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
|
# Copyright 2018 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.
# ==============================================================================
"""Experimental `dataset` API for parsing example."""
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 nest
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import sparse_tensor
from tensorflow.python.ops import gen_dataset_ops
from tensorflow.python.ops import parsing_ops
class _ParseExampleDataset(dataset_ops.Dataset):
"""A `Dataset` that parses `example` dataset into a `dict` dataset."""
def __init__(self, input_dataset, features, num_parallel_calls):
super(_ParseExampleDataset, self).__init__()
self._input_dataset = input_dataset
if not all(types == dtypes.string
for types in nest.flatten(input_dataset.output_types)):
raise TypeError("Input dataset should be a dataset of vectors of strings")
self._num_parallel_calls = num_parallel_calls
# pylint: disable=protected-access
self._features = parsing_ops._prepend_none_dimension(features)
# sparse_keys and dense_keys come back sorted here.
(sparse_keys, sparse_types, dense_keys, dense_types, dense_defaults,
dense_shapes) = parsing_ops._features_to_raw_params(
self._features, [
parsing_ops.VarLenFeature, parsing_ops.SparseFeature,
parsing_ops.FixedLenFeature, parsing_ops.FixedLenSequenceFeature
])
# TODO(b/112859642): Pass sparse_index and sparse_values for SparseFeature.
(_, dense_defaults_vec, sparse_keys, sparse_types, dense_keys, dense_shapes,
dense_shape_as_shape) = parsing_ops._process_raw_parameters(
None, dense_defaults, sparse_keys, sparse_types, dense_keys,
dense_types, dense_shapes)
# pylint: enable=protected-access
self._sparse_keys = sparse_keys
self._sparse_types = sparse_types
self._dense_keys = dense_keys
self._dense_defaults = dense_defaults_vec
self._dense_shapes = dense_shapes
self._dense_types = dense_types
dense_output_shapes = [
self._input_dataset.output_shapes.concatenate(shape)
for shape in dense_shape_as_shape
]
sparse_output_shapes = [
self._input_dataset.output_shapes.concatenate([None])
for _ in range(len(sparse_keys))
]
self._output_shapes = dict(
zip(self._dense_keys + self._sparse_keys,
dense_output_shapes + sparse_output_shapes))
self._output_types = dict(
zip(self._dense_keys + self._sparse_keys,
self._dense_types + self._sparse_types))
self._output_classes = dict(
zip(self._dense_keys + self._sparse_keys,
[ops.Tensor for _ in range(len(self._dense_defaults))] +
[sparse_tensor.SparseTensor for _ in range(len(self._sparse_keys))
]))
def _as_variant_tensor(self):
return gen_dataset_ops.parse_example_dataset(
self._input_dataset._as_variant_tensor(), # pylint: disable=protected-access
self._num_parallel_calls,
self._dense_defaults,
self._sparse_keys,
self._dense_keys,
self._sparse_types,
self._dense_shapes,
**dataset_ops.flat_structure(self))
@property
def output_shapes(self):
return self._output_shapes
@property
def output_types(self):
return self._output_types
@property
def output_classes(self):
return self._output_classes
# TODO(b/38416882): Properly export in the `tf.contrib.data` API when stable
# or make private / remove.
# TODO(b/111553342): add arguments names and example names as well.
def parse_example_dataset(features, num_parallel_calls=1):
"""A transformation that parses `Example` protos into a `dict` of tensors.
Parses a number of serialized `Example` protos given in `serialized`. We refer
to `serialized` as a batch with `batch_size` many entries of individual
`Example` protos.
This op parses serialized examples into a dictionary mapping keys to `Tensor`
and `SparseTensor` objects. `features` is a dict from keys to `VarLenFeature`,
`SparseFeature`, and `FixedLenFeature` objects. Each `VarLenFeature`
and `SparseFeature` is mapped to a `SparseTensor`, and each
`FixedLenFeature` is mapped to a `Tensor`. See `tf.parse_example` for more
details about feature dictionaries.
Args:
features: A `dict` mapping feature keys to `FixedLenFeature`,
`VarLenFeature`, and `SparseFeature` values.
num_parallel_calls: (Optional.) A `tf.int32` scalar `tf.Tensor`,
representing the number of parsing processes to call in parallel.
Returns:
A dataset transformation function, which can be passed to
`tf.data.Dataset.apply`.
Raises:
ValueError: if features argument is None.
"""
if features is None:
raise ValueError("Missing: features was %s." % features)
def _apply_fn(dataset):
"""Function from `Dataset` to `Dataset` that applies the transformation."""
out_dataset = _ParseExampleDataset(dataset, features, num_parallel_calls)
if any([
isinstance(feature, parsing_ops.SparseFeature)
for _, feature in features.items()
]):
# pylint: disable=protected-access
# pylint: disable=g-long-lambda
out_dataset = out_dataset.map(
lambda x: parsing_ops._construct_sparse_tensors_for_sparse_features(
features, x), num_parallel_calls=num_parallel_calls)
return out_dataset
return _apply_fn
|
