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
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
|
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<link rel="canonical" href="http://keras.io/utils/">
<link rel="shortcut icon" href="../img/favicon.ico">
<title>Utils - Keras Documentation</title>
<link href='https://fonts.googleapis.com/css?family=Lato:400,700|Source+Sans+Pro:400,700|Inconsolata:400,700' rel='stylesheet' type='text/css'>
<link rel="stylesheet" href="../css/theme.css" type="text/css" />
<link rel="stylesheet" href="../css/theme_extra.css" type="text/css" />
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/styles/github.min.css">
<script>
// Current page data
var mkdocs_page_name = "Utils";
var mkdocs_page_input_path = "utils.md";
var mkdocs_page_url = "/utils/";
</script>
<script src="../js/jquery-2.1.1.min.js" defer></script>
<script src="../js/modernizr-2.8.3.min.js" defer></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/9.12.0/highlight.min.js"></script>
<script>hljs.initHighlightingOnLoad();</script>
<script>
(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
})(window,document,'script','https://www.google-analytics.com/analytics.js','ga');
ga('create', 'UA-61785484-1', 'keras.io');
ga('send', 'pageview');
</script>
</head>
<body class="wy-body-for-nav" role="document">
<div class="wy-grid-for-nav">
<nav data-toggle="wy-nav-shift" class="wy-nav-side stickynav">
<div class="wy-side-scroll">
<a href="">
<div class="keras-logo">
<img src="/img/keras-logo-small.jpg" class="keras-logo-img">
Keras Documentation
</div>
</a>
<div class="wy-side-nav-search">
<div role="search">
<form id ="rtd-search-form" class="wy-form" action="../search.html" method="get">
<input type="text" name="q" placeholder="Search docs" title="Type search term here" />
</form>
</div>
</div>
<div class="wy-menu wy-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
<ul>
<li class="toctree-l1"><a class="reference internal" href="..">Home</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../why-use-keras/">Why use Keras</a>
</li>
</ul>
<p class="caption"><span class="caption-text">Getting started</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../getting-started/sequential-model-guide/">Guide to the Sequential model</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../getting-started/functional-api-guide/">Guide to the Functional API</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../getting-started/faq/">FAQ</a>
</li>
</ul>
<p class="caption"><span class="caption-text">Models</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../models/about-keras-models/">About Keras models</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../models/sequential/">Sequential</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../models/model/">Model (functional API)</a>
</li>
</ul>
<p class="caption"><span class="caption-text">Layers</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../layers/about-keras-layers/">About Keras layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/core/">Core Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/convolutional/">Convolutional Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/pooling/">Pooling Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/local/">Locally-connected Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/recurrent/">Recurrent Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/embeddings/">Embedding Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/merge/">Merge Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/advanced-activations/">Advanced Activations Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/normalization/">Normalization Layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/noise/">Noise layers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/wrappers/">Layer wrappers</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../layers/writing-your-own-keras-layers/">Writing your own Keras layers</a>
</li>
</ul>
<p class="caption"><span class="caption-text">Preprocessing</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../preprocessing/sequence/">Sequence Preprocessing</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../preprocessing/text/">Text Preprocessing</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../preprocessing/image/">Image Preprocessing</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../losses/">Losses</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../metrics/">Metrics</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../optimizers/">Optimizers</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../activations/">Activations</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../callbacks/">Callbacks</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../datasets/">Datasets</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../applications/">Applications</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../backend/">Backend</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../initializers/">Initializers</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../regularizers/">Regularizers</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../constraints/">Constraints</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../visualization/">Visualization</a>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../scikit-learn-api/">Scikit-learn API</a>
</li>
</ul>
<ul class="current">
<li class="toctree-l1 current"><a class="reference internal current" href="./">Utils</a>
<ul class="current">
</ul>
</li>
</ul>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../contributing/">Contributing</a>
</li>
</ul>
<p class="caption"><span class="caption-text">Examples</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../examples/addition_rnn/">Addition RNN</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/antirectifier/">Custom layer - antirectifier</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/babi_rnn/">Baby RNN</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/babi_memnn/">Baby MemNN</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/cifar10_cnn/">CIFAR-10 CNN</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/cifar10_resnet/">CIFAR-10 ResNet</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/conv_filter_visualization/">Convolution filter visualization</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/conv_lstm/">Convolutional LSTM</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/deep_dream/">Deep Dream</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/image_ocr/">Image OCR</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/imdb_bidirectional_lstm/">Bidirectional LSTM</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/imdb_cnn/">1D CNN for text classification</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/imdb_cnn_lstm/">Sentiment classification CNN-LSTM</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/imdb_fasttext/">Fasttext for text classification</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/imdb_lstm/">Sentiment classification LSTM</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/lstm_seq2seq/">Sequence to sequence - training</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/lstm_seq2seq_restore/">Sequence to sequence - prediction</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/lstm_stateful/">Stateful LSTM</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/lstm_text_generation/">LSTM for text generation</a>
</li>
<li class="toctree-l1"><a class="reference internal" href="../examples/mnist_acgan/">Auxiliary Classifier GAN</a>
</li>
</ul>
</div>
</div>
</nav>
<section data-toggle="wy-nav-shift" class="wy-nav-content-wrap">
<nav class="wy-nav-top" role="navigation" aria-label="top navigation">
<i data-toggle="wy-nav-top" class="fa fa-bars"></i>
<a href="..">Keras Documentation</a>
</nav>
<div class="wy-nav-content">
<div class="rst-content">
<div role="navigation" aria-label="breadcrumbs navigation">
<ul class="wy-breadcrumbs">
<li><a href="..">Docs</a> »</li>
<li>Utils</li>
<li class="wy-breadcrumbs-aside">
<a href="https://github.com/keras-team/keras/tree/master/docs"
class="icon icon-github"> Edit on GitHub</a>
</li>
</ul>
<hr/>
</div>
<div role="main">
<div class="section">
<p><span style="float:right;"><a href="https://github.com/keras-team/keras/blob/master/keras/utils/generic_utils.py#L21">[source]</a></span></p>
<h3 id="customobjectscope">CustomObjectScope</h3>
<pre><code class="python">keras.utils.CustomObjectScope()
</code></pre>
<p>Provides a scope that changes to <code>_GLOBAL_CUSTOM_OBJECTS</code> cannot escape.</p>
<p>Code within a <code>with</code> statement will be able to access custom objects
by name. Changes to global custom objects persist
within the enclosing <code>with</code> statement. At end of the <code>with</code> statement,
global custom objects are reverted to state
at beginning of the <code>with</code> statement.</p>
<p><strong>Example</strong></p>
<p>Consider a custom object <code>MyObject</code> (e.g. a class):</p>
<pre><code class="python">with CustomObjectScope({'MyObject':MyObject}):
layer = Dense(..., kernel_regularizer='MyObject')
# save, load, etc. will recognize custom object by name
</code></pre>
<hr />
<p><span style="float:right;"><a href="https://github.com/keras-team/keras/blob/master/keras/utils/io_utils.py#L26">[source]</a></span></p>
<h3 id="hdf5matrix">HDF5Matrix</h3>
<pre><code class="python">keras.utils.HDF5Matrix(datapath, dataset, start=0, end=None, normalizer=None)
</code></pre>
<p>Representation of HDF5 dataset to be used instead of a Numpy array.</p>
<p><strong>Example</strong></p>
<pre><code class="python">x_data = HDF5Matrix('input/file.hdf5', 'data')
model.predict(x_data)
</code></pre>
<p>Providing <code>start</code> and <code>end</code> allows use of a slice of the dataset.</p>
<p>Optionally, a normalizer function (or lambda) can be given. This will
be called on every slice of data retrieved.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>datapath</strong>: string, path to a HDF5 file</li>
<li><strong>dataset</strong>: string, name of the HDF5 dataset in the file specified
in datapath</li>
<li><strong>start</strong>: int, start of desired slice of the specified dataset</li>
<li><strong>end</strong>: int, end of desired slice of the specified dataset</li>
<li><strong>normalizer</strong>: function to be called on data when retrieved</li>
</ul>
<p><strong>Returns</strong></p>
<p>An array-like HDF5 dataset.</p>
<hr />
<p><span style="float:right;"><a href="https://github.com/keras-team/keras/blob/master/keras/utils/data_utils.py#L305">[source]</a></span></p>
<h3 id="sequence">Sequence</h3>
<pre><code class="python">keras.utils.Sequence()
</code></pre>
<p>Base object for fitting to a sequence of data, such as a dataset.</p>
<p>Every <code>Sequence</code> must implement the <code>__getitem__</code> and the <code>__len__</code> methods.
If you want to modify your dataset between epochs you may implement
<code>on_epoch_end</code>. The method <code>__getitem__</code> should return a complete batch.</p>
<p><strong>Notes</strong></p>
<p><code>Sequence</code> are a safer way to do multiprocessing. This structure guarantees
that the network will only train once on each sample per epoch which is not
the case with generators.</p>
<p><strong>Examples</strong></p>
<pre><code class="python">from skimage.io import imread
from skimage.transform import resize
import numpy as np
# Here, `x_set` is list of path to the images
# and `y_set` are the associated classes.
class CIFAR10Sequence(Sequence):
def __init__(self, x_set, y_set, batch_size):
self.x, self.y = x_set, y_set
self.batch_size = batch_size
def __len__(self):
return int(np.ceil(len(self.x) / float(self.batch_size)))
def __getitem__(self, idx):
batch_x = self.x[idx * self.batch_size:(idx + 1) * self.batch_size]
batch_y = self.y[idx * self.batch_size:(idx + 1) * self.batch_size]
return np.array([
resize(imread(file_name), (200, 200))
for file_name in batch_x]), np.array(batch_y)
</code></pre>
<hr />
<h3 id="to_categorical">to_categorical</h3>
<pre><code class="python">keras.utils.to_categorical(y, num_classes=None, dtype='float32')
</code></pre>
<p>Converts a class vector (integers) to binary class matrix.</p>
<p>E.g. for use with categorical_crossentropy.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>y</strong>: class vector to be converted into a matrix
(integers from 0 to num_classes).</li>
<li><strong>num_classes</strong>: total number of classes.</li>
<li><strong>dtype</strong>: The data type expected by the input, as a string
(<code>float32</code>, <code>float64</code>, <code>int32</code>...)</li>
</ul>
<p><strong>Returns</strong></p>
<p>A binary matrix representation of the input. The classes axis
is placed last.</p>
<p><strong>Example</strong></p>
<pre><code class="python"># Consider an array of 5 labels out of a set of 3 classes {0, 1, 2}:
> labels
array([0, 2, 1, 2, 0])
# `to_categorical` converts this into a matrix with as many
# columns as there are classes. The number of rows
# stays the same.
> to_categorical(labels)
array([[ 1., 0., 0.],
[ 0., 0., 1.],
[ 0., 1., 0.],
[ 0., 0., 1.],
[ 1., 0., 0.]], dtype=float32)
</code></pre>
<hr />
<h3 id="normalize">normalize</h3>
<pre><code class="python">keras.utils.normalize(x, axis=-1, order=2)
</code></pre>
<p>Normalizes a Numpy array.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>x</strong>: Numpy array to normalize.</li>
<li><strong>axis</strong>: axis along which to normalize.</li>
<li><strong>order</strong>: Normalization order (e.g. 2 for L2 norm).</li>
</ul>
<p><strong>Returns</strong></p>
<p>A normalized copy of the array.</p>
<hr />
<h3 id="get_file">get_file</h3>
<pre><code class="python">keras.utils.get_file(fname, origin, untar=False, md5_hash=None, file_hash=None, cache_subdir='datasets', hash_algorithm='auto', extract=False, archive_format='auto', cache_dir=None)
</code></pre>
<p>Downloads a file from a URL if it not already in the cache.</p>
<p>By default the file at the url <code>origin</code> is downloaded to the
cache_dir <code>~/.keras</code>, placed in the cache_subdir <code>datasets</code>,
and given the filename <code>fname</code>. The final location of a file
<code>example.txt</code> would therefore be <code>~/.keras/datasets/example.txt</code>.</p>
<p>Files in tar, tar.gz, tar.bz, and zip formats can also be extracted.
Passing a hash will verify the file after download. The command line
programs <code>shasum</code> and <code>sha256sum</code> can compute the hash.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>fname</strong>: Name of the file. If an absolute path <code>/path/to/file.txt</code> is
specified the file will be saved at that location.</li>
<li><strong>origin</strong>: Original URL of the file.</li>
<li><strong>untar</strong>: Deprecated in favor of 'extract'.
boolean, whether the file should be decompressed</li>
<li><strong>md5_hash</strong>: Deprecated in favor of 'file_hash'.
md5 hash of the file for verification</li>
<li><strong>file_hash</strong>: The expected hash string of the file after download.
The sha256 and md5 hash algorithms are both supported.</li>
<li><strong>cache_subdir</strong>: Subdirectory under the Keras cache dir where the file is
saved. If an absolute path <code>/path/to/folder</code> is
specified the file will be saved at that location.</li>
<li><strong>hash_algorithm</strong>: Select the hash algorithm to verify the file.
options are 'md5', 'sha256', and 'auto'.
The default 'auto' detects the hash algorithm in use.</li>
<li><strong>extract</strong>: True tries extracting the file as an Archive, like tar or zip.</li>
<li><strong>archive_format</strong>: Archive format to try for extracting the file.
Options are 'auto', 'tar', 'zip', and None.
'tar' includes tar, tar.gz, and tar.bz files.
The default 'auto' is ['tar', 'zip'].
None or an empty list will return no matches found.</li>
<li><strong>cache_dir</strong>: Location to store cached files, when None it
defaults to the <a href="/faq/#where-is-the-keras-configuration-filed-stored">Keras Directory</a>.</li>
</ul>
<p><strong>Returns</strong></p>
<p>Path to the downloaded file</p>
<hr />
<h3 id="print_summary">print_summary</h3>
<pre><code class="python">keras.utils.print_summary(model, line_length=None, positions=None, print_fn=None)
</code></pre>
<p>Prints a summary of a model.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>model</strong>: Keras model instance.</li>
<li><strong>line_length</strong>: Total length of printed lines
(e.g. set this to adapt the display to different
terminal window sizes).</li>
<li><strong>positions</strong>: Relative or absolute positions of log elements in each line.
If not provided, defaults to <code>[.33, .55, .67, 1.]</code>.</li>
<li><strong>print_fn</strong>: Print function to use.
It will be called on each line of the summary.
You can set it to a custom function
in order to capture the string summary.
It defaults to <code>print</code> (prints to stdout).</li>
</ul>
<hr />
<h3 id="plot_model">plot_model</h3>
<pre><code class="python">keras.utils.plot_model(model, to_file='model.png', show_shapes=False, show_layer_names=True, rankdir='TB', expand_nested=False, dpi=96)
</code></pre>
<p>Converts a Keras model to dot format and save to a file.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>model</strong>: A Keras model instance</li>
<li><strong>to_file</strong>: File name of the plot image.</li>
<li><strong>show_shapes</strong>: whether to display shape information.</li>
<li><strong>show_layer_names</strong>: whether to display layer names.</li>
<li><strong>rankdir</strong>: <code>rankdir</code> argument passed to PyDot,
a string specifying the format of the plot:
'TB' creates a vertical plot;
'LR' creates a horizontal plot.</li>
<li><strong>expand_nested</strong>: whether to expand nested models into clusters.</li>
<li><strong>dpi</strong>: dot DPI.</li>
</ul>
<p><strong>Returns</strong></p>
<p>A Jupyter notebook Image object if Jupyter is installed.
This enables in-line display of the model plots in notebooks.</p>
<hr />
<h3 id="multi_gpu_model">multi_gpu_model</h3>
<pre><code class="python">keras.utils.multi_gpu_model(model, gpus=None, cpu_merge=True, cpu_relocation=False)
</code></pre>
<p>Replicates a model on different GPUs.</p>
<p>Specifically, this function implements single-machine
multi-GPU data parallelism. It works in the following way:</p>
<ul>
<li>Divide the model's input(s) into multiple sub-batches.</li>
<li>Apply a model copy on each sub-batch. Every model copy
is executed on a dedicated GPU.</li>
<li>Concatenate the results (on CPU) into one big batch.</li>
</ul>
<p>E.g. if your <code>batch_size</code> is 64 and you use <code>gpus=2</code>,
then we will divide the input into 2 sub-batches of 32 samples,
process each sub-batch on one GPU, then return the full
batch of 64 processed samples.</p>
<p>This induces quasi-linear speedup on up to 8 GPUs.</p>
<p>This function is only available with the TensorFlow backend
for the time being.</p>
<p><strong>Arguments</strong></p>
<ul>
<li><strong>model</strong>: A Keras model instance. To avoid OOM errors,
this model could have been built on CPU, for instance
(see usage example below).</li>
<li><strong>gpus</strong>: Integer >= 2 or list of integers, number of GPUs or
list of GPU IDs on which to create model replicas.</li>
<li><strong>cpu_merge</strong>: A boolean value to identify whether to force
merging model weights under the scope of the CPU or not.</li>
<li><strong>cpu_relocation</strong>: A boolean value to identify whether to
create the model's weights under the scope of the CPU.
If the model is not defined under any preceding device
scope, you can still rescue it by activating this option.</li>
</ul>
<p><strong>Returns</strong></p>
<p>A Keras <code>Model</code> instance which can be used just like the initial
<code>model</code> argument, but which distributes its workload on multiple GPUs.</p>
<p><strong>Examples</strong></p>
<p>Example 1 - Training models with weights merge on CPU</p>
<pre><code class="python">import tensorflow as tf
from keras.applications import Xception
from keras.utils import multi_gpu_model
import numpy as np
num_samples = 1000
height = 224
width = 224
num_classes = 1000
# Instantiate the base model (or "template" model).
# We recommend doing this with under a CPU device scope,
# so that the model's weights are hosted on CPU memory.
# Otherwise they may end up hosted on a GPU, which would
# complicate weight sharing.
with tf.device('/cpu:0'):
model = Xception(weights=None,
input_shape=(height, width, 3),
classes=num_classes)
# Replicates the model on 8 GPUs.
# This assumes that your machine has 8 available GPUs.
parallel_model = multi_gpu_model(model, gpus=8)
parallel_model.compile(loss='categorical_crossentropy',
optimizer='rmsprop')
# Generate dummy data.
x = np.random.random((num_samples, height, width, 3))
y = np.random.random((num_samples, num_classes))
# This `fit` call will be distributed on 8 GPUs.
# Since the batch size is 256, each GPU will process 32 samples.
parallel_model.fit(x, y, epochs=20, batch_size=256)
# Save model via the template model (which shares the same weights):
model.save('my_model.h5')
</code></pre>
<p>Example 2 - Training models with weights merge on CPU using cpu_relocation</p>
<pre><code class="python">..
# Not needed to change the device scope for model definition:
model = Xception(weights=None, ..)
try:
parallel_model = multi_gpu_model(model, cpu_relocation=True)
print("Training using multiple GPUs..")
except ValueError:
parallel_model = model
print("Training using single GPU or CPU..")
parallel_model.compile(..)
..
</code></pre>
<p>Example 3 - Training models with weights merge on GPU (recommended for NV-link)</p>
<pre><code class="python">..
# Not needed to change the device scope for model definition:
model = Xception(weights=None, ..)
try:
parallel_model = multi_gpu_model(model, cpu_merge=False)
print("Training using multiple GPUs..")
except:
parallel_model = model
print("Training using single GPU or CPU..")
parallel_model.compile(..)
..
</code></pre>
<p><strong>On model saving</strong></p>
<p>To save the multi-gpu model, use <code>.save(fname)</code> or <code>.save_weights(fname)</code>
with the template model (the argument you passed to <code>multi_gpu_model</code>),
rather than the model returned by <code>multi_gpu_model</code>.</p>
</div>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../contributing/" class="btn btn-neutral float-right" title="Contributing">Next <span class="icon icon-circle-arrow-right"></span></a>
<a href="../scikit-learn-api/" class="btn btn-neutral" title="Scikit-learn API"><span class="icon icon-circle-arrow-left"></span> Previous</a>
</div>
<hr/>
<div role="contentinfo">
<!-- Copyright etc -->
</div>
Built with <a href="https://www.mkdocs.org/">MkDocs</a> using a <a href="https://github.com/snide/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
</footer>
</div>
</div>
</section>
</div>
<div class="rst-versions" role="note" aria-label="versions">
<span class="rst-current-version" data-toggle="rst-current-version">
<a href="http://github.com/keras-team/keras/" class="fa fa-github" style="float: left; color: #fcfcfc"> GitHub</a>
<span><a href="../scikit-learn-api/" style="color: #fcfcfc;">« Previous</a></span>
<span style="margin-left: 15px"><a href="../contributing/" style="color: #fcfcfc">Next »</a></span>
</span>
</div>
<script>var base_url = '..';</script>
<script src="../js/theme.js" defer></script>
<script src="../search/main.js" defer></script>
<script type="text/javascript" defer>
window.onload = function () {
SphinxRtdTheme.Navigation.enable(true);
};
</script>
</body>
</html>
|
