/
serializers.py
392 lines (294 loc) · 12.9 KB
/
serializers.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
# Copyright Amazon.com, Inc. or its affiliates. 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. A copy of
# the License is located at
#
# http://aws.amazon.com/apache2.0/
#
# or in the "license" file accompanying this file. This file 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.
"""Implements methods for serializing data for an inference endpoint."""
from __future__ import absolute_import
import abc
from collections.abc import Iterable
import csv
import io
import json
import numpy as np
from six import with_metaclass
from sagemaker.utils import DeferredError
try:
import scipy.sparse
except ImportError as e:
scipy = DeferredError(e)
class BaseSerializer(abc.ABC):
"""Abstract base class for creation of new serializers.
Provides a skeleton for customization requiring the overriding of the method
serialize and the class attribute CONTENT_TYPE.
"""
@abc.abstractmethod
def serialize(self, data):
"""Serialize data into the media type specified by CONTENT_TYPE.
Args:
data (object): Data to be serialized.
Returns:
object: Serialized data used for a request.
"""
@property
@abc.abstractmethod
def CONTENT_TYPE(self):
"""The MIME type of the data sent to the inference endpoint."""
class SimpleBaseSerializer(with_metaclass(abc.ABCMeta, BaseSerializer)):
"""Abstract base class for creation of new serializers.
This class extends the API of :class:~`sagemaker.serializers.BaseSerializer` with more
user-friendly options for setting the Content-Type header, in situations where it can be
provided at init and freely updated.
"""
def __init__(self, content_type="application/json"):
"""Initialize a ``SimpleBaseSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "application/json").
"""
super(SimpleBaseSerializer, self).__init__()
if not isinstance(content_type, str):
raise ValueError(
"content_type must be a string specifying the MIME type of the data sent in "
"requests: e.g. 'application/json', 'text/csv', etc. Got %s" % content_type
)
self.content_type = content_type
@property
def CONTENT_TYPE(self):
"""The data MIME type set in the Content-Type header on prediction endpoint requests."""
return self.content_type
class CSVSerializer(SimpleBaseSerializer):
"""Serialize data of various formats to a CSV-formatted string."""
def __init__(self, content_type="text/csv"):
"""Initialize a ``CSVSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "text/csv").
"""
super(CSVSerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Serialize data of various formats to a CSV-formatted string.
Args:
data (object): Data to be serialized. Can be a NumPy array, list,
file, or buffer.
Returns:
str: The data serialized as a CSV-formatted string.
"""
if hasattr(data, "read"):
return data.read()
is_mutable_sequence_like = self._is_sequence_like(data) and hasattr(data, "__setitem__")
has_multiple_rows = len(data) > 0 and self._is_sequence_like(data[0])
if is_mutable_sequence_like and has_multiple_rows:
return "\n".join([self._serialize_row(row) for row in data])
return self._serialize_row(data)
def _serialize_row(self, data):
"""Serialize data as a CSV-formatted row.
Args:
data (object): Data to be serialized in a row.
Returns:
str: The data serialized as a CSV-formatted row.
"""
if isinstance(data, str):
return data
if isinstance(data, np.ndarray):
data = np.ndarray.flatten(data)
if hasattr(data, "__len__"):
if len(data) == 0:
raise ValueError("Cannot serialize empty array")
csv_buffer = io.StringIO()
csv_writer = csv.writer(csv_buffer, delimiter=",")
csv_writer.writerow(data)
return csv_buffer.getvalue().rstrip("\r\n")
raise ValueError("Unable to handle input format: %s" % type(data))
def _is_sequence_like(self, data):
"""Returns true if obj is iterable and subscriptable."""
return hasattr(data, "__iter__") and hasattr(data, "__getitem__")
class NumpySerializer(SimpleBaseSerializer):
"""Serialize data to a buffer using the .npy format."""
def __init__(self, dtype=None, content_type="application/x-npy"):
"""Initialize a ``NumpySerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "application/x-npy").
dtype (str): The dtype of the data.
"""
super(NumpySerializer, self).__init__(content_type=content_type)
self.dtype = dtype
def serialize(self, data):
"""Serialize data to a buffer using the .npy format.
Args:
data (object): Data to be serialized. Can be a NumPy array, list,
file, or buffer.
Returns:
io.BytesIO: A buffer containing data serialzied in the .npy format.
"""
if isinstance(data, np.ndarray):
if data.size == 0:
raise ValueError("Cannot serialize empty array.")
return self._serialize_array(data)
if isinstance(data, list):
if len(data) == 0:
raise ValueError("Cannot serialize empty array.")
return self._serialize_array(np.array(data, self.dtype))
# files and buffers. Assumed to hold npy-formatted data.
if hasattr(data, "read"):
return data.read()
return self._serialize_array(np.array(data))
def _serialize_array(self, array):
"""Saves a NumPy array in a buffer.
Args:
array (numpy.ndarray): The array to serialize.
Returns:
io.BytesIO: A buffer containing the serialized array.
"""
buffer = io.BytesIO()
np.save(buffer, array)
return buffer.getvalue()
class JSONSerializer(SimpleBaseSerializer):
"""Serialize data to a JSON formatted string."""
def serialize(self, data):
"""Serialize data of various formats to a JSON formatted string.
Args:
data (object): Data to be serialized.
Returns:
str: The data serialized as a JSON string.
"""
if isinstance(data, dict):
return json.dumps(
{
key: value.tolist() if isinstance(value, np.ndarray) else value
for key, value in data.items()
}
)
if hasattr(data, "read"):
return data.read()
if isinstance(data, np.ndarray):
return json.dumps(data.tolist())
return json.dumps(data)
class IdentitySerializer(SimpleBaseSerializer):
"""Serialize data by returning data without modification.
This serializer may be useful if, for example, you're sending raw bytes such as from an image
file's .read() method.
"""
def __init__(self, content_type="application/octet-stream"):
"""Initialize an ``IdentitySerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "application/octet-stream").
"""
super(IdentitySerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Return data without modification.
Args:
data (object): Data to be serialized.
Returns:
object: The unmodified data.
"""
return data
class JSONLinesSerializer(SimpleBaseSerializer):
"""Serialize data to a JSON Lines formatted string."""
def __init__(self, content_type="application/jsonlines"):
"""Initialize a ``JSONLinesSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "application/jsonlines").
"""
super(JSONLinesSerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Serialize data of various formats to a JSON Lines formatted string.
Args:
data (object): Data to be serialized. The data can be a string,
iterable of JSON serializable objects, or a file-like object.
Returns:
str: The data serialized as a string containing newline-separated
JSON values.
"""
if isinstance(data, str):
return data
if hasattr(data, "read"):
return data.read()
if isinstance(data, Iterable):
return "\n".join(json.dumps(element) for element in data)
raise ValueError("Object of type %s is not JSON Lines serializable." % type(data))
class SparseMatrixSerializer(SimpleBaseSerializer):
"""Serialize a sparse matrix to a buffer using the .npz format."""
def __init__(self, content_type="application/x-npz"):
"""Initialize a ``SparseMatrixSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "application/x-npz").
"""
super(SparseMatrixSerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Serialize a sparse matrix to a buffer using the .npz format.
Sparse matrices can be in the ``csc``, ``csr``, ``bsr``, ``dia`` or
``coo`` formats.
Args:
data (scipy.sparse.spmatrix): The sparse matrix to serialize.
Returns:
io.BytesIO: A buffer containing the serialized sparse matrix.
"""
buffer = io.BytesIO()
scipy.sparse.save_npz(buffer, data)
return buffer.getvalue()
class LibSVMSerializer(SimpleBaseSerializer):
"""Serialize data of various formats to a LibSVM-formatted string.
The data must already be in LIBSVM file format:
<label> <index1>:<value1> <index2>:<value2> ...
It is suitable for sparse datasets since it does not store zero-valued
features.
"""
def __init__(self, content_type="text/libsvm"):
"""Initialize a ``LibSVMSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "text/libsvm").
"""
super(LibSVMSerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Serialize data of various formats to a LibSVM-formatted string.
Args:
data (object): Data to be serialized. Can be a string or a
file-like object.
Returns:
str: The data serialized as a LibSVM-formatted string.
"""
if isinstance(data, str):
return data
if hasattr(data, "read"):
return data.read()
raise ValueError("Unable to handle input format: %s" % type(data))
class DataSerializer(SimpleBaseSerializer):
"""Serialize data in any file by extracting raw bytes from the file."""
def __init__(self, content_type="file-path/raw-bytes"):
"""Initialize a ``DataSerializer`` instance.
Args:
content_type (str): The MIME type to signal to the inference endpoint when sending
request data (default: "file-path/raw-bytes").
"""
super(DataSerializer, self).__init__(content_type=content_type)
def serialize(self, data):
"""Serialize file data to a raw bytes.
Args:
data (object): Data to be serialized. The data can be a string
representing file-path or the raw bytes from a file.
Returns:
raw-bytes: The data serialized as raw-bytes from the input.
"""
if isinstance(data, str):
try:
with open(data, "rb") as data_file:
data_file_info = data_file.read()
return data_file_info
except Exception as e:
raise ValueError(f"Could not open/read file: {data}. {e}")
if isinstance(data, bytes):
return data
raise ValueError(f"Object of type {type(data)} is not Data serializable.")