-
Notifications
You must be signed in to change notification settings - Fork 74k
/
readers.py
496 lines (417 loc) · 20 KB
/
readers.py
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
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
# Copyright 2017 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.
# ==============================================================================
"""Python wrappers for reader Datasets."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from tensorflow.python.data.ops import dataset_ops
from tensorflow.python.data.util import convert
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import tensor_shape
from tensorflow.python.framework import tensor_spec
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import gen_dataset_ops
from tensorflow.python.ops import gen_experimental_dataset_ops as ged_ops
from tensorflow.python.util.tf_export import tf_export
# TODO(b/64974358): Increase default buffer size to 256 MB.
_DEFAULT_READER_BUFFER_SIZE_BYTES = 256 * 1024 # 256 KB
def _create_or_validate_filenames_dataset(filenames):
"""Creates (or validates) a dataset of filenames.
Args:
filenames: Either a list or dataset of filenames. If it is a list, it is
convert to a dataset. If it is a dataset, its type and shape is validated.
Returns:
A dataset of filenames.
"""
if isinstance(filenames, dataset_ops.DatasetV2):
if dataset_ops.get_legacy_output_types(filenames) != dtypes.string:
raise TypeError(
"`filenames` must be a `tf.data.Dataset` of `tf.string` elements.")
if not dataset_ops.get_legacy_output_shapes(filenames).is_compatible_with(
tensor_shape.TensorShape([])):
raise TypeError(
"`filenames` must be a `tf.data.Dataset` of scalar `tf.string` "
"elements.")
else:
filenames = ops.convert_to_tensor(filenames, dtype=dtypes.string)
filenames = array_ops.reshape(filenames, [-1], name="flat_filenames")
filenames = dataset_ops.DatasetV2.from_tensor_slices(filenames)
return filenames
def _create_dataset_reader(dataset_creator, filenames, num_parallel_reads=None):
"""Creates a dataset that reads the given files using the given reader.
Args:
dataset_creator: A function that takes in a single file name and returns a
dataset.
filenames: A `tf.data.Dataset` containing one or more filenames.
num_parallel_reads: The number of parallel reads we should do.
Returns:
A `Dataset` that reads data from `filenames`.
"""
def read_one_file(filename):
filename = ops.convert_to_tensor(filename, dtypes.string, name="filename")
return dataset_creator(filename)
if num_parallel_reads is None:
return filenames.flat_map(read_one_file)
else:
return ParallelInterleaveDataset(
filenames, read_one_file, cycle_length=num_parallel_reads,
block_length=1, sloppy=False, buffer_output_elements=None,
prefetch_input_elements=None)
class _TextLineDataset(dataset_ops.DatasetSource):
"""A `Dataset` comprising records from one or more text files."""
def __init__(self, filenames, compression_type=None, buffer_size=None):
"""Creates a `TextLineDataset`.
Args:
filenames: A `tf.string` tensor containing one or more filenames.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
buffer_size: (Optional.) A `tf.int64` scalar denoting the number of bytes
to buffer. A value of 0 results in the default buffering values chosen
based on the compression type.
"""
self._filenames = filenames
self._compression_type = convert.optional_param_to_tensor(
"compression_type",
compression_type,
argument_default="",
argument_dtype=dtypes.string)
self._buffer_size = convert.optional_param_to_tensor(
"buffer_size",
buffer_size,
argument_default=_DEFAULT_READER_BUFFER_SIZE_BYTES)
variant_tensor = gen_dataset_ops.text_line_dataset(
self._filenames, self._compression_type, self._buffer_size)
super(_TextLineDataset, self).__init__(variant_tensor)
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
@tf_export("data.TextLineDataset", v1=[])
class TextLineDatasetV2(dataset_ops.DatasetSource):
"""A `Dataset` comprising lines from one or more text files."""
def __init__(self, filenames, compression_type=None, buffer_size=None,
num_parallel_reads=None):
"""Creates a `TextLineDataset`.
Args:
filenames: A `tf.string` tensor or `tf.data.Dataset` containing one or
more filenames.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
buffer_size: (Optional.) A `tf.int64` scalar denoting the number of bytes
to buffer. A value of 0 results in the default buffering values chosen
based on the compression type.
num_parallel_reads: (Optional.) A `tf.int64` scalar representing the
number of files to read in parallel. If greater than one, the records of
files read in parallel are outputted in an interleaved order. If your
input pipeline is I/O bottlenecked, consider setting this parameter to a
value greater than one to parallelize the I/O. If `None`, files will be
read sequentially.
"""
filenames = _create_or_validate_filenames_dataset(filenames)
self._filenames = filenames
self._compression_type = compression_type
self._buffer_size = buffer_size
def creator_fn(filename):
return _TextLineDataset(filename, compression_type, buffer_size)
self._impl = _create_dataset_reader(creator_fn, filenames,
num_parallel_reads)
variant_tensor = self._impl._variant_tensor # pylint: disable=protected-access
super(TextLineDatasetV2, self).__init__(variant_tensor)
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
@tf_export(v1=["data.TextLineDataset"])
class TextLineDatasetV1(dataset_ops.DatasetV1Adapter):
"""A `Dataset` comprising lines from one or more text files."""
def __init__(self, filenames, compression_type=None, buffer_size=None,
num_parallel_reads=None):
wrapped = TextLineDatasetV2(filenames, compression_type, buffer_size,
num_parallel_reads)
super(TextLineDatasetV1, self).__init__(wrapped)
__init__.__doc__ = TextLineDatasetV2.__init__.__doc__
@property
def _filenames(self):
return self._dataset._filenames # pylint: disable=protected-access
@_filenames.setter
def _filenames(self, value):
self._dataset._filenames = value # pylint: disable=protected-access
class _TFRecordDataset(dataset_ops.DatasetSource):
"""A `Dataset` comprising records from one or more TFRecord files."""
def __init__(self, filenames, compression_type=None, buffer_size=None):
"""Creates a `TFRecordDataset`.
Args:
filenames: A `tf.string` tensor containing one or more filenames.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
buffer_size: (Optional.) A `tf.int64` scalar representing the number of
bytes in the read buffer. 0 means no buffering.
"""
self._filenames = filenames
self._compression_type = convert.optional_param_to_tensor(
"compression_type",
compression_type,
argument_default="",
argument_dtype=dtypes.string)
self._buffer_size = convert.optional_param_to_tensor(
"buffer_size",
buffer_size,
argument_default=_DEFAULT_READER_BUFFER_SIZE_BYTES)
variant_tensor = gen_dataset_ops.tf_record_dataset(
self._filenames, self._compression_type, self._buffer_size)
super(_TFRecordDataset, self).__init__(variant_tensor)
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
class ParallelInterleaveDataset(dataset_ops.UnaryDataset):
"""A `Dataset` that maps a function over its input and flattens the result."""
def __init__(self, input_dataset, map_func, cycle_length, block_length,
sloppy, buffer_output_elements, prefetch_input_elements):
"""See `tf.data.experimental.parallel_interleave()` for details."""
self._input_dataset = input_dataset
self._map_func = dataset_ops.StructuredFunctionWrapper(
map_func, self._transformation_name(), dataset=input_dataset)
if not isinstance(self._map_func.output_structure, dataset_ops.DatasetSpec):
raise TypeError("`map_func` must return a `Dataset` object.")
self._element_spec = self._map_func.output_structure._element_spec # pylint: disable=protected-access
self._cycle_length = ops.convert_to_tensor(
cycle_length, dtype=dtypes.int64, name="cycle_length")
self._block_length = ops.convert_to_tensor(
block_length, dtype=dtypes.int64, name="block_length")
self._sloppy = ops.convert_to_tensor(
sloppy, dtype=dtypes.bool, name="sloppy")
self._buffer_output_elements = convert.optional_param_to_tensor(
"buffer_output_elements",
buffer_output_elements,
argument_default=2 * block_length)
self._prefetch_input_elements = convert.optional_param_to_tensor(
"prefetch_input_elements",
prefetch_input_elements,
argument_default=2 * cycle_length)
variant_tensor = ged_ops.parallel_interleave_dataset(
self._input_dataset._variant_tensor, # pylint: disable=protected-access
self._map_func.function.captured_inputs,
self._cycle_length,
self._block_length,
self._sloppy,
self._buffer_output_elements,
self._prefetch_input_elements,
f=self._map_func.function,
**self._flat_structure)
super(ParallelInterleaveDataset, self).__init__(input_dataset,
variant_tensor)
def _functions(self):
return [self._map_func]
@property
def element_spec(self):
return self._element_spec
def _transformation_name(self):
return "tf.data.experimental.parallel_interleave()"
@tf_export("data.TFRecordDataset", v1=[])
class TFRecordDatasetV2(dataset_ops.DatasetV2):
"""A `Dataset` comprising records from one or more TFRecord files."""
def __init__(self, filenames, compression_type=None, buffer_size=None,
num_parallel_reads=None):
"""Creates a `TFRecordDataset` to read one or more TFRecord files.
Args:
filenames: A `tf.string` tensor or `tf.data.Dataset` containing one or
more filenames.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
buffer_size: (Optional.) A `tf.int64` scalar representing the number of
bytes in the read buffer. If your input pipeline is I/O bottlenecked,
consider setting this parameter to a value 1-100 MBs. If `None`, a
sensible default for both local and remote file systems is used.
num_parallel_reads: (Optional.) A `tf.int64` scalar representing the
number of files to read in parallel. If greater than one, the records of
files read in parallel are outputted in an interleaved order. If your
input pipeline is I/O bottlenecked, consider setting this parameter to a
value greater than one to parallelize the I/O. If `None`, files will be
read sequentially.
Raises:
TypeError: If any argument does not have the expected type.
ValueError: If any argument does not have the expected shape.
"""
filenames = _create_or_validate_filenames_dataset(filenames)
self._filenames = filenames
self._compression_type = compression_type
self._buffer_size = buffer_size
self._num_parallel_reads = num_parallel_reads
def creator_fn(filename):
return _TFRecordDataset(filename, compression_type, buffer_size)
self._impl = _create_dataset_reader(creator_fn, filenames,
num_parallel_reads)
variant_tensor = self._impl._variant_tensor # pylint: disable=protected-access
super(TFRecordDatasetV2, self).__init__(variant_tensor)
def _clone(self,
filenames=None,
compression_type=None,
buffer_size=None,
num_parallel_reads=None):
return TFRecordDatasetV2(filenames or self._filenames,
compression_type or self._compression_type,
buffer_size or self._buffer_size,
num_parallel_reads or self._num_parallel_reads)
def _inputs(self):
return self._impl._inputs() # pylint: disable=protected-access
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
@tf_export(v1=["data.TFRecordDataset"])
class TFRecordDatasetV1(dataset_ops.DatasetV1Adapter):
"""A `Dataset` comprising records from one or more TFRecord files."""
def __init__(self, filenames, compression_type=None, buffer_size=None,
num_parallel_reads=None):
wrapped = TFRecordDatasetV2(
filenames, compression_type, buffer_size, num_parallel_reads)
super(TFRecordDatasetV1, self).__init__(wrapped)
__init__.__doc__ = TFRecordDatasetV2.__init__.__doc__
def _clone(self,
filenames=None,
compression_type=None,
buffer_size=None,
num_parallel_reads=None):
# pylint: disable=protected-access
return TFRecordDatasetV1(
filenames or self._dataset._filenames,
compression_type or self._dataset._compression_type,
buffer_size or self._dataset._buffer_size,
num_parallel_reads or self._dataset._num_parallel_reads)
@property
def _filenames(self):
return self._dataset._filenames # pylint: disable=protected-access
@_filenames.setter
def _filenames(self, value):
self._dataset._filenames = value # pylint: disable=protected-access
class _FixedLengthRecordDataset(dataset_ops.DatasetSource):
"""A `Dataset` of fixed-length records from one or more binary files."""
def __init__(self,
filenames,
record_bytes,
header_bytes=None,
footer_bytes=None,
buffer_size=None,
compression_type=None):
"""Creates a `FixedLengthRecordDataset`.
Args:
filenames: A `tf.string` tensor containing one or more filenames.
record_bytes: A `tf.int64` scalar representing the number of bytes in
each record.
header_bytes: (Optional.) A `tf.int64` scalar representing the number of
bytes to skip at the start of a file.
footer_bytes: (Optional.) A `tf.int64` scalar representing the number of
bytes to ignore at the end of a file.
buffer_size: (Optional.) A `tf.int64` scalar representing the number of
bytes to buffer when reading.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
"""
self._filenames = filenames
self._record_bytes = ops.convert_to_tensor(
record_bytes, dtype=dtypes.int64, name="record_bytes")
self._header_bytes = convert.optional_param_to_tensor(
"header_bytes", header_bytes)
self._footer_bytes = convert.optional_param_to_tensor(
"footer_bytes", footer_bytes)
self._buffer_size = convert.optional_param_to_tensor(
"buffer_size", buffer_size, _DEFAULT_READER_BUFFER_SIZE_BYTES)
self._compression_type = convert.optional_param_to_tensor(
"compression_type",
compression_type,
argument_default="",
argument_dtype=dtypes.string)
variant_tensor = gen_dataset_ops.fixed_length_record_dataset_v2(
self._filenames, self._header_bytes, self._record_bytes,
self._footer_bytes, self._buffer_size, self._compression_type)
super(_FixedLengthRecordDataset, self).__init__(variant_tensor)
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
@tf_export("data.FixedLengthRecordDataset", v1=[])
class FixedLengthRecordDatasetV2(dataset_ops.DatasetSource):
"""A `Dataset` of fixed-length records from one or more binary files."""
def __init__(self,
filenames,
record_bytes,
header_bytes=None,
footer_bytes=None,
buffer_size=None,
compression_type=None,
num_parallel_reads=None):
"""Creates a `FixedLengthRecordDataset`.
Args:
filenames: A `tf.string` tensor or `tf.data.Dataset` containing one or
more filenames.
record_bytes: A `tf.int64` scalar representing the number of bytes in
each record.
header_bytes: (Optional.) A `tf.int64` scalar representing the number of
bytes to skip at the start of a file.
footer_bytes: (Optional.) A `tf.int64` scalar representing the number of
bytes to ignore at the end of a file.
buffer_size: (Optional.) A `tf.int64` scalar representing the number of
bytes to buffer when reading.
compression_type: (Optional.) A `tf.string` scalar evaluating to one of
`""` (no compression), `"ZLIB"`, or `"GZIP"`.
num_parallel_reads: (Optional.) A `tf.int64` scalar representing the
number of files to read in parallel. If greater than one, the records of
files read in parallel are outputted in an interleaved order. If your
input pipeline is I/O bottlenecked, consider setting this parameter to a
value greater than one to parallelize the I/O. If `None`, files will be
read sequentially.
"""
filenames = _create_or_validate_filenames_dataset(filenames)
self._filenames = filenames
self._record_bytes = record_bytes
self._header_bytes = header_bytes
self._footer_bytes = footer_bytes
self._buffer_size = buffer_size
self._compression_type = compression_type
def creator_fn(filename):
return _FixedLengthRecordDataset(filename, record_bytes, header_bytes,
footer_bytes, buffer_size,
compression_type)
self._impl = _create_dataset_reader(creator_fn, filenames,
num_parallel_reads)
variant_tensor = self._impl._variant_tensor # pylint: disable=protected-access
super(FixedLengthRecordDatasetV2, self).__init__(variant_tensor)
@property
def element_spec(self):
return tensor_spec.TensorSpec([], dtypes.string)
@tf_export(v1=["data.FixedLengthRecordDataset"])
class FixedLengthRecordDatasetV1(dataset_ops.DatasetV1Adapter):
"""A `Dataset` of fixed-length records from one or more binary files."""
def __init__(self,
filenames,
record_bytes,
header_bytes=None,
footer_bytes=None,
buffer_size=None,
compression_type=None,
num_parallel_reads=None):
wrapped = FixedLengthRecordDatasetV2(
filenames, record_bytes, header_bytes, footer_bytes, buffer_size,
compression_type, num_parallel_reads)
super(FixedLengthRecordDatasetV1, self).__init__(wrapped)
__init__.__doc__ = FixedLengthRecordDatasetV2.__init__.__doc__
@property
def _filenames(self):
return self._dataset._filenames # pylint: disable=protected-access
@_filenames.setter
def _filenames(self, value):
self._dataset._filenames = value # pylint: disable=protected-access
# TODO(b/119044825): Until all `tf.data` unit tests are converted to V2, keep
# these aliases in place.
FixedLengthRecordDataset = FixedLengthRecordDatasetV1
TFRecordDataset = TFRecordDatasetV1
TextLineDataset = TextLineDatasetV1