
use_locking: If `True` use locks for update operations.
name: Optional name prefix for the operations created when applying
gradients. Defaults to "Adagrad".
Raises:
ValueError: If the `initial_accumulator_value` is invalid.
"""
if initial_accumulator_value <= 0.0:
raise ValueError("initial_accumulator_value must be positive: %s" %
initial_accumulator_value)
super(AdagradOptimizer, self).__init__(use_locking, name)
self._learning_rate = learning_rate
self._initial_accumulator_value = initial_accumulator_value
# Created in Initialize.
self._learning_rate_tensor = None
def _create_slots(self, var_list):
for v in var_list:
with ops.colocate_with(v):
val = constant_op.constant(self._initial_accumulator_value,
shape=v.get_shape(),
dtype=v.dtype.base_dtype)
self._get_or_make_slot(v, val, "accumulator", self._name)
def _prepare(self):
self._learning_rate_tensor = ops.convert_to_tensor(self._learning_rate,
name="learning_rate")
def _apply_dense(self, grad, var):
acc = self.get_slot(var, "accumulator")
return training_ops.apply_adagrad(
var,
acc,
math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
grad,
use_locking=self._use_locking)
def _resource_apply_dense(self, grad, var):
acc = self.get_slot(var, "accumulator")
return training_ops.resource_apply_adagrad(
var.handle,
acc.handle,
math_ops.cast(self._learning_rate_tensor, grad.dtype.base_dtype),
grad,
use_locking=self._use_locking)
def _apply_sparse(self, grad, var):
acc = self.get_slot(var, "accumulator")
return training_ops.sparse_apply_adagrad(
var,
acc,
math_ops.cast(self._learning_rate_tensor, var.dtype.base_dtype),
grad.values,
grad.indices,
use_locking=self._use_locking)
def _resource_apply_sparse(self, grad, var, indices):
acc = self.get_slot(var, "accumulator")
return training_ops.resource_sparse_apply_adagrad(
var.handle,
acc.handle,
math_ops.cast(self._learning_rate_tensor, grad.dtype),
grad,
indices,
use_locking=self._use_locking)
|
