blib/lib/Paws/SageMaker/InputConfig.pm untested

filter: Statementstmt 0.0% 0 / 7 Branchbran n/a Conditioncond n/a MC/DCmcdc n/a Subroutinesub 0.0% 0 / 1 Podpod n/a totaltotal 0.0% 0 / 8 CC 1 SCAR 6.9
CC
1
Cov
0%
BEGIN
2.0
CRAP 2.0
1 # Generated by default/object.tt
2 package Paws::SageMaker::InputConfig;
3 use Moose;
4 has DataInputConfig => (is => 'ro', isa => 'Str', required => 1);
5 has Framework => (is => 'ro', isa => 'Str', required => 1);
6 has FrameworkVersion => (is => 'ro', isa => 'Str');
7 has S3Uri => (is => 'ro', isa => 'Str', required => 1);
8  
9 1;
10  
11 ### main pod documentation begin ###
12  
13 =head1 NAME
14  
15 Paws::SageMaker::InputConfig
16  
17 =head1 USAGE
18  
19 This class represents one of two things:
20  
21 =head3 Arguments in a call to a service
22  
23 Use the attributes of this class as arguments to methods. You shouldn't make instances of this class.
24 Each attribute should be used as a named argument in the calls that expect this type of object.
25  
26 As an example, if Att1 is expected to be a Paws::SageMaker::InputConfig object:
27  
28 $service_obj->Method(Att1 => { DataInputConfig => $value, ..., S3Uri => $value });
29  
30 =head3 Results returned from an API call
31  
32 Use accessors for each attribute. If Att1 is expected to be an Paws::SageMaker::InputConfig object:
33  
34 $result = $service_obj->Method(...);
35 $result->Att1->DataInputConfig
36  
37 =head1 DESCRIPTION
38  
39 Contains information about the location of input model artifacts, the
40 name and shape of the expected data inputs, and the framework in which
41 the model was trained.
42  
43 =head1 ATTRIBUTES
44  
45  
46 =head2 B<REQUIRED> DataInputConfig => Str
47  
48 Specifies the name and shape of the expected data inputs for your
49 trained model with a JSON dictionary form. The data inputs are
50 InputConfig$Framework specific.
51  
52 =over
53  
54 =item *
55  
56 C<TensorFlow>: You must specify the name and shape (NHWC format) of the
57 expected data inputs using a dictionary format for your trained model.
58 The dictionary formats required for the console and CLI are different.
59  
60 =over
61  
62 =item *
63  
64 Examples for one input:
65  
66 =over
67  
68 =item *
69  
70 If using the console, C<{"input":[1,1024,1024,3]}>
71  
72 =item *
73  
74 If using the CLI, C<{\"input\":[1,1024,1024,3]}>
75  
76 =back
77  
78 =item *
79  
80 Examples for two inputs:
81  
82 =over
83  
84 =item *
85  
86 If using the console, C<{"data1": [1,28,28,1], "data2":[1,28,28,1]}>
87  
88 =item *
89  
90 If using the CLI, C<{\"data1\": [1,28,28,1], \"data2\":[1,28,28,1]}>
91  
92 =back
93  
94 =back
95  
96 =item *
97  
98 C<KERAS>: You must specify the name and shape (NCHW format) of expected
99 data inputs using a dictionary format for your trained model. Note that
100 while Keras model artifacts should be uploaded in NHWC (channel-last)
101 format, C<DataInputConfig> should be specified in NCHW (channel-first)
102 format. The dictionary formats required for the console and CLI are
103 different.
104  
105 =over
106  
107 =item *
108  
109 Examples for one input:
110  
111 =over
112  
113 =item *
114  
115 If using the console, C<{"input_1":[1,3,224,224]}>
116  
117 =item *
118  
119 If using the CLI, C<{\"input_1\":[1,3,224,224]}>
120  
121 =back
122  
123 =item *
124  
125 Examples for two inputs:
126  
127 =over
128  
129 =item *
130  
131 If using the console, C<{"input_1": [1,3,224,224],
132 "input_2":[1,3,224,224]}>
133  
134 =item *
135  
136 If using the CLI, C<{\"input_1\": [1,3,224,224],
137 \"input_2\":[1,3,224,224]}>
138  
139 =back
140  
141 =back
142  
143 =item *
144  
145 C<MXNET/ONNX/DARKNET>: You must specify the name and shape (NCHW
146 format) of the expected data inputs in order using a dictionary format
147 for your trained model. The dictionary formats required for the console
148 and CLI are different.
149  
150 =over
151  
152 =item *
153  
154 Examples for one input:
155  
156 =over
157  
158 =item *
159  
160 If using the console, C<{"data":[1,3,1024,1024]}>
161  
162 =item *
163  
164 If using the CLI, C<{\"data\":[1,3,1024,1024]}>
165  
166 =back
167  
168 =item *
169  
170 Examples for two inputs:
171  
172 =over
173  
174 =item *
175  
176 If using the console, C<{"var1": [1,1,28,28], "var2":[1,1,28,28]}>
177  
178 =item *
179  
180 If using the CLI, C<{\"var1\": [1,1,28,28], \"var2\":[1,1,28,28]}>
181  
182 =back
183  
184 =back
185  
186 =item *
187  
188 C<PyTorch>: You can either specify the name and shape (NCHW format) of
189 expected data inputs in order using a dictionary format for your
190 trained model or you can specify the shape only using a list format.
191 The dictionary formats required for the console and CLI are different.
192 The list formats for the console and CLI are the same.
193  
194 =over
195  
196 =item *
197  
198 Examples for one input in dictionary format:
199  
200 =over
201  
202 =item *
203  
204 If using the console, C<{"input0":[1,3,224,224]}>
205  
206 =item *
207  
208 If using the CLI, C<{\"input0\":[1,3,224,224]}>
209  
210 =back
211  
212 =item *
213  
214 Example for one input in list format: C<[[1,3,224,224]]>
215  
216 =item *
217  
218 Examples for two inputs in dictionary format:
219  
220 =over
221  
222 =item *
223  
224 If using the console, C<{"input0":[1,3,224,224],
225 "input1":[1,3,224,224]}>
226  
227 =item *
228  
229 If using the CLI, C<{\"input0\":[1,3,224,224],
230 \"input1\":[1,3,224,224]}>
231  
232 =back
233  
234 =item *
235  
236 Example for two inputs in list format: C<[[1,3,224,224],
237 [1,3,224,224]]>
238  
239 =back
240  
241 =item *
242  
243 C<XGBOOST>: input data name and shape are not needed.
244  
245 =back
246  
247 C<DataInputConfig> supports the following parameters for C<CoreML>
248 OutputConfig$TargetDevice (ML Model format):
249  
250 =over
251  
252 =item *
253  
254 C<shape>: Input shape, for example C<{"input_1": {"shape":
255 [1,224,224,3]}}>. In addition to static input shapes, CoreML converter
256 supports Flexible input shapes:
257  
258 =over
259  
260 =item *
261  
262 Range Dimension. You can use the Range Dimension feature if you know
263 the input shape will be within some specific interval in that
264 dimension, for example: C<{"input_1": {"shape": ["1..10", 224, 224,
265 3]}}>
266  
267 =item *
268  
269 Enumerated shapes. Sometimes, the models are trained to work only on a
270 select set of inputs. You can enumerate all supported input shapes, for
271 example: C<{"input_1": {"shape": [[1, 224, 224, 3], [1, 160, 160,
272 3]]}}>
273  
274 =back
275  
276 =item *
277  
278 C<default_shape>: Default input shape. You can set a default shape
279 during conversion for both Range Dimension and Enumerated Shapes. For
280 example C<{"input_1": {"shape": ["1..10", 224, 224, 3],
281 "default_shape": [1, 224, 224, 3]}}>
282  
283 =item *
284  
285 C<type>: Input type. Allowed values: C<Image> and C<Tensor>. By
286 default, the converter generates an ML Model with inputs of type Tensor
287 (MultiArray). User can set input type to be Image. Image input type
288 requires additional input parameters such as C<bias> and C<scale>.
289  
290 =item *
291  
292 C<bias>: If the input type is an Image, you need to provide the bias
293 vector.
294  
295 =item *
296  
297 C<scale>: If the input type is an Image, you need to provide a scale
298 factor.
299  
300 =back
301  
302 CoreML C<ClassifierConfig> parameters can be specified using
303 OutputConfig$CompilerOptions. CoreML converter supports Tensorflow and
304 PyTorch models. CoreML conversion examples:
305  
306 =over
307  
308 =item *
309  
310 Tensor type input:
311  
312 =over
313  
314 =item *
315  
316 C<"DataInputConfig": {"input_1": {"shape": [[1,224,224,3],
317 [1,160,160,3]], "default_shape": [1,224,224,3]}}>
318  
319 =back
320  
321 =item *
322  
323 Tensor type input without input name (PyTorch):
324  
325 =over
326  
327 =item *
328  
329 C<"DataInputConfig": [{"shape": [[1,3,224,224], [1,3,160,160]],
330 "default_shape": [1,3,224,224]}]>
331  
332 =back
333  
334 =item *
335  
336 Image type input:
337  
338 =over
339  
340 =item *
341  
342 C<"DataInputConfig": {"input_1": {"shape": [[1,224,224,3],
343 [1,160,160,3]], "default_shape": [1,224,224,3], "type": "Image",
344 "bias": [-1,-1,-1], "scale": 0.007843137255}}>
345  
346 =item *
347  
348 C<"CompilerOptions": {"class_labels": "imagenet_labels_1000.txt"}>
349  
350 =back
351  
352 =item *
353  
354 Image type input without input name (PyTorch):
355  
356 =over
357  
358 =item *
359  
360 C<"DataInputConfig": [{"shape": [[1,3,224,224], [1,3,160,160]],
361 "default_shape": [1,3,224,224], "type": "Image", "bias": [-1,-1,-1],
362 "scale": 0.007843137255}]>
363  
364 =item *
365  
366 C<"CompilerOptions": {"class_labels": "imagenet_labels_1000.txt"}>
367  
368 =back
369  
370 =back
371  
372 Depending on the model format, C<DataInputConfig> requires the
373 following parameters for C<ml_eia2> OutputConfig:TargetDevice
374 (https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_OutputConfig.html#sagemaker-Type-OutputConfig-TargetDevice).
375  
376 =over
377  
378 =item *
379  
380 For TensorFlow models saved in the SavedModel format, specify the input
381 names from C<signature_def_key> and the input model shapes for
382 C<DataInputConfig>. Specify the C<signature_def_key> in
383 C<OutputConfig:CompilerOptions>
384 (https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_OutputConfig.html#sagemaker-Type-OutputConfig-CompilerOptions)
385 if the model does not use TensorFlow's default signature def key. For
386 example:
387  
388 =over
389  
390 =item *
391  
392 C<"DataInputConfig": {"inputs": [1, 224, 224, 3]}>
393  
394 =item *
395  
396 C<"CompilerOptions": {"signature_def_key": "serving_custom"}>
397  
398 =back
399  
400 =item *
401  
402 For TensorFlow models saved as a frozen graph, specify the input tensor
403 names and shapes in C<DataInputConfig> and the output tensor names for
404 C<output_names> in C<OutputConfig:CompilerOptions>
405 (https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_OutputConfig.html#sagemaker-Type-OutputConfig-CompilerOptions).
406 For example:
407  
408 =over
409  
410 =item *
411  
412 C<"DataInputConfig": {"input_tensor:0": [1, 224, 224, 3]}>
413  
414 =item *
415  
416 C<"CompilerOptions": {"output_names": ["output_tensor:0"]}>
417  
418 =back
419  
420 =back
421  
422  
423  
424 =head2 B<REQUIRED> Framework => Str
425  
426 Identifies the framework in which the model was trained. For example:
427 TENSORFLOW.
428  
429  
430 =head2 FrameworkVersion => Str
431  
432 Specifies the framework version to use.
433  
434 This API field is only supported for PyTorch framework versions C<1.4>,
435 C<1.5>, and C<1.6> for cloud instance target devices: C<ml_c4>,
436 C<ml_c5>, C<ml_m4>, C<ml_m5>, C<ml_p2>, C<ml_p3>, and C<ml_g4dn>.
437  
438  
439 =head2 B<REQUIRED> S3Uri => Str
440  
441 The S3 path where the model artifacts, which result from model
442 training, are stored. This path must point to a single gzip compressed
443 tar archive (.tar.gz suffix).
444  
445  
446  
447 =head1 SEE ALSO
448  
449 This class forms part of L<Paws>, describing an object used in L<Paws::SageMaker>
450  
451 =head1 BUGS and CONTRIBUTIONS
452  
453 The source code is located here: L<https://github.com/pplu/aws-sdk-perl>
454  
455 Please report bugs to: L<https://github.com/pplu/aws-sdk-perl/issues>
456  
457 =cut
458