-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathjit.html
2034 lines (1721 loc) · 184 KB
/
jit.html
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
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!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 name="robots" content="noindex">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>TorchScript — PyTorch 1.5.0 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/jit.html"/>
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="_static/css/jit.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="_static/katex-math.css" type="text/css" />
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="TorchScript Builtins" href="jit_builtin_functions.html" />
<link rel="prev" title="torch.hub" href="hub.html" />
<script src="_static/js/modernizr.min.js"></script>
<!-- Preload the theme fonts -->
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-book.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-Medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium-italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-SemiBold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<!-- Preload the katex fonts -->
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Caligraphic-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
</head>
<div class="container-fluid header-holder tutorials-header" id="header-holder">
<div class="container">
<div class="header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<div class="ecosystem-dropdown">
<a id="dropdownMenuButton" data-toggle="ecosystem-dropdown">
Ecosystem
</a>
<div class="ecosystem-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/hub"">
<span class=dropdown-title>Models (Beta)</span>
<p>Discover, publish, and reuse pre-trained models</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/ecosystem">
<span class=dropdown-title>Tools & Libraries</span>
<p>Explore the ecosystem of tools and libraries</p>
</a>
</div>
</div>
</li>
<li>
<a href="https://pytorch.org/mobile">Mobile</a>
</li>
<li>
<a href="https://pytorch.org/blog/">Blog</a>
</li>
<li>
<a href="https://pytorch.org/tutorials">Tutorials</a>
</li>
<li class="active">
<a href="https://pytorch.org/docs/stable/index.html">Docs</a>
</li>
<li>
<div class="resources-dropdown">
<a id="resourcesDropdownButton" data-toggle="resources-dropdown">
Resources
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/resources"">
<span class=dropdown-title>Developer Resources</span>
<p>Find resources and get questions answered</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/features">
<span class=dropdown-title>About</span>
<p>Learn about PyTorch’s features and capabilities</p>
</a>
</div>
</div>
</li>
<li>
<a href="https://github.com/pytorch/pytorch">Github</a>
</li>
</ul>
</div>
<a class="main-menu-open-button" href="#" data-behavior="open-mobile-menu"></a>
</div>
</div>
</div>
<body class="pytorch-body">
<div class="table-of-contents-link-wrapper">
<span>Table of Contents</span>
<a href="#" class="toggle-table-of-contents" data-behavior="toggle-table-of-contents"></a>
</div>
<nav data-toggle="wy-nav-shift" class="pytorch-left-menu" id="pytorch-left-menu">
<div class="pytorch-side-scroll">
<div class="pytorch-menu pytorch-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
<div class="pytorch-left-menu-search">
<div class="version">
<a href='http://pytorch.org/docs/versions.html'>1.5.0 ▼</a>
</div>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="search.html" method="get">
<input type="text" name="q" placeholder="Search Docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div>
<p class="caption"><span class="caption-text">Notes</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="notes/amp_examples.html">Automatic Mixed Precision examples</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/autograd.html">Autograd mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/broadcasting.html">Broadcasting semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cpu_threading_torchscript_inference.html">CPU threading and TorchScript inference</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cuda.html">CUDA semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/ddp.html">Distributed Data Parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/extending.html">Extending PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/faq.html">Frequently Asked Questions</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/large_scale_deployments.html">Features for large-scale deployments</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/multiprocessing.html">Multiprocessing best practices</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/randomness.html">Reproducibility</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/serialization.html">Serialization semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/windows.html">Windows FAQ</a></li>
</ul>
<p class="caption"><span class="caption-text">Language Bindings</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="cpp_index.html">C++</a></li>
<li class="toctree-l1"><a class="reference internal" href="packages.html">Javadoc</a></li>
</ul>
<p class="caption"><span class="caption-text">Python API</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="torch.html">torch</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.html">torch.nn</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.functional.html">torch.nn.functional</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensors.html">torch.Tensor</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_attributes.html">Tensor Attributes</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_view.html">Tensor Views</a></li>
<li class="toctree-l1"><a class="reference internal" href="autograd.html">torch.autograd</a></li>
<li class="toctree-l1"><a class="reference internal" href="cuda.html">torch.cuda</a></li>
<li class="toctree-l1"><a class="reference internal" href="amp.html">torch.cuda.amp</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.html">torch.distributed</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributions.html">torch.distributions</a></li>
<li class="toctree-l1"><a class="reference internal" href="hub.html">torch.hub</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">torch.jit</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.init.html">torch.nn.init</a></li>
<li class="toctree-l1"><a class="reference internal" href="onnx.html">torch.onnx</a></li>
<li class="toctree-l1"><a class="reference internal" href="optim.html">torch.optim</a></li>
<li class="toctree-l1"><a class="reference internal" href="quantization.html">Quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="rpc/index.html">Distributed RPC Framework</a></li>
<li class="toctree-l1"><a class="reference internal" href="random.html">torch.random</a></li>
<li class="toctree-l1"><a class="reference internal" href="sparse.html">torch.sparse</a></li>
<li class="toctree-l1"><a class="reference internal" href="storage.html">torch.Storage</a></li>
<li class="toctree-l1"><a class="reference internal" href="bottleneck.html">torch.utils.bottleneck</a></li>
<li class="toctree-l1"><a class="reference internal" href="checkpoint.html">torch.utils.checkpoint</a></li>
<li class="toctree-l1"><a class="reference internal" href="cpp_extension.html">torch.utils.cpp_extension</a></li>
<li class="toctree-l1"><a class="reference internal" href="data.html">torch.utils.data</a></li>
<li class="toctree-l1"><a class="reference internal" href="dlpack.html">torch.utils.dlpack</a></li>
<li class="toctree-l1"><a class="reference internal" href="model_zoo.html">torch.utils.model_zoo</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensorboard.html">torch.utils.tensorboard</a></li>
<li class="toctree-l1"><a class="reference internal" href="type_info.html">Type Info</a></li>
<li class="toctree-l1"><a class="reference internal" href="named_tensor.html">Named Tensors</a></li>
<li class="toctree-l1"><a class="reference internal" href="name_inference.html">Named Tensors operator coverage</a></li>
<li class="toctree-l1"><a class="reference internal" href="__config__.html">torch.__config__</a></li>
</ul>
<p class="caption"><span class="caption-text">Libraries</span></p>
<ul>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/audio">torchaudio</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/text">torchtext</a></li>
<li class="toctree-l1"><a class="reference internal" href="torchvision/index.html">torchvision</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/elastic/">TorchElastic</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/serve">TorchServe</a></li>
<li class="toctree-l1"><a class="reference external" href="http://pytorch.org/xla/">PyTorch on XLA Devices</a></li>
</ul>
<p class="caption"><span class="caption-text">Community</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="community/contribution_guide.html">PyTorch Contribution Guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/governance.html">PyTorch Governance</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/persons_of_interest.html">PyTorch Governance | Persons of Interest</a></li>
</ul>
</div>
</div>
</nav>
<div class="pytorch-container">
<div class="pytorch-page-level-bar" id="pytorch-page-level-bar">
<div class="pytorch-breadcrumbs-wrapper">
<div role="navigation" aria-label="breadcrumbs navigation">
<ul class="pytorch-breadcrumbs">
<li>
<a href="index.html">
Docs
</a> >
</li>
<li>TorchScript</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/jit.rst.txt" rel="nofollow"><img src="_static/images/view-page-source-icon.svg"></a>
</li>
</ul>
</div>
</div>
<div class="pytorch-shortcuts-wrapper" id="pytorch-shortcuts-wrapper">
Shortcuts
</div>
</div>
<section data-toggle="wy-nav-shift" id="pytorch-content-wrap" class="pytorch-content-wrap">
<div class="pytorch-content-left">
<div class="rst-content">
<div role="main" class="main-content" itemscope="itemscope" itemtype="http://schema.org/Article">
<article itemprop="articleBody" id="pytorch-article" class="pytorch-article">
<div class="section" id="torchscript">
<h1>TorchScript<a class="headerlink" href="#torchscript" title="Permalink to this headline">¶</a></h1>
<div class="toctree-wrapper compound">
</div>
<div class="toctree-wrapper compound">
</div>
<div class="contents local topic" id="contents">
<ul class="simple">
<li><p><a class="reference internal" href="#creating-torchscript-code" id="id4">Creating TorchScript Code</a></p></li>
<li><p><a class="reference internal" href="#mixing-tracing-and-scripting" id="id5">Mixing Tracing and Scripting</a></p></li>
<li><p><a class="reference internal" href="#torchscript-language" id="id6">TorchScript Language</a></p></li>
<li><p><a class="reference internal" href="#built-in-functions-and-modules" id="id7">Built-in Functions and Modules</a></p>
<ul>
<li><p><a class="reference internal" href="#pytorch-functions-and-modules" id="id8">PyTorch Functions and Modules</a></p></li>
<li><p><a class="reference internal" href="#python-functions-and-modules" id="id9">Python Functions and Modules</a></p></li>
<li><p><a class="reference internal" href="#python-language-reference-comparison" id="id10">Python Language Reference Comparison</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#debugging" id="id11">Debugging</a></p>
<ul>
<li><p><a class="reference internal" href="#disable-jit-for-debugging" id="id12">Disable JIT for Debugging</a></p></li>
<li><p><a class="reference internal" href="#inspecting-code" id="id13">Inspecting Code</a></p></li>
<li><p><a class="reference internal" href="#interpreting-graphs" id="id14">Interpreting Graphs</a></p></li>
<li><p><a class="reference internal" href="#tracer" id="id15">Tracer</a></p></li>
<li><p><a class="reference internal" href="#id1" id="id16">Built-in Functions and Modules</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#frequently-asked-questions" id="id17">Frequently Asked Questions</a></p></li>
<li><p><a class="reference internal" href="#appendix" id="id18">Appendix</a></p>
<ul>
<li><p><a class="reference internal" href="#migrating-to-pytorch-1-2-recursive-scripting-api" id="id19">Migrating to PyTorch 1.2 Recursive Scripting API</a></p></li>
</ul>
</li>
</ul>
</div>
<span class="target" id="module-torch.jit"></span><p>TorchScript is a way to create serializable and optimizable models from PyTorch code.
Any TorchScript program can be saved from a Python
process and loaded in a process where there is no Python dependency.</p>
<p>We provide tools to incrementally transition a model from a pure Python program
to a TorchScript program that can be run independently from Python, such as in a standalone C++ program.
This makes it possible to train models in PyTorch using familiar tools in Python and then export
the model via TorchScript to a production environment where Python programs may be disadvantageous
for performance and multi-threading reasons.</p>
<p>For a gentle introduction to TorchScript, see the <a class="reference external" href="https://pytorch.org/tutorials/beginner/Intro_to_TorchScript_tutorial.html">Introduction to TorchScript</a> tutorial.</p>
<p>For an end-to-end example of converting a PyTorch model to TorchScript and running it in C++, see the
<a class="reference external" href="https://pytorch.org/tutorials/advanced/cpp_export.html">Loading a PyTorch Model in C++</a> tutorial.</p>
<div class="section" id="creating-torchscript-code">
<h2><a class="toc-backref" href="#id4">Creating TorchScript Code</a><a class="headerlink" href="#creating-torchscript-code" title="Permalink to this headline">¶</a></h2>
<dl class="function">
<dt id="torch.jit.script">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">script</code><span class="sig-paren">(</span><em class="sig-param">obj</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/jit.html#script"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.script" title="Permalink to this definition">¶</a></dt>
<dd><p>Scripting a function or <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> will inspect the source code, compile
it as TorchScript code using the TorchScript compiler, and return a <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> or
<a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a>. TorchScript itself is a subset of the Python language, so not all
features in Python work, but we provide enough functionality to compute on
tensors and do control-dependent operations. For a complete guide, see the
<a class="reference internal" href="jit_language_reference.html#language-reference"><span class="std std-ref">TorchScript Language Reference</span></a>.</p>
<p><code class="docutils literal notranslate"><span class="pre">torch.jit.script</span></code> can be used as a function for modules and functions, and as a decorator
<code class="docutils literal notranslate"><span class="pre">@torch.jit.script</span></code> for <a class="reference internal" href="jit_language_reference.html#id2"><span class="std std-ref">TorchScript Classes</span></a> and functions.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>obj</strong> (callable, class, or <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>) – The <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>, function, or class type to
compile.</p>
</dd>
<dt class="field-even">Returns</dt>
<dd class="field-even"><p>If <code class="docutils literal notranslate"><span class="pre">obj</span></code> is <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>, <code class="docutils literal notranslate"><span class="pre">script</span></code> returns
a <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> object. The returned <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> will
have the same set of sub-modules and parameters as the
original <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>. If <code class="docutils literal notranslate"><span class="pre">obj</span></code> is a standalone function,
a <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a> will be returned.</p>
</dd>
</dl>
<dl>
<dt><strong>Scripting a function</strong></dt><dd><p>The <code class="docutils literal notranslate"><span class="pre">@torch.jit.script</span></code> decorator will construct a <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a>
by compiling the body of the function.</p>
<p>Example (scripting a function):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span>
<span class="k">def</span> <span class="nf">foo</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">):</span>
<span class="k">if</span> <span class="n">x</span><span class="o">.</span><span class="n">max</span><span class="p">()</span> <span class="o">></span> <span class="n">y</span><span class="o">.</span><span class="n">max</span><span class="p">():</span>
<span class="n">r</span> <span class="o">=</span> <span class="n">x</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">r</span> <span class="o">=</span> <span class="n">y</span>
<span class="k">return</span> <span class="n">r</span>
<span class="nb">print</span><span class="p">(</span><span class="nb">type</span><span class="p">(</span><span class="n">foo</span><span class="p">))</span> <span class="c1"># torch.jit.ScriptFuncion</span>
<span class="c1"># See the compiled graph as Python code</span>
<span class="nb">print</span><span class="p">(</span><span class="n">foo</span><span class="o">.</span><span class="n">code</span><span class="p">)</span>
<span class="c1"># Call the function using the TorchScript interpreter</span>
<span class="n">foo</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">ones</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span>
</pre></div>
</div>
</dd>
<dt><strong>Scripting an nn.Module</strong></dt><dd><p>Scripting an <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> by default will compile the <code class="docutils literal notranslate"><span class="pre">forward</span></code> method and recursively
compile any methods, submodules, and functions called by <code class="docutils literal notranslate"><span class="pre">forward</span></code>. If a <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> only uses
features supported in TorchScript, no changes to the original module code should be necessary. <code class="docutils literal notranslate"><span class="pre">script</span></code>
will construct <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> that has copies of the attributes, parameters, and methods of
the original module.</p>
<p>Example (scripting a simple module with a Parameter):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">N</span><span class="p">,</span> <span class="n">M</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">MyModule</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="c1"># This parameter will be copied to the new ScriptModule</span>
<span class="bp">self</span><span class="o">.</span><span class="n">weight</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Parameter</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="n">N</span><span class="p">,</span> <span class="n">M</span><span class="p">))</span>
<span class="c1"># When this submodule is used, it will be compiled</span>
<span class="bp">self</span><span class="o">.</span><span class="n">linear</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">N</span><span class="p">,</span> <span class="n">M</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
<span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">mv</span><span class="p">(</span><span class="nb">input</span><span class="p">)</span>
<span class="c1"># This calls the `forward` method of the `nn.Linear` module, which will</span>
<span class="c1"># cause the `self.linear` submodule to be compiled to a `ScriptModule` here</span>
<span class="n">output</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">linear</span><span class="p">(</span><span class="n">output</span><span class="p">)</span>
<span class="k">return</span> <span class="n">output</span>
<span class="n">scripted_module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">))</span>
</pre></div>
</div>
<p>Example (scripting a module with traced submodules):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="kn">import</span> <span class="nn">torch.nn.functional</span> <span class="k">as</span> <span class="nn">F</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">MyModule</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="c1"># torch.jit.trace produces a ScriptModule's conv1 and conv2</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">20</span><span class="p">,</span> <span class="mi">5</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">16</span><span class="p">))</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">20</span><span class="p">,</span> <span class="mi">20</span><span class="p">,</span> <span class="mi">5</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">20</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">16</span><span class="p">))</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
<span class="nb">input</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="nb">input</span><span class="p">))</span>
<span class="nb">input</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="nb">input</span><span class="p">))</span>
<span class="k">return</span> <span class="nb">input</span>
<span class="n">scripted_module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">())</span>
</pre></div>
</div>
<p>To compile a method other than <code class="docutils literal notranslate"><span class="pre">forward</span></code> (and recursively compile anything it calls), add
the <a class="reference internal" href="#torch.jit.export" title="torch.jit.export"><code class="xref py py-func docutils literal notranslate"><span class="pre">@torch.jit.export</span></code></a> decorator to the method. To opt out of compilation
use <a class="reference internal" href="#torch.jit.ignore" title="torch.jit.ignore"><code class="xref py py-func docutils literal notranslate"><span class="pre">@torch.jit.ignore</span></code></a> or <a class="reference internal" href="#torch.jit.unused" title="torch.jit.unused"><code class="xref py py-func docutils literal notranslate"><span class="pre">@torch.jit.unused</span></code></a>.</p>
<p>Example (an exported and ignored method in a module):</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">MyModule</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">export</span>
<span class="k">def</span> <span class="nf">some_entry_point</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
<span class="k">return</span> <span class="nb">input</span> <span class="o">+</span> <span class="mi">10</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">ignore</span>
<span class="k">def</span> <span class="nf">python_only_fn</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
<span class="c1"># This function won't be compiled, so any</span>
<span class="c1"># Python APIs can be used</span>
<span class="kn">import</span> <span class="nn">pdb</span>
<span class="n">pdb</span><span class="o">.</span><span class="n">set_trace</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="nb">input</span><span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">python_only_fn</span><span class="p">(</span><span class="nb">input</span><span class="p">)</span>
<span class="k">return</span> <span class="nb">input</span> <span class="o">*</span> <span class="mi">99</span>
<span class="n">scripted_module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">())</span>
<span class="nb">print</span><span class="p">(</span><span class="n">scripted_module</span><span class="o">.</span><span class="n">some_entry_point</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">)))</span>
<span class="nb">print</span><span class="p">(</span><span class="n">scripted_module</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">)))</span>
</pre></div>
</div>
</dd>
</dl>
</dd></dl>
<dl class="function">
<dt id="torch.jit.trace">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">trace</code><span class="sig-paren">(</span><em class="sig-param">func</em>, <em class="sig-param">example_inputs</em>, <em class="sig-param">optimize=None</em>, <em class="sig-param">check_trace=True</em>, <em class="sig-param">check_inputs=None</em>, <em class="sig-param">check_tolerance=1e-5</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/jit.html#trace"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.trace" title="Permalink to this definition">¶</a></dt>
<dd><p>Trace a function and return an executable or <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a>
that will be optimized using just-in-time compilation. Tracing is ideal for
code that operates only on <code class="docutils literal notranslate"><span class="pre">Tensor</span></code>s and lists, dictionaries, and tuples of <code class="docutils literal notranslate"><span class="pre">Tensor</span></code>s.</p>
<p>Using <code class="docutils literal notranslate"><span class="pre">torch.jit.trace</span></code> and <a class="reference internal" href="#torch.jit.trace_module" title="torch.jit.trace_module"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.trace_module</span></code></a>, you can turn an existing module or Python
function into a TorchScript <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a> or <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a>. You must provide example inputs,
and we run the function, recording the operations performed on all the tensors.</p>
<ul class="simple">
<li><p>The resulting recording of a standalone function produces <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a>.</p></li>
<li><p>The resulting recording of <code class="docutils literal notranslate"><span class="pre">forward</span></code> function of <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> or <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> produces <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a>.</p></li>
</ul>
<p>This module also contains any parameters that the original
module had as well.</p>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p>Tracing only correctly records functions and modules which are not data
dependent (e.g., do not have conditionals on data in tensors) and do not have
any untracked external dependencies (e.g., perform input/output or
access global variables). Tracing only records operations done when the given
function is run on the given
tensors. Therefore, the returned <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> will always run the same traced
graph on any input. This has some important implications when your module is
expected to run different sets of operations, depending on the input and/or the
module state. For example,</p>
<ul class="simple">
<li><p>Tracing will not record any control-flow like if-statements or loops.
When this control-flow is constant across your module, this is fine and it often
inlines the control-flow decisions. But sometimes the control-flow is actually part
of the model itself. For instance, a recurrent network is a loop over
the (possibly dynamic) length of an input sequence.</p></li>
<li><p>In the returned <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a>, operations that have different
behaviors in <code class="docutils literal notranslate"><span class="pre">training</span></code> and <code class="docutils literal notranslate"><span class="pre">eval</span></code> modes will always behave as if it
is in the mode it was in during tracing, no matter which mode the
<a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> is in.</p></li>
</ul>
<p>In cases like these, tracing would not be appropriate and <a class="reference internal" href="#torch.jit.script" title="torch.jit.script"><code class="xref py py-func docutils literal notranslate"><span class="pre">scripting</span></code></a> is a better
choice. If you trace such models, you may silently get
incorrect results on subsequent invocations of the model. The tracer
will try to emit warnings when doing something that may cause an
incorrect trace to be produced.</p>
</div>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>func</strong> (<em>callable</em><em> or </em><a class="reference internal" href="nn.html#torch.nn.Module" title="torch.nn.Module"><em>torch.nn.Module</em></a>) – A Python function or <code class="docutils literal notranslate"><span class="pre">torch.nn.Module</span></code>
that will be run with <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code>.
arguments and returns to <code class="docutils literal notranslate"><span class="pre">func</span></code> must be tensors
or (possibly nested) tuples that
contain tensors. When a module is passed to
<a class="reference internal" href="#torch.jit.trace" title="torch.jit.trace"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.trace</span></code></a>, only the
<code class="docutils literal notranslate"><span class="pre">forward</span></code> method is run and traced
(see <a class="reference internal" href="#torch.jit.trace_module" title="torch.jit.trace_module"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.trace</span></code></a> for details).</p></li>
<li><p><strong>example_inputs</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#tuple" title="(in Python v3.8)"><em>tuple</em></a>) – A tuple of example inputs that will be passed to the function
while tracing. The resulting trace can be run with
inputs of different types and shapes assuming the traced operations
support those types and shapes. <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code> may also be a single
Tensor in which case it is automatically wrapped in a tuple.</p></li>
</ul>
</dd>
<dt class="field-even">Keyword Arguments</dt>
<dd class="field-even"><ul class="simple">
<li><p><strong>check_trace</strong> (<a class="reference internal" href="storage.html#torch.FloatStorage.bool" title="torch.FloatStorage.bool"><em>bool</em></a><em>, </em><em>optional</em>) – Check if the same inputs run through
traced code produce the same outputs. Default: <code class="docutils literal notranslate"><span class="pre">True</span></code>. You might want
to disable this if, for example, your network contains non-
deterministic ops or if you are sure that the network is correct despite
a checker failure.</p></li>
<li><p><strong>check_inputs</strong> (<em>list of tuples</em><em>, </em><em>optional</em>) – A list of tuples of input arguments that should be used
to check the trace against what is expected. Each tuple
is equivalent to a set of input arguments that would
be specified in <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code>. For best results, pass in a
set of checking inputs representative of the space of
shapes and types of inputs you expect the network to see.
If not specified, the original <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code> are used for checking</p></li>
<li><p><strong>check_tolerance</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#float" title="(in Python v3.8)"><em>float</em></a><em>, </em><em>optional</em>) – Floating-point comparison tolerance to use in the checker procedure.
This can be used to relax the checker strictness in the event that
results diverge numerically for a known reason, such as operator fusion.</p></li>
</ul>
</dd>
<dt class="field-odd">Returns</dt>
<dd class="field-odd"><p>If <code class="docutils literal notranslate"><span class="pre">callable</span></code> is <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> or <code class="docutils literal notranslate"><span class="pre">forward</span></code> of <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>, <code class="docutils literal notranslate"><span class="pre">trace</span></code> returns
a <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> object with a single <code class="docutils literal notranslate"><span class="pre">forward</span></code> method containing the traced code.
The returned <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> will have the same set of sub-modules and parameters as the
original <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code>.
If <code class="docutils literal notranslate"><span class="pre">callable</span></code> is a standalone function, <code class="docutils literal notranslate"><span class="pre">trace</span></code> returns <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a></p>
</dd>
</dl>
<p>Example (tracing a function):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="k">def</span> <span class="nf">foo</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">):</span>
<span class="k">return</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">x</span> <span class="o">+</span> <span class="n">y</span>
<span class="c1"># Run `foo` with the provided inputs and record the tensor operations</span>
<span class="n">traced_foo</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">foo</span><span class="p">,</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">3</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">3</span><span class="p">)))</span>
<span class="c1"># `traced_foo` can now be run with the TorchScript interpreter or saved</span>
<span class="c1"># and loaded in a Python-free environment</span>
</pre></div>
</div>
<p>Example (tracing an existing module):</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">Net</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">Net</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="n">n</span> <span class="o">=</span> <span class="n">Net</span><span class="p">()</span>
<span class="n">example_weight</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="n">example_forward_input</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="c1"># Trace a specific method and construct `ScriptModule` with</span>
<span class="c1"># a single `forward` method</span>
<span class="n">module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">n</span><span class="o">.</span><span class="n">forward</span><span class="p">,</span> <span class="n">example_forward_input</span><span class="p">)</span>
<span class="c1"># Trace a module (implicitly traces `forward`) and construct a</span>
<span class="c1"># `ScriptModule` with a single `forward` method</span>
<span class="n">module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">n</span><span class="p">,</span> <span class="n">example_forward_input</span><span class="p">)</span>
</pre></div>
</div>
</dd></dl>
<dl class="function">
<dt id="torch.jit.trace_module">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">trace_module</code><span class="sig-paren">(</span><em class="sig-param">mod</em>, <em class="sig-param">inputs</em>, <em class="sig-param">optimize=None</em>, <em class="sig-param">check_trace=True</em>, <em class="sig-param">check_inputs=None</em>, <em class="sig-param">check_tolerance=1e-5</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/jit.html#trace_module"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.trace_module" title="Permalink to this definition">¶</a></dt>
<dd><p>Trace a module and return an executable <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> that will be optimized
using just-in-time compilation. When a module is passed to <a class="reference internal" href="#torch.jit.trace" title="torch.jit.trace"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.trace</span></code></a>, only
the <code class="docutils literal notranslate"><span class="pre">forward</span></code> method is run and traced. With <code class="docutils literal notranslate"><span class="pre">trace_module</span></code>, you can specify a dictionary of
method names to example inputs to trace (see the <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code>) argument below.</p>
<p>See <a class="reference internal" href="#torch.jit.trace" title="torch.jit.trace"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.trace</span></code></a> for more information on tracing.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>mod</strong> (<a class="reference internal" href="nn.html#torch.nn.Module" title="torch.nn.Module"><em>torch.nn.Module</em></a>) – A <code class="docutils literal notranslate"><span class="pre">torch.nn.Module</span></code> containing methods whose names are
specified in <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code>. The given methods will be compiled
as a part of a single <cite>ScriptModule</cite>.</p></li>
<li><p><strong>example_inputs</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#dict" title="(in Python v3.8)"><em>dict</em></a>) – A dict containing sample inputs indexed by method names in <code class="docutils literal notranslate"><span class="pre">mod</span></code>.
The inputs will be passed to methods whose names correspond to inputs’
keys while tracing.
<code class="docutils literal notranslate"><span class="pre">{</span> <span class="pre">'forward'</span> <span class="pre">:</span> <span class="pre">example_forward_input,</span> <span class="pre">'method2':</span> <span class="pre">example_method2_input}</span></code></p></li>
</ul>
</dd>
<dt class="field-even">Keyword Arguments</dt>
<dd class="field-even"><ul class="simple">
<li><p><strong>check_trace</strong> (<code class="docutils literal notranslate"><span class="pre">bool</span></code>, optional) – Check if the same inputs run through
traced code produce the same outputs. Default: <code class="docutils literal notranslate"><span class="pre">True</span></code>. You might want
to disable this if, for example, your network contains non-
deterministic ops or if you are sure that the network is correct despite
a checker failure.</p></li>
<li><p><strong>check_inputs</strong> (<em>list of dicts</em><em>, </em><em>optional</em>) – A list of dicts of input arguments that should be used
to check the trace against what is expected. Each tuple
is equivalent to a set of input arguments that would
be specified in <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code>. For best results, pass in a
set of checking inputs representative of the space of
shapes and types of inputs you expect the network to see.
If not specified, the original <code class="docutils literal notranslate"><span class="pre">example_inputs</span></code> are used for checking</p></li>
<li><p><strong>check_tolerance</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#float" title="(in Python v3.8)"><em>float</em></a><em>, </em><em>optional</em>) – Floating-point comparison tolerance to use in the checker procedure.
This can be used to relax the checker strictness in the event that
results diverge numerically for a known reason, such as operator fusion.</p></li>
</ul>
</dd>
<dt class="field-odd">Returns</dt>
<dd class="field-odd"><p>A <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> object with a single <code class="docutils literal notranslate"><span class="pre">forward</span></code> method containing the traced code.
When <code class="docutils literal notranslate"><span class="pre">func</span></code> is a <code class="docutils literal notranslate"><span class="pre">torch.nn.Module</span></code>, the returned <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> will have the same set of
sub-modules and parameters as <code class="docutils literal notranslate"><span class="pre">func</span></code>.</p>
</dd>
</dl>
<p>Example (tracing a module with multiple methods):</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">Net</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">Net</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">weighted_kernel_sum</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">weight</span><span class="p">):</span>
<span class="k">return</span> <span class="n">weight</span> <span class="o">*</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="o">.</span><span class="n">weight</span>
<span class="n">n</span> <span class="o">=</span> <span class="n">Net</span><span class="p">()</span>
<span class="n">example_weight</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="n">example_forward_input</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="c1"># Trace a specific method and construct `ScriptModule` with</span>
<span class="c1"># a single `forward` method</span>
<span class="n">module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">n</span><span class="o">.</span><span class="n">forward</span><span class="p">,</span> <span class="n">example_forward_input</span><span class="p">)</span>
<span class="c1"># Trace a module (implicitly traces `forward`) and construct a</span>
<span class="c1"># `ScriptModule` with a single `forward` method</span>
<span class="n">module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">n</span><span class="p">,</span> <span class="n">example_forward_input</span><span class="p">)</span>
<span class="c1"># Trace specific methods on a module (specified in `inputs`), constructs</span>
<span class="c1"># a `ScriptModule` with `forward` and `weighted_kernel_sum` methods</span>
<span class="n">inputs</span> <span class="o">=</span> <span class="p">{</span><span class="s1">'forward'</span> <span class="p">:</span> <span class="n">example_forward_input</span><span class="p">,</span> <span class="s1">'weighted_kernel_sum'</span> <span class="p">:</span> <span class="n">example_weight</span><span class="p">}</span>
<span class="n">module</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace_module</span><span class="p">(</span><span class="n">n</span><span class="p">,</span> <span class="n">inputs</span><span class="p">)</span>
</pre></div>
</div>
</dd></dl>
<dl class="class">
<dt id="torch.jit.ScriptModule">
<em class="property">class </em><code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">ScriptModule</code><a class="reference internal" href="_modules/torch/jit.html#ScriptModule"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.ScriptModule" title="Permalink to this definition">¶</a></dt>
<dd><p><code class="docutils literal notranslate"><span class="pre">ScriptModule``s</span> <span class="pre">wrap</span> <span class="pre">a</span> <span class="pre">C++</span> <span class="pre">``torch::jit::Module</span></code>. <code class="docutils literal notranslate"><span class="pre">ScriptModule``s</span>
<span class="pre">contain</span> <span class="pre">methods,</span> <span class="pre">attributes,</span> <span class="pre">parameters,</span> <span class="pre">and</span>
<span class="pre">constants.</span> <span class="pre">These</span> <span class="pre">can</span> <span class="pre">be</span> <span class="pre">accessed</span> <span class="pre">the</span> <span class="pre">same</span> <span class="pre">as</span> <span class="pre">on</span> <span class="pre">a</span> <span class="pre">normal</span> <span class="pre">``nn.Module</span></code>.</p>
<dl class="method">
<dt id="torch.jit.ScriptModule.code">
<em class="property">property </em><code class="sig-name descname">code</code><a class="headerlink" href="#torch.jit.ScriptModule.code" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns a pretty-printed representation (as valid Python syntax) of
the internal graph for the <code class="docutils literal notranslate"><span class="pre">forward</span></code> method. See <a class="reference internal" href="#inspecting-code">Inspecting Code</a>
for details.</p>
</dd></dl>
<dl class="method">
<dt id="torch.jit.ScriptModule.graph">
<em class="property">property </em><code class="sig-name descname">graph</code><a class="headerlink" href="#torch.jit.ScriptModule.graph" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns a string representation of the internal graph for the
<code class="docutils literal notranslate"><span class="pre">forward</span></code> method. See <a class="reference internal" href="#interpreting-graphs">Interpreting Graphs</a> for details.</p>
</dd></dl>
<dl class="method">
<dt id="torch.jit.ScriptModule.inlined_graph">
<em class="property">property </em><code class="sig-name descname">inlined_graph</code><a class="headerlink" href="#torch.jit.ScriptModule.inlined_graph" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns a string representation of the internal graph for the
<code class="docutils literal notranslate"><span class="pre">forward</span></code> method. This graph will be preprocessed to inline all function and method calls.
See <a class="reference internal" href="#interpreting-graphs">Interpreting Graphs</a> for details.</p>
</dd></dl>
<dl class="method">
<dt id="torch.jit.ScriptModule.save">
<code class="sig-name descname">save</code><span class="sig-paren">(</span><em class="sig-param">f</em>, <em class="sig-param">_extra_files=ExtraFilesMap{}</em><span class="sig-paren">)</span><a class="headerlink" href="#torch.jit.ScriptModule.save" title="Permalink to this definition">¶</a></dt>
<dd><p>See <a class="reference internal" href="#torch.jit.save" title="torch.jit.save"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.save</span></code></a> for details.</p>
</dd></dl>
</dd></dl>
<dl class="class">
<dt id="torch.jit.ScriptFunction">
<em class="property">class </em><code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">ScriptFunction</code><a class="headerlink" href="#torch.jit.ScriptFunction" title="Permalink to this definition">¶</a></dt>
<dd><p>Functionally equivalent to a <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a>, but represents a single
function and does not have any attributes or Parameters.</p>
</dd></dl>
<dl class="function">
<dt id="torch.jit.save">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">save</code><span class="sig-paren">(</span><em class="sig-param">m</em>, <em class="sig-param">f</em>, <em class="sig-param">_extra_files=ExtraFilesMap{}</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/jit.html#save"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.save" title="Permalink to this definition">¶</a></dt>
<dd><p>Save an offline version of this module for use in a separate process. The saved
module serializes all of the methods, submodules, parameters, and attributes of this
module. It can be loaded into the C++ API using <code class="docutils literal notranslate"><span class="pre">torch::jit::load(filename)</span></code> or into the Python
API with <a class="reference internal" href="#torch.jit.load" title="torch.jit.load"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.load</span></code></a>.</p>
<p>To be able to save a module, it must not make any calls to native Python functions.
This means that all submodules must be subclasses of <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> as well.</p>
<div class="admonition danger">
<p class="admonition-title">Danger</p>
<p>All modules, no matter their device, are always loaded onto the CPU during loading.
This is different from <a class="reference internal" href="torch.html#torch.load" title="torch.load"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.load()</span></code></a>’s semantics and may change in the future.</p>
</div>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>m</strong> – A <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> to save.</p></li>
<li><p><strong>f</strong> – A file-like object (has to implement write and flush) or a string
containing a file name.</p></li>
<li><p><strong>_extra_files</strong> – Map from filename to contents which will be stored as part of ‘f’.</p></li>
</ul>
</dd>
</dl>
<div class="admonition warning">
<p class="admonition-title">Warning</p>
<p>If you are using Python 2, <code class="docutils literal notranslate"><span class="pre">torch.jit.save</span></code> does NOT support <code class="xref any docutils literal notranslate"><span class="pre">StringIO.StringIO</span></code>
as a valid file-like object. This is because the write method should return
the number of bytes written; <code class="docutils literal notranslate"><span class="pre">StringIO.write()</span></code> does not do this.</p>
<p>Please use something like <code class="docutils literal notranslate"><span class="pre">io.BytesIO</span></code> instead.</p>
</div>
<p>Example:</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">io</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="k">return</span> <span class="n">x</span> <span class="o">+</span> <span class="mi">10</span>
<span class="n">m</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">())</span>
<span class="c1"># Save to file</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="s1">'scriptmodule.pt'</span><span class="p">)</span>
<span class="c1"># This line is equivalent to the previous</span>
<span class="n">m</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s2">"scriptmodule.pt"</span><span class="p">)</span>
<span class="c1"># Save to io.BytesIO buffer</span>
<span class="n">buffer</span> <span class="o">=</span> <span class="n">io</span><span class="o">.</span><span class="n">BytesIO</span><span class="p">()</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">buffer</span><span class="p">)</span>
<span class="c1"># Save with extra files</span>
<span class="n">extra_files</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">_C</span><span class="o">.</span><span class="n">ExtraFilesMap</span><span class="p">()</span>
<span class="n">extra_files</span><span class="p">[</span><span class="s1">'foo.txt'</span><span class="p">]</span> <span class="o">=</span> <span class="s1">'bar'</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="s1">'scriptmodule.pt'</span><span class="p">,</span> <span class="n">_extra_files</span><span class="o">=</span><span class="n">extra_files</span><span class="p">)</span>
</pre></div>
</div>
</dd></dl>
<dl class="function">
<dt id="torch.jit.load">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">load</code><span class="sig-paren">(</span><em class="sig-param">f</em>, <em class="sig-param">map_location=None</em>, <em class="sig-param">_extra_files=ExtraFilesMap{}</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/jit.html#load"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.load" title="Permalink to this definition">¶</a></dt>
<dd><p>Load a <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> or <a class="reference internal" href="#torch.jit.ScriptFunction" title="torch.jit.ScriptFunction"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptFunction</span></code></a> previously
saved with <a class="reference internal" href="#torch.jit.save" title="torch.jit.save"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.jit.save</span></code></a></p>
<p>All previously saved modules, no matter their device, are first loaded onto CPU,
and then are moved to the devices they were saved from. If this fails (e.g. because
the run time system doesn’t have certain devices), an exception is raised.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>f</strong> – a file-like object (has to implement read, readline, tell, and seek),
or a string containing a file name</p></li>
<li><p><strong>map_location</strong> (<em>string</em><em> or </em><a class="reference internal" href="tensor_attributes.html#torch.torch.device" title="torch.torch.device"><em>torch.device</em></a>) – A simplified version of <code class="docutils literal notranslate"><span class="pre">map_location</span></code> in
<code class="docutils literal notranslate"><span class="pre">torch.save</span></code> used to dynamically remap storages to an alternative set of devices.</p></li>
<li><p><strong>_extra_files</strong> (<em>dictionary of filename to content</em>) – The extra
filenames given in the map would be loaded and their content
would be stored in the provided map.</p></li>
</ul>
</dd>
<dt class="field-even">Returns</dt>
<dd class="field-even"><p>A <a class="reference internal" href="#torch.jit.ScriptModule" title="torch.jit.ScriptModule"><code class="xref py py-class docutils literal notranslate"><span class="pre">ScriptModule</span></code></a> object.</p>
</dd>
</dl>
<p>Example:</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">io</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="s1">'scriptmodule.pt'</span><span class="p">)</span>
<span class="c1"># Load ScriptModule from io.BytesIO object</span>
<span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="s1">'scriptmodule.pt'</span><span class="p">,</span> <span class="s1">'rb'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
<span class="n">buffer</span> <span class="o">=</span> <span class="n">io</span><span class="o">.</span><span class="n">BytesIO</span><span class="p">(</span><span class="n">f</span><span class="o">.</span><span class="n">read</span><span class="p">())</span>
<span class="c1"># Load all tensors to the original device</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">buffer</span><span class="p">)</span>
<span class="c1"># Load all tensors onto CPU, using a device</span>
<span class="n">buffer</span><span class="o">.</span><span class="n">seek</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="n">map_location</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s1">'cpu'</span><span class="p">))</span>
<span class="c1"># Load all tensors onto CPU, using a string</span>
<span class="n">buffer</span><span class="o">.</span><span class="n">seek</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">buffer</span><span class="p">,</span> <span class="n">map_location</span><span class="o">=</span><span class="s1">'cpu'</span><span class="p">)</span>
<span class="c1"># Load with extra files.</span>
<span class="n">extra_files</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">_C</span><span class="o">.</span><span class="n">ExtraFilesMap</span><span class="p">()</span>
<span class="n">extra_files</span><span class="p">[</span><span class="s1">'foo.txt'</span><span class="p">]</span> <span class="o">=</span> <span class="s1">'bar'</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="s1">'scriptmodule.pt'</span><span class="p">,</span> <span class="n">_extra_files</span><span class="o">=</span><span class="n">extra_files</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="n">extra_files</span><span class="p">[</span><span class="s1">'foo.txt'</span><span class="p">])</span>
</pre></div>
</div>
</dd></dl>
<dl class="function">
<dt id="torch.jit.ignore">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">ignore</code><span class="sig-paren">(</span><em class="sig-param">drop=False</em>, <em class="sig-param">**kwargs</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/_jit_internal.html#ignore"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.ignore" title="Permalink to this definition">¶</a></dt>
<dd><p>This decorator indicates to the compiler that a function or method should
be ignored and left as a Python function. This allows you to leave code in
your model that is not yet TorchScript compatible. If called from TorchScript,
ignored functions will dispatch the call to the Python interpreter. Models with ignored
functions cannot be exported; use <a class="reference internal" href="#torch.jit.unused" title="torch.jit.unused"><code class="xref py py-func docutils literal notranslate"><span class="pre">@torch.jit.unused</span></code></a> instead.</p>
<p>Example (using <code class="docutils literal notranslate"><span class="pre">@torch.jit.ignore</span></code> on a method):</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">ignore</span>
<span class="k">def</span> <span class="nf">debugger</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="kn">import</span> <span class="nn">pdb</span>
<span class="n">pdb</span><span class="o">.</span><span class="n">set_trace</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="n">x</span> <span class="o">+=</span> <span class="mi">10</span>
<span class="c1"># The compiler would normally try to compile `debugger`,</span>
<span class="c1"># but since it is `@ignore`d, it will be left as a call</span>
<span class="c1"># to Python</span>
<span class="bp">self</span><span class="o">.</span><span class="n">debugger</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
<span class="n">m</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">())</span>
<span class="c1"># Error! The call `debugger` cannot be saved since it calls into Python</span>
<span class="n">m</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s2">"m.pt"</span><span class="p">)</span>
</pre></div>
</div>
<p>Example (using <code class="docutils literal notranslate"><span class="pre">@torch.jit.ignore(drop=True)</span></code> on a method):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">ignore</span><span class="p">(</span><span class="n">drop</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">training_method</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="kn">import</span> <span class="nn">pdb</span>
<span class="n">pdb</span><span class="o">.</span><span class="n">set_trace</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">training</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">training_method</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
<span class="n">m</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">())</span>
<span class="c1"># This is OK since `training_method` is not saved, the call is replaced</span>
<span class="c1"># with a `raise`.</span>
<span class="n">m</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s2">"m.pt"</span><span class="p">)</span>
</pre></div>
</div>
</dd></dl>
<dl class="function">
<dt id="torch.jit.unused">
<code class="sig-prename descclassname">torch.jit.</code><code class="sig-name descname">unused</code><span class="sig-paren">(</span><em class="sig-param">fn</em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/_jit_internal.html#unused"><span class="viewcode-link">[source]</span></a><a class="headerlink" href="#torch.jit.unused" title="Permalink to this definition">¶</a></dt>
<dd><p>This decorator indicates to the compiler that a function or method should
be ignored and replaced with the raising of an exception. This allows you
to leave code in your model that is not yet TorchScript compatible and still
export your model.</p>
<blockquote>
<div><p>Example (using <code class="docutils literal notranslate"><span class="pre">@torch.jit.unused</span></code> on a method):</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="k">class</span> <span class="nc">MyModule</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">use_memory_efficent</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">MyModule</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">use_memory_efficent</span> <span class="o">=</span> <span class="n">use_memory_efficent</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">unused</span>
<span class="k">def</span> <span class="nf">memory_efficient</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="kn">import</span> <span class="nn">pdb</span>
<span class="n">pdb</span><span class="o">.</span><span class="n">set_trace</span><span class="p">()</span>
<span class="k">return</span> <span class="n">x</span> <span class="o">+</span> <span class="mi">10</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="c1"># Use not-yet-scriptable memory efficient mode</span>
<span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_memory_efficient</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">memory_efficient</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">x</span> <span class="o">+</span> <span class="mi">10</span>
<span class="n">m</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">(</span><span class="n">use_memory_efficent</span><span class="o">=</span><span class="kc">False</span><span class="p">))</span>
<span class="n">m</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="s2">"m.pt"</span><span class="p">)</span>
<span class="n">m</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">MyModule</span><span class="p">(</span><span class="n">use_memory_efficient</span><span class="o">=</span><span class="kc">True</span><span class="p">))</span>
<span class="c1"># exception raised</span>
<span class="n">m</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">100</span><span class="p">))</span>
</pre></div>
</div>
</div></blockquote>
</dd></dl>
</div>
<div class="section" id="mixing-tracing-and-scripting">
<h2><a class="toc-backref" href="#id5">Mixing Tracing and Scripting</a><a class="headerlink" href="#mixing-tracing-and-scripting" title="Permalink to this headline">¶</a></h2>
<p>In many cases either tracing or scripting is an easier approach for converting a model to TorchScript.
Tracing and scripting can be composed to suit the particular requirements
of a part of a model.</p>
<p>Scripted functions can call traced functions. This is particularly useful when you need
to use control-flow around a simple feed-forward model. For instance the beam search
of a sequence to sequence model will typically be written in script but can call an
encoder module generated using tracing.</p>
<p>Example (calling a traced function in script):</p>
<div class="highlight-python3 notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="k">def</span> <span class="nf">foo</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">y</span><span class="p">):</span>
<span class="k">return</span> <span class="mi">2</span> <span class="o">*</span> <span class="n">x</span> <span class="o">+</span> <span class="n">y</span>
<span class="n">traced_foo</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">trace</span><span class="p">(</span><span class="n">foo</span><span class="p">,</span> <span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">3</span><span class="p">),</span> <span class="n">torch</span><span class="o">.</span><span class="n">rand</span><span class="p">(</span><span class="mi">3</span><span class="p">)))</span>
<span class="nd">@torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span>
<span class="k">def</span> <span class="nf">bar</span><span class="p">(</span><span class="n">x</span><span class="p">):</span>
<span class="k">return</span> <span class="n">traced_foo</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">x</span><span class="p">)</span>
</pre></div>
</div>