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
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
|
# Copyright 2022 Google LLC
#
# 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.
"""Tests for the `append_slices` and `merge_appended_slices` custom ops."""
import os
import tensorflow as tf
from fcp.tensorflow import append_slices
from fcp.tensorflow import delete_file
class AppendSlicesTest(tf.test.TestCase):
def new_tempfile_path(self):
"""Returns a path that can be used to store a new tempfile."""
return os.path.join(self.create_tempdir(), 'checkpoint.ckp')
def test_converts_single_element_once_appended_file_to_checkpoint(self):
checkpoint_path = self.new_tempfile_path()
tensor_name = 'a'
tensor = tf.constant(42, dtype=tf.int32)
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[tensor_name],
data=[tensor],
shapes_and_slices=[''])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=[tensor_name],
shape_and_slices=[''],
dtypes=[tf.int32])
self.assertEqual(restored[0], 42)
def test_converts_single_element_twice_appended_file_to_checkpoint(self):
checkpoint_path = self.new_tempfile_path()
tensor_names = ['a', 'b']
tensor_values = [tf.constant(x, dtype=tf.int32) for x in (7, 11)]
for (tensor_name, tensor_value) in zip(tensor_names, tensor_values):
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[tensor_name],
data=[tensor_value],
shapes_and_slices=[''])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=tensor_names,
shape_and_slices=[''] * 2,
dtypes=[tf.int32] * 2)
self.assertEqual(restored[0], 7)
self.assertEqual(restored[1], 11)
def test_converts_two_element_once_appended_file_to_checkpoint(self):
checkpoint_path = self.new_tempfile_path()
tensors = [('a', 16), ('b', 17)]
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[name for (name, value) in tensors],
data=[tf.constant(value, tf.int32) for (name, value) in tensors],
shapes_and_slices=['' for _ in tensors])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=['a', 'b'],
shape_and_slices=[''] * 2,
dtypes=[tf.int32] * 2)
self.assertEqual(restored[0], 16)
self.assertEqual(restored[1], 17)
def test_converts_two_element_multi_twice_appended_file_to_checkpoint(self):
# Note: the interleaved ordering ensures that the resulting merged
# checkpoint is able to mix together the two input checkpoints properly.
checkpoint_path = self.new_tempfile_path()
tensors = [
[('a', 12), ('c', 55)],
[('b', 40), ('d', 88)],
]
for tensors_for_checkpoint in tensors:
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[name for (name, value) in tensors_for_checkpoint],
data=[
tf.constant(value, tf.int32)
for (name, value) in tensors_for_checkpoint
],
shapes_and_slices=['' for _ in tensors_for_checkpoint])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=['a', 'b', 'c', 'd'],
shape_and_slices=[''] * 4,
dtypes=[tf.int32] * 4)
self.assertEqual(restored[0], 12)
self.assertEqual(restored[1], 40)
self.assertEqual(restored[2], 55)
self.assertEqual(restored[3], 88)
def test_converts_nonalphabetical_two_element_multi_twice_appended_file_to_checkpoint(
self):
# Note: the interleaved ordering ensures that the resulting merged
# checkpoint is able to mix together the two input checkpoints properly.
checkpoint_path = self.new_tempfile_path()
tensors = [
[('b', 12), ('a', 55)],
[('d', 40), ('c', 88)],
]
for tensors_for_checkpoint in tensors:
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[name for (name, value) in tensors_for_checkpoint],
data=[
tf.constant(value, tf.int32)
for (name, value) in tensors_for_checkpoint
],
shapes_and_slices=['' for _ in tensors_for_checkpoint])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=['d', 'c', 'b', 'a'],
shape_and_slices=[''] * 4,
dtypes=[tf.int32] * 4)
self.assertEqual(restored[0], 40)
self.assertEqual(restored[1], 88)
self.assertEqual(restored[2], 12)
self.assertEqual(restored[3], 55)
def test_merge_missing_checkpoint_file_raises(self):
checkpoint_path = self.new_tempfile_path()
with self.assertRaises(tf.errors.NotFoundError):
append_slices.merge_appended_slices(checkpoint_path)
def test_duplicate_named_tensor_raises(self):
checkpoint_path = self.new_tempfile_path()
tensor_values = [tf.constant(x, dtype=tf.int32) for x in (7, 11)]
for tensor_value in tensor_values:
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=['a'],
data=[tensor_value],
shapes_and_slices=[''])
with self.assertRaisesRegex(
tf.errors.InvalidArgumentError,
'Attempted to merge two checkpoint entries for slice name: `a`'):
append_slices.merge_appended_slices(checkpoint_path)
def test_append_and_merge_using_same_filename(self):
checkpoint_path = self.new_tempfile_path()
for _ in range(2):
# Without calling this we might append to a previously used file.
delete_file.delete_file(checkpoint_path)
tensor_names = ['a', 'b']
tensor_values = [tf.constant(x, dtype=tf.int32) for x in (7, 11)]
for (tensor_name, tensor_value) in zip(tensor_names, tensor_values):
append_slices.append_slices(
filename=checkpoint_path,
tensor_names=[tensor_name],
data=[tensor_value],
shapes_and_slices=[''])
append_slices.merge_appended_slices(checkpoint_path)
restored = tf.raw_ops.RestoreV2(
prefix=checkpoint_path,
tensor_names=tensor_names,
shape_and_slices=[''] * 2,
dtypes=[tf.int32] * 2)
self.assertEqual(restored[0], 7)
self.assertEqual(restored[1], 11)
if __name__ == '__main__':
tf.test.main()
|
