forked from tensorflow/tensorflow
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathfunctional_ops.py
More file actions
384 lines (324 loc) · 14.4 KB
/
Copy pathfunctional_ops.py
File metadata and controls
384 lines (324 loc) · 14.4 KB
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
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
# 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.
# =============================================================================
"""Functional operations.
## Higher Order Operators
TensorFlow provides several higher order operators to simplify the common
map-reduce programming patterns.
@@map_fn
@@foldl
@@foldr
@@scan
"""
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 array_ops
from tensorflow.python.ops import constant_op
from tensorflow.python.ops import control_flow_ops
from tensorflow.python.ops import tensor_array_ops
from tensorflow.python.ops import variable_scope as vs
# go/tf-wildcard-import
# pylint: disable=wildcard-import
from tensorflow.python.ops.gen_functional_ops import *
# pylint: enable=wildcard-import
# pylint: disable=unused-import
from tensorflow.python.ops.gen_functional_ops import _symbolic_gradient
# pylint: enable=unused-import
# TODO(yuanbyu, mrry): Handle stride to support sliding windows.
def foldl(fn, elems, initializer=None, parallel_iterations=10, back_prop=True,
swap_memory=False, name=None):
"""foldl on the list of tensors unpacked from `elems` on dimension 0.
This foldl operator repeatedly applies the callable `fn` to a sequence
of elements from first to last. The elements are made of the tensors
unpacked from `elems` on dimension 0. The callable fn takes two tensors as
arguments. The first argument is the accumulated value computed from the
preceding invocation of fn. If `initializer` is None, `elems` must contain
at least one element, and its first element is used as the initializer.
Suppose that `elems` is unpacked into `values`, a list of tensors. The shape
of the result tensor is fn(initializer, values[0]).shape`.
Args:
fn: The callable to be performed.
elems: A tensor to be unpacked on dimension 0.
initializer: (optional) The initial value for the accumulator.
parallel_iterations: (optional) The number of iterations allowed to run
in parallel.
back_prop: (optional) True enables back propagation.
swap_memory: (optional) True enables GPU-CPU memory swapping.
name: (optional) Name prefix for the returned tensors.
Returns:
A tensor resulting from applying `fn` consecutively to the list of tensors
unpacked from `elems`, from first to last.
Raises:
TypeError: if `fn` is not callable.
Example:
```python
elems = [1, 2, 3, 4, 5, 6]
sum = foldl(lambda a, x: a + x, elems)
# sum == 21
```
"""
if not callable(fn):
raise TypeError("fn must be callable.")
with ops.op_scope([elems], name, "foldl"):
# Any get_variable calls in fn will cache the first call locally
# and not issue repeated network I/O requests for each iteration.
varscope = vs.get_variable_scope()
varscope_caching_device_was_none = False
if varscope.caching_device is None:
# TODO(ebrevdo): Change to using colocate_with here and in other methods.
varscope.set_caching_device(lambda op: op.device)
varscope_caching_device_was_none = True
# Convert elems to tensor array.
elems = ops.convert_to_tensor(elems, name="elems")
n = array_ops.shape(elems)[0]
elems_ta = tensor_array_ops.TensorArray(dtype=elems.dtype, size=n,
dynamic_size=False,
infer_shape=True)
elems_ta = elems_ta.unpack(elems)
if initializer is None:
a = elems_ta.read(0)
i = constant_op.constant(1)
else:
a = ops.convert_to_tensor(initializer)
i = constant_op.constant(0)
def compute(i, a):
a = fn(a, elems_ta.read(i))
return [i + 1, a]
_, r_a = control_flow_ops.while_loop(
lambda i, a: i < n, compute, [i, a],
parallel_iterations=parallel_iterations,
back_prop=back_prop,
swap_memory=swap_memory)
if varscope_caching_device_was_none:
varscope.set_caching_device(None)
return r_a
def foldr(fn, elems, initializer=None, parallel_iterations=10, back_prop=True,
swap_memory=False, name=None):
"""foldr on the list of tensors unpacked from `elems` on dimension 0.
This foldr operator repeatedly applies the callable `fn` to a sequence
of elements from last to first. The elements are made of the tensors
unpacked from `elems`. The callable fn takes two tensors as arguments.
The first argument is the accumulated value computed from the preceding
invocation of fn. If `initializer` is None, `elems` must contain at least
one element, and its first element is used as the initializer.
Suppose that `elems` is unpacked into `values`, a list of tensors. The shape
of the result tensor is `fn(initializer, values[0]).shape`.
Args:
fn: The callable to be performed.
elems: A tensor that is unpacked into a sequence of tensors to apply `fn`.
initializer: (optional) The initial value for the accumulator.
parallel_iterations: (optional) The number of iterations allowed to run
in parallel.
back_prop: (optional) True enables back propagation.
swap_memory: (optional) True enables GPU-CPU memory swapping.
name: (optional) Name prefix for the returned tensors.
Returns:
A tensor resulting from applying `fn` consecutively to the list of tensors
unpacked from `elems`, from last to first.
Raises:
TypeError: if `fn` is not callable.
Example:
```python
elems = [1, 2, 3, 4, 5, 6]
sum = foldr(lambda a, x: a + x, elems)
# sum == 21
```
"""
if not callable(fn):
raise TypeError("fn must be callable.")
with ops.op_scope([elems], name, "foldr"):
# Any get_variable calls in fn will cache the first call locally
# and not issue repeated network I/O requests for each iteration.
varscope = vs.get_variable_scope()
varscope_caching_device_was_none = False
if varscope.caching_device is None:
# TODO(ebrevdo): Change to using colocate_with here and in other methods.
varscope.set_caching_device(lambda op: op.device)
varscope_caching_device_was_none = True
# Convert elems to tensor array.
elems = ops.convert_to_tensor(elems, name="elems")
n = array_ops.shape(elems)[0]
elems_ta = tensor_array_ops.TensorArray(dtype=elems.dtype, size=n,
dynamic_size=False,
infer_shape=True)
elems_ta = elems_ta.unpack(elems)
if initializer is None:
i = n - 1
a = elems_ta.read(i)
else:
i = n
a = ops.convert_to_tensor(initializer)
def compute(i, a):
i -= 1
a = fn(a, elems_ta.read(i))
return [i, a]
_, r_a = control_flow_ops.while_loop(
lambda i, a: i > 0, compute, [i, a],
parallel_iterations=parallel_iterations,
back_prop=back_prop,
swap_memory=swap_memory)
if varscope_caching_device_was_none:
varscope.set_caching_device(None)
return r_a
def map_fn(fn, elems, dtype=None, parallel_iterations=10, back_prop=True,
swap_memory=False, name=None):
"""map on the list of tensors unpacked from `elems` on dimension 0.
This map operator repeatedly applies the callable `fn` to a sequence of
elements from first to last. The elements are made of the tensors unpacked
from `elems`. `dtype` is the data type of the return value of `fn`. Users
must provide `dtype` if it is different from the data type of `elems`.
Suppose that `elems` is unpacked into `values`, a list of tensors. The shape
of the result tensor is `[len(values)] + fn(values[0]).shape`.
Args:
fn: The callable to be performed.
elems: A tensor to be unpacked to apply `fn`.
dtype: (optional) The output type of `fn`.
parallel_iterations: (optional) The number of iterations allowed to run
in parallel.
back_prop: (optional) True enables back propagation.
swap_memory: (optional) True enables GPU-CPU memory swapping.
name: (optional) Name prefix for the returned tensors.
Returns:
A tensor that packs the results of applying `fn` to the list of tensors
unpacked from `elems`, from first to last.
Raises:
TypeError: if `fn` is not callable.
Example:
```python
elems = [1, 2, 3, 4, 5, 6]
squares = map_fn(lambda x: x * x, elems)
# squares == [1, 4, 9, 16, 25, 36]
```
"""
if not callable(fn):
raise TypeError("fn must be callable.")
with ops.op_scope([elems], name, "map"):
# Any get_variable calls in fn will cache the first call locally
# and not issue repeated network I/O requests for each iteration.
varscope = vs.get_variable_scope()
varscope_caching_device_was_none = False
if varscope.caching_device is None:
# TODO(ebrevdo): Change to using colocate_with here and in other methods.
varscope.set_caching_device(lambda op: op.device)
varscope_caching_device_was_none = True
elems = ops.convert_to_tensor(elems, name="elems")
dtype = dtype if dtype else elems.dtype
# Convert elems to tensor array.
n = array_ops.shape(elems)[0]
elems_ta = tensor_array_ops.TensorArray(dtype=elems.dtype, size=n,
dynamic_size=False,
infer_shape=True)
elems_ta = elems_ta.unpack(elems)
i = constant_op.constant(0)
acc_ta = tensor_array_ops.TensorArray(dtype=dtype, size=n,
dynamic_size=False,
infer_shape=True)
def compute(i, ta):
ta = ta.write(i, fn(elems_ta.read(i)))
return [i + 1, ta]
_, r_a = control_flow_ops.while_loop(
lambda i, a: i < n, compute, [i, acc_ta],
parallel_iterations=parallel_iterations,
back_prop=back_prop,
swap_memory=swap_memory)
result = r_a.pack()
result.set_shape(elems.get_shape().with_rank_at_least(1)[0:1].concatenate(
result.get_shape()[1:]))
if varscope_caching_device_was_none:
varscope.set_caching_device(None)
return result
def scan(fn, elems, initializer=None, parallel_iterations=10, back_prop=True,
swap_memory=False, name=None):
"""scan on the list of tensors unpacked from `elems` on dimension 0.
This scan operator repeatedly applies the callable `fn` to a sequence
of elements from first to last. The elements are made of the tensors
unpacked from `elems` on dimension 0. The callable fn takes two tensors as
arguments. The first argument is the accumulated value computed from the
preceding invocation of fn. If `initializer` is None, `elems` must contain
at least one element, and its first element is used as the initializer.
Suppose that `elems` is unpacked into `values`, a list of tensors. The shape
of the result tensor is `[len(values)] + fn(initializer, values[0]).shape`.
Args:
fn: The callable to be performed.
elems: A tensor to be unpacked on dimension 0.
initializer: (optional) The initial value for the accumulator.
parallel_iterations: (optional) The number of iterations allowed to run
in parallel.
back_prop: (optional) True enables back propagation.
swap_memory: (optional) True enables GPU-CPU memory swapping.
name: (optional) Name prefix for the returned tensors.
Returns:
A tensor that packs the results of applying `fn` to the list of tensors
unpacked from `elems`, from first to last.
Raises:
TypeError: if `fn` is not callable.
Example:
```python
elems = [1, 2, 3, 4, 5, 6]
sum = scan(lambda a, x: a + x, elems)
# sum == [1, 3, 6, 10, 15, 21]
```
"""
if not callable(fn):
raise TypeError("fn must be callable.")
with ops.op_scope([elems], name, "scan"):
# Any get_variable calls in fn will cache the first call locally
# and not issue repeated network I/O requests for each iteration.
varscope = vs.get_variable_scope()
varscope_caching_device_was_none = False
if varscope.caching_device is None:
# TODO(ebrevdo): Change to using colocate_with here and in other methods.
varscope.set_caching_device(lambda op: op.device)
varscope_caching_device_was_none = True
# Convert elems to tensor array.
elems = ops.convert_to_tensor(elems, name="elems")
n = array_ops.shape(elems)[0]
elems_ta = tensor_array_ops.TensorArray(dtype=elems.dtype, size=n,
dynamic_size=False,
infer_shape=True)
elems_ta = elems_ta.unpack(elems)
if initializer is None:
a = elems_ta.read(0)
i = constant_op.constant(1)
else:
a = ops.convert_to_tensor(initializer)
i = constant_op.constant(0)
# Create a tensor array to store the intermediate values.
acc_ta = tensor_array_ops.TensorArray(dtype=a.dtype, size=n,
dynamic_size=False,
infer_shape=True)
if initializer is None:
acc_ta = acc_ta.write(0, a)
def compute(i, a, ta):
a = fn(a, elems_ta.read(i))
ta = ta.write(i, a)
return [i + 1, a, ta]
_, _, r_a = control_flow_ops.while_loop(
lambda i, a, ta: i < n, compute, [i, a, acc_ta],
parallel_iterations=parallel_iterations,
back_prop=back_prop, swap_memory=swap_memory)
result = r_a.pack()
result.set_shape(elems.get_shape().with_rank_at_least(1)[0:1].concatenate(
result.get_shape()[1:]))
if varscope_caching_device_was_none:
varscope.set_caching_device(None)
return result
@ops.RegisterShape("SymbolicGradient")
def _symbolic_gradient_shape(op):
# Say, (u, v) = f(x, y, z), _symbolic_gradient(f) is a function of
# (x, y, z, du, dv) -> (dx, dy, dz). Therefore, shapes of its
# outputs (dx, dy, dz) are the same as (x, y, z).
return [op.inputs[i].get_shape() for i in range(len(op.outputs))]