forked from pytorch/pytorch.github.io
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathget-started.html
973 lines (776 loc) · 57.8 KB
/
get-started.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
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta name="generator" content="Docutils 0.18.1: http://docutils.sourceforge.net/" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Getting Started — PyTorch master documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/dynamo/get-started.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="../_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="../_static/copybutton.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="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="stylesheet" href="../_static/sphinx-dropdown.css" type="text/css" />
<link rel="stylesheet" href="../_static/panels-bootstrap.min.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/jit.css" type="text/css" />
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="Guards Overview" href="guards-overview.html" />
<link rel="prev" title="Installing TorchDynamo" href="installation.html" />
<!--
Search engines should not index the master version of documentation.
Stable documentation are built without release == 'master'.
-->
<meta name="robots" content="noindex">
<!-- Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-117752657-2"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-117752657-2');
</script>
<!-- End Google Analytics -->
<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">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.15.2/css/all.css" integrity="sha384-vSIIfh2YWi9wW0r9iZe7RJPrKwp6bG+s9QZMoITbCckVJqGCCRhc+ccxNcdpHuYu" 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>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</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 docs-active">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-orange-arrow">
Docs
</a>
<div class="resources-dropdown-menu">
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/docs/stable/index.html">
<span class="dropdown-title">PyTorch</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/audio/stable/index.html">
<span class="dropdown-title">torchaudio</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/text/stable/index.html">
<span class="dropdown-title">torchtext</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/vision/stable/index.html">
<span class="dropdown-title">torchvision</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torcharrow">
<span class="dropdown-title">torcharrow</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/data">
<span class="dropdown-title">TorchData</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torchrec">
<span class="dropdown-title">TorchRec</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/serve/">
<span class="dropdown-title">TorchServe</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torchx/">
<span class="dropdown-title">TorchX</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/xla">
<span class="dropdown-title">PyTorch on XLA Devices</span>
<p></p>
</a>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-arrow">
Resources
</a>
<div class="resources-dropdown-menu">
<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>
<a class="nav-dropdown-item" href="https://pytorch.org/foundation">
<span class="dropdown-title">PyTorch Foundation</span>
<p>Learn about the PyTorch foundation</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/#community-module">
<span class="dropdown-title">Community</span>
<p>Join the PyTorch developer community to contribute, learn, and get your questions answered.</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/community-stories">
<span class="dropdown-title">Community Stories</span>
<p>Learn how our community solves real, everyday machine learning problems with PyTorch.</p>
</a>
<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/events">
<span class="dropdown-title">Events</span>
<p>Find events, webinars, and podcasts</p>
</a>
<a class="nav-dropdown-item" href="https://discuss.pytorch.org/" target="_blank">
<span class="dropdown-title">Forums</span>
<p>A place to discuss PyTorch code, issues, install, research</p>
</a>
<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>
</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='https://pytorch.org/docs/versions.html'>master (1.14.0a0+git876b702 ) ▼</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>
<div>
<a style="color:#F05732" href="https://pytorch.org/docs/stable/dynamo/get-started.html">
You are viewing unstable developer preview docs.
Click here to view docs for latest stable release.
</a>
</div>
<p class="caption" role="heading"><span class="caption-text">Community</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../community/build_ci_governance.html">PyTorch Governance | Build + CI</a></li>
<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/design.html">PyTorch Design Philosophy</a></li>
<li class="toctree-l1"><a class="reference internal" href="../community/governance.html">PyTorch Governance | Mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="../community/persons_of_interest.html">PyTorch Governance | Maintainers</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Developer Notes</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../notes/amp_examples.html">CUDA 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/gradcheck.html">Gradcheck mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="../notes/hip.html">HIP (ROCm) semantics</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/modules.html">Modules</a></li>
<li class="toctree-l1"><a class="reference internal" href="../notes/mps.html">MPS backend</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/numerical_accuracy.html">Numerical accuracy</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" role="heading"><span class="caption-text">torch.compile</span></p>
<ul class="current">
<li class="toctree-l1 current"><a class="reference internal" href="index.html">TorchDynamo Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="installation.html">Installing TorchDynamo</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">Getting Started</a></li>
<li class="toctree-l1"><a class="reference internal" href="guards-overview.html">Guards Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="custom-backends.html">Custom Backends</a></li>
<li class="toctree-l1"><a class="reference internal" href="deep-dive.html">TorchDynamo Deeper Dive</a></li>
<li class="toctree-l1"><a class="reference internal" href="troubleshooting.html">TorchDynamo Troubleshooting</a></li>
<li class="toctree-l1"><a class="reference internal" href="faq.html">Frequently Asked Questions</a></li>
</ul>
<p class="caption" role="heading"><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 external" href="https://pytorch.org/javadoc/">Javadoc</a></li>
<li class="toctree-l1"><a class="reference internal" href="../deploy.html">torch::deploy</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Python API</span></p>
<ul>
<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="../amp.html">torch.amp</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="../library.html">torch.library</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="../backends.html">torch.backends</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="../distributed.algorithms.join.html">torch.distributed.algorithms.join</a></li>
<li class="toctree-l1"><a class="reference internal" href="../distributed.elastic.html">torch.distributed.elastic</a></li>
<li class="toctree-l1"><a class="reference internal" href="../fsdp.html">torch.distributed.fsdp</a></li>
<li class="toctree-l1"><a class="reference internal" href="../distributed.optim.html">torch.distributed.optim</a></li>
<li class="toctree-l1"><a class="reference internal" href="../distributed.tensor.parallel.html">torch.distributed.tensor.parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="../distributed.checkpoint.html">torch.distributed.checkpoint</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="../_dynamo.html">torch._dynamo</a></li>
<li class="toctree-l1"><a class="reference internal" href="../fft.html">torch.fft</a></li>
<li class="toctree-l1"><a class="reference internal" href="../futures.html">torch.futures</a></li>
<li class="toctree-l1"><a class="reference internal" href="../fx.html">torch.fx</a></li>
<li class="toctree-l1"><a class="reference internal" href="../hub.html">torch.hub</a></li>
<li class="toctree-l1"><a class="reference internal" href="../jit.html">torch.jit</a></li>
<li class="toctree-l1"><a class="reference internal" href="../linalg.html">torch.linalg</a></li>
<li class="toctree-l1"><a class="reference internal" href="../monitor.html">torch.monitor</a></li>
<li class="toctree-l1"><a class="reference internal" href="../signal.html">torch.signal</a></li>
<li class="toctree-l1"><a class="reference internal" href="../special.html">torch.special</a></li>
<li class="toctree-l1"><a class="reference internal" href="../torch.overrides.html">torch.overrides</a></li>
<li class="toctree-l1"><a class="reference internal" href="../package.html">torch.package</a></li>
<li class="toctree-l1"><a class="reference internal" href="../profiler.html">torch.profiler</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="../onnx_diagnostics.html">torch.onnx diagnostics</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="../complex_numbers.html">Complex Numbers</a></li>
<li class="toctree-l1"><a class="reference internal" href="../ddp_comm_hooks.html">DDP Communication Hooks</a></li>
<li class="toctree-l1"><a class="reference internal" href="../pipeline.html">Pipeline Parallelism</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.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="../masked.html">torch.masked</a></li>
<li class="toctree-l1"><a class="reference internal" href="../nested.html">torch.nested</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="../testing.html">torch.testing</a></li>
<li class="toctree-l1"><a class="reference internal" href="../benchmark_utils.html">torch.utils.benchmark</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="../jit_utils.html">torch.utils.jit</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="../mobile_optimizer.html">torch.utils.mobile_optimizer</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_mod.html">torch.__config__</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Libraries</span></p>
<ul>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/audio/stable">torchaudio</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/data">TorchData</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/torchrec">TorchRec</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="https://pytorch.org/text/stable">torchtext</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/vision/stable">torchvision</a></li>
<li class="toctree-l1"><a class="reference external" href="http://pytorch.org/xla/">PyTorch on XLA Devices</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><a href="index.html">TorchDynamo Overview</a> ></li>
<li>Getting Started</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/dynamo/get-started.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">
<section id="getting-started">
<h1>Getting Started<a class="headerlink" href="#getting-started" title="Permalink to this heading">¶</a></h1>
<p>Let’s start with a simple example. Note that you are likely to see more
significant speedups the newer your GPU is.</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">torch._dynamo</span> <span class="kn">import</span> <span class="n">optimize</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="k">def</span> <span class="nf">fn</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="n">a</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cos</span><span class="p">(</span><span class="n">x</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">()</span>
<span class="n">b</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sin</span><span class="p">(</span><span class="n">y</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">()</span>
<span class="k">return</span> <span class="n">a</span> <span class="o">+</span> <span class="n">b</span>
<span class="n">new_fn</span> <span class="o">=</span> <span class="n">optimize</span><span class="p">(</span><span class="s2">"inductor"</span><span class="p">)(</span><span class="n">fn</span><span class="p">)</span>
<span class="n">input_tensor</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">10000</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="s2">"cuda:0"</span><span class="p">)</span>
<span class="n">a</span> <span class="o">=</span> <span class="n">new_fn</span><span class="p">(</span><span class="n">input_tensor</span><span class="p">,</span> <span class="n">input_tensor</span><span class="p">)</span>
</pre></div>
</div>
<p>This example will not actually run faster. Its purpose is to demonstrate
the <code class="docutils literal notranslate"><span class="pre">torch.cos()</span></code> and <code class="docutils literal notranslate"><span class="pre">torch.sin()</span></code> features which are
examples of pointwise ops as in they operate element by element on a
vector. A more famous pointwise op you might want to use would
be something like <code class="docutils literal notranslate"><span class="pre">torch.relu()</span></code>. Pointwise ops in eager mode are
suboptimal because each one would need to read a tensor from
memory, make some changes, and then write back those changes. The single
most important optimization that inductor does is fusion. So back to our
example we can turn 2 reads and 2 writes into 1 read and 1 write which
is crucial especially for newer GPUs where the bottleneck is memory
bandwidth (how quickly you can send data to a GPU) rather than compute
(how quickly your GPU can crunch floating point operations).</p>
<p>Another major optimization that inductor makes available is automatic
support for CUDA graphs.
CUDA graphs help eliminate the overhead from launching individual
kernels from a Python program which is especially relevant for newer GPUs.</p>
<p>TorchDynamo supports many different backends but inductor specifically works
by generating <a class="reference external" href="https://github.com/openai/triton">Triton</a> kernels and
we can inspect them by running <code class="docutils literal notranslate"><span class="pre">TORCHINDUCTOR_TRACE=1</span> <span class="pre">python</span> <span class="pre">trig.py</span></code>
with the actual generated kernel being</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@pointwise</span><span class="p">(</span><span class="n">size_hints</span><span class="o">=</span><span class="p">[</span><span class="mi">16384</span><span class="p">],</span> <span class="n">filename</span><span class="o">=</span><span class="vm">__file__</span><span class="p">,</span> <span class="n">meta</span><span class="o">=</span><span class="p">{</span><span class="s1">'signature'</span><span class="p">:</span> <span class="p">{</span><span class="mi">0</span><span class="p">:</span> <span class="s1">'*fp32'</span><span class="p">,</span> <span class="mi">1</span><span class="p">:</span> <span class="s1">'*fp32'</span><span class="p">,</span> <span class="mi">2</span><span class="p">:</span> <span class="s1">'i32'</span><span class="p">},</span> <span class="s1">'device'</span><span class="p">:</span> <span class="mi">0</span><span class="p">,</span> <span class="s1">'constants'</span><span class="p">:</span> <span class="p">{},</span> <span class="s1">'configs'</span><span class="p">:</span> <span class="p">[</span><span class="n">instance_descriptor</span><span class="p">(</span><span class="n">divisible_by_16</span><span class="o">=</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">),</span> <span class="n">equal_to_1</span><span class="o">=</span><span class="p">())]})</span>
<span class="nd">@triton</span><span class="o">.</span><span class="n">jit</span>
<span class="k">def</span> <span class="nf">kernel</span><span class="p">(</span><span class="n">in_ptr0</span><span class="p">,</span> <span class="n">out_ptr0</span><span class="p">,</span> <span class="n">xnumel</span><span class="p">,</span> <span class="n">XBLOCK</span> <span class="p">:</span> <span class="n">tl</span><span class="o">.</span><span class="n">constexpr</span><span class="p">):</span>
<span class="n">xnumel</span> <span class="o">=</span> <span class="mi">10000</span>
<span class="n">xoffset</span> <span class="o">=</span> <span class="n">tl</span><span class="o">.</span><span class="n">program_id</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">*</span> <span class="n">XBLOCK</span>
<span class="n">xindex</span> <span class="o">=</span> <span class="n">xoffset</span> <span class="o">+</span> <span class="n">tl</span><span class="o">.</span><span class="n">reshape</span><span class="p">(</span><span class="n">tl</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">XBLOCK</span><span class="p">),</span> <span class="p">[</span><span class="n">XBLOCK</span><span class="p">])</span>
<span class="n">xmask</span> <span class="o">=</span> <span class="n">xindex</span> <span class="o"><</span> <span class="n">xnumel</span>
<span class="n">x0</span> <span class="o">=</span> <span class="n">xindex</span>
<span class="n">tmp0</span> <span class="o">=</span> <span class="n">tl</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">in_ptr0</span> <span class="o">+</span> <span class="p">(</span><span class="n">x0</span><span class="p">),</span> <span class="n">xmask</span><span class="p">)</span>
<span class="n">tmp1</span> <span class="o">=</span> <span class="n">tl</span><span class="o">.</span><span class="n">sin</span><span class="p">(</span><span class="n">tmp0</span><span class="p">)</span>
<span class="n">tmp2</span> <span class="o">=</span> <span class="n">tl</span><span class="o">.</span><span class="n">sin</span><span class="p">(</span><span class="n">tmp1</span><span class="p">)</span>
<span class="n">tl</span><span class="o">.</span><span class="n">store</span><span class="p">(</span><span class="n">out_ptr0</span> <span class="o">+</span> <span class="p">(</span><span class="n">x0</span> <span class="o">+</span> <span class="n">tl</span><span class="o">.</span><span class="n">zeros</span><span class="p">([</span><span class="n">XBLOCK</span><span class="p">],</span> <span class="n">tl</span><span class="o">.</span><span class="n">int32</span><span class="p">)),</span> <span class="n">tmp2</span><span class="p">,</span> <span class="n">xmask</span><span class="p">)</span>
</pre></div>
</div>
<p>And you can verify that fusing the two <code class="docutils literal notranslate"><span class="pre">sins</span></code> did actually occur
because the two <code class="docutils literal notranslate"><span class="pre">sin</span></code> operations occur within a single Triton kernel
and the temporary variables are held in registers with very fast access.</p>
<p>You can read up a lot more on Triton’s performance
<a class="reference external" href="https://openai.com/blog/triton/">here</a> but the key is it’s in Python
so you can easily understand it even if you have not written all that
many CUDA kernels.</p>
<p>Next, let’s try a real model like resnet50 from the PyTorch
hub.</p>
<div class="highlight-python 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._dynamo</span> <span class="k">as</span> <span class="nn">dynamo</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">hub</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="s1">'pytorch/vision:v0.10.0'</span><span class="p">,</span> <span class="s1">'resnet18'</span><span class="p">,</span> <span class="n">pretrained</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">opt_model</span> <span class="o">=</span> <span class="n">dynamo</span><span class="o">.</span><span class="n">optimize</span><span class="p">(</span><span class="s2">"inductor"</span><span class="p">)(</span><span class="n">model</span><span class="p">)</span>
<span class="n">model</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">1</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">64</span><span class="p">,</span><span class="mi">64</span><span class="p">))</span>
</pre></div>
</div>
<p>And that is not the only available backend, you can run in a REPL
<code class="docutils literal notranslate"><span class="pre">dynamo.list_backends()</span></code> to see all the available backends. Try out the
<code class="docutils literal notranslate"><span class="pre">aot_cudagraphs</span></code> or <code class="docutils literal notranslate"><span class="pre">nvfuser</span></code> next as inspiration.</p>
<p>Let’s do something a bit more interesting now, our community frequently
uses pretrained models from
<a class="reference external" href="https://github.com/huggingface/transformers">transformers</a> or
<a class="reference external" href="https://github.com/rwightman/pytorch-image-models">TIMM</a> and one of
our design goals is for Dynamo and inductor to work out of the box with
any model that people would like to author.</p>
<p>So we will directly download a pretrained model from the
HuggingFace hub and optimize it:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">from</span> <span class="nn">transformers</span> <span class="kn">import</span> <span class="n">BertTokenizer</span><span class="p">,</span> <span class="n">BertModel</span>
<span class="kn">import</span> <span class="nn">torch._dynamo</span> <span class="k">as</span> <span class="nn">dynamo</span>
<span class="c1"># Copy pasted from here https://huggingface.co/bert-base-uncased</span>
<span class="n">tokenizer</span> <span class="o">=</span> <span class="n">BertTokenizer</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s1">'bert-base-uncased'</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">BertModel</span><span class="o">.</span><span class="n">from_pretrained</span><span class="p">(</span><span class="s2">"bert-base-uncased"</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="s2">"cuda:0"</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">dynamo</span><span class="o">.</span><span class="n">optimize</span><span class="p">(</span><span class="s2">"inductor"</span><span class="p">)(</span><span class="n">model</span><span class="p">)</span> <span class="c1"># This is the only line of code that we changed</span>
<span class="n">text</span> <span class="o">=</span> <span class="s2">"Replace me by any text you'd like."</span>
<span class="n">encoded_input</span> <span class="o">=</span> <span class="n">tokenizer</span><span class="p">(</span><span class="n">text</span><span class="p">,</span> <span class="n">return_tensors</span><span class="o">=</span><span class="s1">'pt'</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="s2">"cuda:0"</span><span class="p">)</span>
<span class="n">output</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="o">**</span><span class="n">encoded_input</span><span class="p">)</span>
</pre></div>
</div>
<p>If you remove the <code class="docutils literal notranslate"><span class="pre">to(device="cuda:0")</span></code> from the model and
<code class="docutils literal notranslate"><span class="pre">encoded_input</span></code>, then Triton will generate C++ kernels that will be
optimized for running on your CPU. You can inspect both Triton or C++
kernels for BERT, they’re obviously more complex than the trigonometry
example we had above but you can similarly skim it and understand if you
understand PyTorch.</p>
<p>Similarly let’s try out a TIMM example</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">timm</span>
<span class="kn">import</span> <span class="nn">torch._dynamo</span> <span class="k">as</span> <span class="nn">dynamo</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">timm</span><span class="o">.</span><span class="n">create_model</span><span class="p">(</span><span class="s1">'resnext101_32x8d'</span><span class="p">,</span> <span class="n">pretrained</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">num_classes</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
<span class="n">opt_model</span> <span class="o">=</span> <span class="n">dynamo</span><span class="o">.</span><span class="n">optimize</span><span class="p">(</span><span class="s2">"inductor"</span><span class="p">)(</span><span class="n">model</span><span class="p">)</span>
<span class="n">opt_model</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">64</span><span class="p">,</span><span class="mi">3</span><span class="p">,</span><span class="mi">7</span><span class="p">,</span><span class="mi">7</span><span class="p">))</span>
</pre></div>
</div>
<p>Our goal with Dynamo and inductor is to build the highest coverage ML compiler
which should work with any model you throw at it.</p>
<section id="existing-backends">
<h2>Existing Backends<a class="headerlink" href="#existing-backends" title="Permalink to this heading">¶</a></h2>
<p>TorchDynamo has a growing list of backends, which can be found in
<a class="reference external" href="https://github.com/pytorch/pytorch/blob/master/torch/_dynamo/optimizations/backends.py">backends.py</a>
or <code class="docutils literal notranslate"><span class="pre">torchdynamo.list_backends()</span></code> each of which with its optional dependencies.</p>
<p>Some of the most commonly used backends include:</p>
<ul class="simple">
<li><p><strong>Debugging backends</strong>:
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("eager")</span></code> - Uses PyTorch
to run the extracted GraphModule. This is quite useful in debugging
TorchDynamo issues.
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("aot_eager")</span></code> - Uses
AotAutograd with no compiler, for example, just using PyTorch eager for the
AotAutograd’s extracted forward and backward graphs. This is useful for
debugging, and unlikely to give speedups.</p></li>
<li><p><strong>Training & inference backends</strong>:
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("inductor")</span></code> - Uses <code class="docutils literal notranslate"><span class="pre">TorchInductor</span></code> backend
with AotAutograd and cudagraphs by leveraging
codegened Triton kernels <a class="reference external" href="https://dev-discuss.pytorch.org/t/torchinductor-a-pytorch-native-compiler-with-define-by-run-ir-and-symbolic-shapes/747">Read
more</a>
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("nvfuser")</span></code> - nvFuser with TorchScript. <a class="reference external" href="https://dev-discuss.pytorch.org/t/tracing-with-primitives-update-1-nvfuser-and-its-primitives/593">Read more</a>
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("aot_nvfuser")</span></code> - nvFuser with AotAutograd. <a class="reference external" href="https://dev-discuss.pytorch.org/t/tracing-with-primitives-update-1-nvfuser-and-its-primitives/593">Read more</a>
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("aot_cudagraphs")</span></code> - cudagraphs with AotAutograd. <a class="reference external" href="https://github.com/pytorch/torchdynamo/pull/757">Read more</a></p></li>
<li><p><strong>Inference-only backends</strong>:
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("ofi")</span></code> - Uses
Torchscript <code class="docutils literal notranslate"><span class="pre">optimize_for_inference</span></code>. <a class="reference external" href="https://pytorch.org/docs/stable/generated/torch.jit.optimize_for_inference.html">Read
more</a>
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("fx2trt")</span></code> - Uses Nvidia TensorRT for inference optimizations. <a class="reference external" href="https://github.com/pytorch/TensorRT/blob/master/docsrc/tutorials/getting_started_with_fx_path.rst">Read more</a>
* <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("onnxrt")</span></code> - Uses ONNXRT for inference on CPU/GPU. <a class="reference external" href="https://onnxruntime.ai/">Read more</a> * <code class="docutils literal notranslate"><span class="pre">dynamo.optimize("ipex")</span></code> - Uses IPEX for inference on CPU. <a class="reference external" href="https://github.com/intel/intel-extension-for-pytorch">Read more</a></p></li>
</ul>
<section id="why-do-you-need-another-way-of-optimizing-pytorch-code">
<h3>Why do you need another way of optimizing PyTorch code?<a class="headerlink" href="#why-do-you-need-another-way-of-optimizing-pytorch-code" title="Permalink to this heading">¶</a></h3>
<p>While a number of other code optimization tools exist in the PyTorch
ecosystem, each of them has its own flow.
Here is a few examples of existing methods and their limitations:</p>
<ul class="simple">
<li><p><code class="docutils literal notranslate"><span class="pre">torch.jit.trace()</span></code> is silently wrong if it cannot trace, for example:
during control flow</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">torch.jit.script()</span></code> requires modifications to user or library code
by adding type annotations and removing non PyTorch code</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">torch.fx.symbolic_trace()</span></code> either traces correctly or gives a hard
error but it’s limited to traceable code so still can’t handle
control flow</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">torch._dynamo</span></code> works out of the box and produces partial graphs.
It still has the option of producing a single graph with
<code class="docutils literal notranslate"><span class="pre">nopython=True</span></code> which are needed for <a class="reference external" href="./documentation/FAQ.md#do-i-still-need-to-export-whole-graphs">some
situations</a>
but allows a smoother transition where partial graphs can be
optimized without code modification</p></li>
</ul>
</section>
</section>
</section>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="guards-overview.html" class="btn btn-neutral float-right" title="Guards Overview" accesskey="n" rel="next">Next <img src="../_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="installation.html" class="btn btn-neutral" title="Installing TorchDynamo" accesskey="p" rel="prev"><img src="../_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr>
<div role="contentinfo">
<p>
© Copyright 2022, PyTorch Contributors.
</p>
</div>
<div>
Built with <a href="http://sphinx-doc.org/">Sphinx</a> using a <a href="https://github.com/rtfd/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
</div>
</footer>
</div>
</div>
<div class="pytorch-content-right" id="pytorch-content-right">
<div class="pytorch-right-menu" id="pytorch-right-menu">
<div class="pytorch-side-scroll" id="pytorch-side-scroll-right">
<ul>
<li><a class="reference internal" href="#">Getting Started</a><ul>
<li><a class="reference internal" href="#existing-backends">Existing Backends</a><ul>
<li><a class="reference internal" href="#why-do-you-need-another-way-of-optimizing-pytorch-code">Why do you need another way of optimizing PyTorch code?</a></li>
</ul>
</li>
</ul>
</li>
</ul>
</div>
</div>
</div>
</section>
</div>
<script type="text/javascript" id="documentation_options" data-url_root="../" src="../_static/documentation_options.js"></script>
<script data-url_root="../" id="documentation_options" src="../_static/documentation_options.js"></script>
<script src="../_static/jquery.js"></script>
<script src="../_static/underscore.js"></script>
<script src="../_static/_sphinx_javascript_frameworks_compat.js"></script>
<script src="../_static/doctools.js"></script>
<script src="../_static/clipboard.min.js"></script>
<script src="../_static/copybutton.js"></script>
<script type="text/javascript" src="../_static/js/vendor/popper.min.js"></script>
<script type="text/javascript" src="../_static/js/vendor/bootstrap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/list.js/1.5.0/list.min.js"></script>
<script type="text/javascript" src="../_static/js/theme.js"></script>
<script type="text/javascript">
jQuery(function () {
SphinxRtdTheme.Navigation.enable(true);
});
</script>
<script script type="text/javascript">
var collapsedSections = ['Developer Notes', 'Language Bindings', 'Libraries', 'Community'];
</script>
<img height="1" width="1" style="border-style:none;" alt="" src="https://www.googleadservices.com/pagead/conversion/795629140/?label=txkmCPmdtosBENSssfsC&guid=ON&script=0"/>
<!-- Begin Footer -->
<div class="container-fluid docs-tutorials-resources" id="docs-tutorials-resources">
<div class="container">
<div class="row">
<div class="col-md-4 text-center">
<h2>Docs</h2>
<p>Access comprehensive developer documentation for PyTorch</p>
<a class="with-right-arrow" href="https://pytorch.org/docs/stable/index.html">View Docs</a>
</div>
<div class="col-md-4 text-center">
<h2>Tutorials</h2>
<p>Get in-depth tutorials for beginners and advanced developers</p>
<a class="with-right-arrow" href="https://pytorch.org/tutorials">View Tutorials</a>
</div>
<div class="col-md-4 text-center">
<h2>Resources</h2>
<p>Find development resources and get your questions answered</p>
<a class="with-right-arrow" href="https://pytorch.org/resources">View Resources</a>
</div>
</div>
</div>
</div>
<footer class="site-footer">
<div class="container footer-container">
<div class="footer-logo-wrapper">
<a href="https://pytorch.org/" class="footer-logo"></a>
</div>
<div class="footer-links-wrapper">
<div class="footer-links-col">
<ul>
<li class="list-title"><a href="https://pytorch.org/">PyTorch</a></li>
<li><a href="https://pytorch.org/get-started">Get Started</a></li>
<li><a href="https://pytorch.org/features">Features</a></li>
<li><a href="https://pytorch.org/ecosystem">Ecosystem</a></li>
<li><a href="https://pytorch.org/blog/">Blog</a></li>
<li><a href="https://github.com/pytorch/pytorch/blob/master/CONTRIBUTING.md">Contributing</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title"><a href="https://pytorch.org/resources">Resources</a></li>
<li><a href="https://pytorch.org/tutorials">Tutorials</a></li>
<li><a href="https://pytorch.org/docs/stable/index.html">Docs</a></li>
<li><a href="https://discuss.pytorch.org" target="_blank">Discuss</a></li>
<li><a href="https://github.com/pytorch/pytorch/issues" target="_blank">Github Issues</a></li>
<li><a href="https://pytorch.org/assets/brand-guidelines/PyTorch-Brand-Guidelines.pdf" target="_blank">Brand Guidelines</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title">Stay up to date</li>
<li><a href="https://www.facebook.com/pytorch" target="_blank">Facebook</a></li>
<li><a href="https://twitter.com/pytorch" target="_blank">Twitter</a></li>
<li><a href="https://www.youtube.com/pytorch" target="_blank">YouTube</a></li>
<li><a href="https://www.linkedin.com/company/pytorch" target="_blank">LinkedIn</a></li>
</ul>
</div>
<div class="footer-links-col">
<ul>
<li class="list-title">PyTorch Podcasts</li>
<li><a href="https://open.spotify.com/show/6UzHKeiy368jKfQMKKvJY5" target="_blank">Spotify</a></li>
<li><a href="https://podcasts.apple.com/us/podcast/pytorch-developer-podcast/id1566080008" target="_blank">Apple</a></li>
<li><a href="https://www.google.com/podcasts?feed=aHR0cHM6Ly9mZWVkcy5zaW1wbGVjYXN0LmNvbS9PQjVGa0lsOA%3D%3D" target="_blank">Google</a></li>
<li><a href="https://music.amazon.com/podcasts/7a4e6f0e-26c2-49e9-a478-41bd244197d0/PyTorch-Developer-Podcast?" target="_blank">Amazon</a></li>
</ul>
</div>
</div>
<div class="privacy-policy">
<ul>
<li class="privacy-policy-links"><a href="https://www.linuxfoundation.org/terms/" target="_blank">Terms</a></li>
<li class="privacy-policy-links">|</li>
<li class="privacy-policy-links"><a href="https://www.linuxfoundation.org/privacy-policy/" target="_blank">Privacy</a></li>
</ul>
</div>
<div class="copyright">
<p>© Copyright The Linux Foundation. The PyTorch Foundation is a project of The Linux Foundation.
For web site terms of use, trademark policy and other policies applicable to The PyTorch Foundation please see
<a href="www.linuxfoundation.org/policies/">www.linuxfoundation.org/policies/</a>. The PyTorch Foundation supports the PyTorch open source
project, which has been established as PyTorch Project a Series of LF Projects, LLC. For policies applicable to the PyTorch Project a Series of LF Projects, LLC,
please see <a href="www.lfprojects.org/policies/">www.lfprojects.org/policies/</a>.</p>
</div>
</div>
</footer>
<div class="cookie-banner-wrapper">
<div class="container">
<p class="gdpr-notice">To analyze traffic and optimize your experience, we serve cookies on this site. By clicking or navigating, you agree to allow our usage of cookies. As the current maintainers of this site, Facebook’s Cookies Policy applies. Learn more, including about available controls: <a href="https://www.facebook.com/policies/cookies/">Cookies Policy</a>.</p>
<img class="close-button" src="../_static/images/pytorch-x.svg">
</div>
</div>
<!-- End Footer -->
<!-- Begin Mobile Menu -->
<div class="mobile-main-menu">
<div class="container-fluid">
<div class="container">
<div class="mobile-main-menu-header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<a class="main-menu-close-button" href="#" data-behavior="close-mobile-menu"></a>
</div>
</div>
</div>
<div class="mobile-main-menu-links-container">
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</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="resources-mobile-menu-title" class="active">
Docs
</li>
<ul class="resources-mobile-menu-items">
<li>
<a href="https://pytorch.org/docs/stable/index.html">PyTorch</a>
</li>
<li>
<a href="https://pytorch.org/audio/stable/index.html">torchaudio</a>
</li>
<li>
<a href="https://pytorch.org/text/stable/index.html">torchtext</a>
</li>
<li>
<a href="https://pytorch.org/vision/stable/index.html">torchvision</a>
</li>
<li>
<a href="https://pytorch.org/torcharrow">torcharrow</a>
</li>
<li>
<a href="https://pytorch.org/data">TorchData</a>
</li>
<li>
<a href="https://pytorch.org/torchrec">TorchRec</a>
</li>
<li>
<a href="https://pytorch.org/serve/">TorchServe</a>
</li>
<li>
<a href="https://pytorch.org/torchx/">TorchX</a>
</li>
<li>
<a href="https://pytorch.org/xla">PyTorch on XLA Devices</a>
</li>
</ul>
<li class="resources-mobile-menu-title">
Resources
</li>
<ul class="resources-mobile-menu-items">
<li>
<a href="https://pytorch.org/features">About</a>
</li>
<li>
<a href="https://pytorch.org/foundation">PyTorch Foundation</a>
</li>
<li>
<a href="https://pytorch.org/#community-module">Community</a>
</li>
<li>
<a href="https://pytorch.org/community-stories">Community Stories</a>
</li>
<li>
<a href="https://pytorch.org/resources">Developer Resources</a>
</li>
<li>
<a href="https://pytorch.org/events">Events</a>
</li>
<li>
<a href="https://discuss.pytorch.org/">Forums</a>
</li>
<li>
<a href="https://pytorch.org/hub">Models (Beta)</a>
</li>
</ul>
<li>
<a href="https://github.com/pytorch/pytorch">Github</a>
</li>
</ul>
</div>
</div>
</div>
<!-- End Mobile Menu -->
<script type="text/javascript" src="../_static/js/vendor/anchor.min.js"></script>
<script type="text/javascript">
$(document).ready(function() {
mobileMenu.bind();
mobileTOC.bind();
pytorchAnchors.bind();
sideMenus.bind();
scrollToAnchor.bind();
highlightNavigation.bind();
mainMenuDropdown.bind();
filterTags.bind();
// Add class to links that have code blocks, since we cannot create links in code blocks
$("article.pytorch-article a span.pre").each(function(e) {
$(this).closest("a").addClass("has-code");
});
})
</script>
</body>
</html>