-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtorch_cuda_memory.html
1002 lines (800 loc) · 56.6 KB
/
torch_cuda_memory.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 charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Understanding CUDA Memory Usage — PyTorch 2.2 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/torch_cuda_memory.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/katex@0.10.0-beta/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.11/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="torch.mps" href="mps.html" />
<link rel="prev" title="CUDA Stream Sanitizer" href="cuda._sanitizer.html" />
<!-- Google Tag Manager -->
<script>(function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':
new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],
j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src=
'https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);
})(window,document,'script','dataLayer','GTM-T8XT4PS');</script>
<!-- End Google Tag Manager -->
<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/katex@0.10.0/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/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>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-arrow">
Edge
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/edge">
<span class="dropdown-title">About PyTorch Edge</span>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/executorch">
<span class="dropdown-title">ExecuTorch</span>
</a>
</div>
</div>
</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'>2.2 ▼</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" 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/extending.func.html">Extending torch.func with autograd.Function</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">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 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="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="cpu.html">torch.cpu</a></li>
<li class="toctree-l1"><a class="reference internal" href="cuda.html">torch.cuda</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">Understanding CUDA Memory Usage</a></li>
<li class="toctree-l1"><a class="reference internal" href="#generating-a-snapshot">Generating a Snapshot</a></li>
<li class="toctree-l1"><a class="reference internal" href="#using-the-visualizer">Using the visualizer</a></li>
<li class="toctree-l1"><a class="reference internal" href="#snapshot-api-reference">Snapshot API Reference</a></li>
<li class="toctree-l1"><a class="reference internal" href="mps.html">torch.mps</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="export.html">torch.export</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="torch.compiler.html">torch.compiler</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="func.html">torch.func</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="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="utils.html">torch.utils</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="deterministic.html">torch.utils.deterministic</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>
<li class="toctree-l1"><a class="reference internal" href="logging.html">torch._logging</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="https://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>Understanding CUDA Memory Usage</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/torch_cuda_memory.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">
<!-- Google Tag Manager (noscript) -->
<noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-T8XT4PS"
height="0" width="0" style="display:none;visibility:hidden"></iframe></noscript>
<!-- End Google Tag Manager (noscript) -->
<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="understanding-cuda-memory-usage">
<span id="torch-cuda-memory"></span><h1>Understanding CUDA Memory Usage<a class="headerlink" href="#understanding-cuda-memory-usage" title="Permalink to this heading">¶</a></h1>
<p>To debug CUDA memory use, PyTorch provides a way to generate memory snapshots that record the state of allocated CUDA memory
at any point in time, and optionally record the history of allocation events that led up to that snapshot.</p>
<p>The generated snapshots can then be drag and dropped onto the interactiver viewer hosted at <a class="reference external" href="https://pytorch.org/memory_viz">pytorch.org/memory_viz</a> which
can be used to explore the snapshot.</p>
</div>
<div class="section" id="generating-a-snapshot">
<h1>Generating a Snapshot<a class="headerlink" href="#generating-a-snapshot" title="Permalink to this heading">¶</a></h1>
<p>The common pattern for recording a snapshot is to enable memory history, run the code to be observed, and then save a file with a pickled snapshot:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># enable memory history, which will</span>
<span class="c1"># add tracebacks and event history to snapshots</span>
<span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">memory</span><span class="o">.</span><span class="n">_record_memory_history</span><span class="p">()</span>
<span class="n">run_your_code</span><span class="p">()</span>
<span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">memory</span><span class="o">.</span><span class="n">_dump_snapshot</span><span class="p">(</span><span class="s2">"my_snapshot.pickle"</span><span class="p">)</span>
</pre></div>
</div>
</div>
<div class="section" id="using-the-visualizer">
<h1>Using the visualizer<a class="headerlink" href="#using-the-visualizer" title="Permalink to this heading">¶</a></h1>
<p>Open <a class="reference external" href="https://pytorch.org/memory_viz">pytorch.org/memory_viz</a> and drag/drop the pickled snapshot file into the visualizer.
The visualizer is a javascript application that runs locally on your computer. It does not upload any snapshot data.</p>
<div class="section" id="active-memory-timeline">
<h2>Active Memory Timeline<a class="headerlink" href="#active-memory-timeline" title="Permalink to this heading">¶</a></h2>
<p>The Active Memory Timeline shows all the live tensors over time in the snapshot on a particular GPU. Pan/Zoom over the plot to look at smaller allocations.
Mouse over allocated blocks to see a stack trace for when that block was allocated, and details like its address. The detail slider can be adjusted to
render fewer allocations and improve performance when there is a lot of data.</p>
<img alt="_images/active_memory_timeline.png" src="_images/active_memory_timeline.png" />
</div>
<div class="section" id="allocator-state-history">
<h2>Allocator State History<a class="headerlink" href="#allocator-state-history" title="Permalink to this heading">¶</a></h2>
<p>The Allocator State History shows individual allocator events in a timeline on the left. Select an event in the timeline to see a visual summary of the
allocator state at that event. This summary shows each individual segment returned from cudaMalloc and how it is split up into blocks of individual allocations
or free space. Mouse over segments and blocks to see the stack trace when the memory was allocated. Mouse over events to see the stack trace when the event occurred,
such as when a tensor was freed. Out of memory errors are reported as OOM events. Looking at the state of memory during an OOM may provide insight into why
an allocation failed even though reserved memory still exists.</p>
<img alt="_images/allocator_state_history.png" src="_images/allocator_state_history.png" />
<p>The stack trace information also reports the address at which an allocation occurred.
The address b7f064c000000_0 refers to the (b)lock at address 7f064c000000 which is the “_0”th time this address was allocated.
This unique string can be looked up in the Active Memory Timeline and searched
in the Active State History to examine the memory state when a tensor was allocated or freed.</p>
</div>
</div>
<div class="section" id="snapshot-api-reference">
<h1>Snapshot API Reference<a class="headerlink" href="#snapshot-api-reference" title="Permalink to this heading">¶</a></h1>
<dl class="py function">
<dt class="sig sig-object py" id="torch.cuda.memory._record_memory_history">
<span class="sig-prename descclassname"><span class="pre">torch.cuda.memory.</span></span><span class="sig-name descname"><span class="pre">_record_memory_history</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">enabled</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'all'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">context</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'all'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stacks</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'all'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">max_entries</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">9223372036854775807</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/cuda/memory.html#_record_memory_history"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.cuda.memory._record_memory_history" title="Permalink to this definition">¶</a></dt>
<dd><p>Enable recording of stack traces associated with memory
allocations, so you can tell what allocated any piece of memory in
<a class="reference internal" href="#torch.cuda.memory._snapshot" title="torch.cuda.memory._snapshot"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.cuda.memory._snapshot()</span></code></a>.</p>
<p>In addition too keeping stack traces with each current allocation and free,
this will also enable recording of a history of all alloc/free events.</p>
<p>Use <a class="reference internal" href="#torch.cuda.memory._snapshot" title="torch.cuda.memory._snapshot"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.cuda.memory._snapshot()</span></code></a> to retrieve this information,
and the tools in <cite>_memory_viz.py</cite> to visualize snapshots.</p>
<p>The Python trace collection is fast (2us per trace), so you may consider
enabling this on production jobs if you anticipate ever having to debug
memory issues.</p>
<p>C++ trace collection is also fast (~50ns/frame), which for many typical programs
works out to ~2us per trace, but can vary depending on stack depth.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>enabled</strong> (<em>Literal</em><em>[</em><em>None</em><em>, </em><em>"state"</em><em>, </em><em>"all"</em><em>]</em><em>, </em><em>optional</em>) – <cite>None</cite>, disable recording memory history.
<cite>“state”</cite>, keep information for currenly allocated memory.
<cite>“all”</cite>, additionally keep a history of all alloc/free calls.
Defaults to “all”.</p></li>
<li><p><strong>context</strong> (<em>Literal</em><em>[</em><em>None</em><em>, </em><em>"state"</em><em>, </em><em>"alloc"</em><em>, </em><em>"all"</em><em>]</em><em>, </em><em>optional</em>) – <cite>None</cite>, Do not record any tracebacks.
<cite>“state”</cite>, Record tracebacks for currently allocated memory.
<cite>“alloc”</cite>, additionally keep tracebacks for alloc calls.
<cite>“all”</cite>, additionally keep tracebacks for free calls.
Defaults to “all”.</p></li>
<li><p><strong>stacks</strong> (<em>Literal</em><em>[</em><em>"python"</em><em>, </em><em>"all"</em><em>]</em><em>, </em><em>optional</em>) – <cite>“python”</cite>, include Python, TorchScript, and inductor frames in tracebacks
<cite>“all”</cite>, additionally include C++ frames
Defaults to “all”.</p></li>
<li><p><strong>max_entries</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#int" title="(in Python v3.12)"><em>int</em></a><em>, </em><em>optional</em>) – Keep a maximum of <cite>max_entries</cite>
alloc/free events in the recorded history recorded.</p></li>
</ul>
</dd>
</dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.cuda.memory._snapshot">
<span class="sig-prename descclassname"><span class="pre">torch.cuda.memory.</span></span><span class="sig-name descname"><span class="pre">_snapshot</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/cuda/memory.html#_snapshot"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.cuda.memory._snapshot" title="Permalink to this definition">¶</a></dt>
<dd><p>Save a snapshot of CUDA memory state at the time it was called.</p>
<p>The state is represented as a dictionary with the following structure.</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">Snapshot</span><span class="p">(</span><span class="n">TypedDict</span><span class="p">):</span>
<span class="n">segments</span> <span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">Segment</span><span class="p">]</span>
<span class="n">device_traces</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">List</span><span class="p">[</span><span class="n">TraceEntry</span><span class="p">]]</span>
<span class="k">class</span> <span class="nc">Segment</span><span class="p">(</span><span class="n">TypedDict</span><span class="p">):</span>
<span class="c1"># Segments are memory returned from a cudaMalloc call.</span>
<span class="c1"># The size of reserved memory is the sum of all Segments.</span>
<span class="c1"># Segments are cached and reused for future allocations.</span>
<span class="c1"># If the reuse is smaller than the segment, the segment</span>
<span class="c1"># is split into more then one Block.</span>
<span class="c1"># empty_cache() frees Segments that are entirely inactive.</span>
<span class="n">address</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">total_size</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># cudaMalloc'd size of segment</span>
<span class="n">stream</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">segment_type</span><span class="p">:</span> <span class="n">Literal</span><span class="p">[</span><span class="s1">'small'</span><span class="p">,</span> <span class="s1">'large'</span><span class="p">]</span> <span class="c1"># 'large' (>1MB)</span>
<span class="n">allocated_size</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># size of memory in use</span>
<span class="n">active_size</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># size of memory in use or in active_awaiting_free state</span>
<span class="n">blocks</span> <span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">Block</span><span class="p">]</span>
<span class="k">class</span> <span class="nc">Block</span><span class="p">(</span><span class="n">TypedDict</span><span class="p">):</span>
<span class="c1"># A piece of memory returned from the allocator, or</span>
<span class="c1"># current cached but inactive.</span>
<span class="n">size</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">requested_size</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># size requested during malloc, may be smaller than</span>
<span class="c1"># size due to rounding</span>
<span class="n">address</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">state</span><span class="p">:</span> <span class="n">Literal</span><span class="p">[</span><span class="s1">'active_allocated'</span><span class="p">,</span> <span class="c1"># used by a tensor</span>
<span class="s1">'active_awaiting_free'</span><span class="p">,</span> <span class="c1"># waiting for another stream to finish using</span>
<span class="c1"># this, then it will become free</span>
<span class="s1">'inactive'</span><span class="p">,]</span> <span class="c1"># free for reuse</span>
<span class="n">frames</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">Frame</span><span class="p">]</span> <span class="c1"># stack trace from where the allocation occurred</span>
<span class="k">class</span> <span class="nc">Frame</span><span class="p">(</span><span class="n">TypedDict</span><span class="p">):</span>
<span class="n">filename</span><span class="p">:</span> <span class="nb">str</span>
<span class="n">line</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">name</span><span class="p">:</span> <span class="nb">str</span>
<span class="k">class</span> <span class="nc">TraceEntry</span><span class="p">(</span><span class="n">TypedDict</span><span class="p">):</span>
<span class="c1"># When `torch.cuda.memory._record_memory_history()` is enabled,</span>
<span class="c1"># the snapshot will contain TraceEntry objects that record each</span>
<span class="c1"># action the allocator took.</span>
<span class="n">action</span><span class="p">:</span> <span class="n">Literal</span><span class="p">[</span>
<span class="s1">'alloc'</span> <span class="c1"># memory allocated</span>
<span class="s1">'free_requested'</span><span class="p">,</span> <span class="c1"># the allocated received a call to free memory</span>
<span class="s1">'free_completed'</span><span class="p">,</span> <span class="c1"># the memory that was requested to be freed is now</span>
<span class="c1"># able to be used in future allocation calls</span>
<span class="s1">'segment_alloc'</span><span class="p">,</span> <span class="c1"># the caching allocator ask cudaMalloc for more memory</span>
<span class="c1"># and added it as a segment in its cache</span>
<span class="s1">'segment_free'</span><span class="p">,</span> <span class="c1"># the caching allocator called cudaFree to return memory</span>
<span class="c1"># to cuda possibly trying free up memory to</span>
<span class="c1"># allocate more segments or because empty_caches was called</span>
<span class="s1">'oom'</span><span class="p">,</span> <span class="c1"># the allocator threw an OOM exception. 'size' is</span>
<span class="c1"># the requested number of bytes that did not succeed</span>
<span class="s1">'snapshot'</span> <span class="c1"># the allocator generated a memory snapshot</span>
<span class="c1"># useful to coorelate a previously taken</span>
<span class="c1"># snapshot with this trace</span>
<span class="p">]</span>
<span class="n">addr</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># not present for OOM</span>
<span class="n">frames</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">Frame</span><span class="p">]</span>
<span class="n">size</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">stream</span><span class="p">:</span> <span class="nb">int</span>
<span class="n">device_free</span><span class="p">:</span> <span class="nb">int</span> <span class="c1"># only present for OOM, the amount of</span>
<span class="c1"># memory cuda still reports to be free</span>
</pre></div>
</div>
<dl class="field-list simple">
<dt class="field-odd">Returns</dt>
<dd class="field-odd"><p>The Snapshot dictionary object</p>
</dd>
</dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.cuda.memory._dump_snapshot">
<span class="sig-prename descclassname"><span class="pre">torch.cuda.memory.</span></span><span class="sig-name descname"><span class="pre">_dump_snapshot</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">filename</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'dump_snapshot.pickle'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/cuda/memory.html#_dump_snapshot"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.cuda.memory._dump_snapshot" title="Permalink to this definition">¶</a></dt>
<dd><p>Save a pickled version of the <cite>torch.memory._snapshot()</cite> dictionary to a file.</p>
<p>This file can be opened by the interactive snapshot viewer at pytorch.org/memory_viz</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>filename</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.12)"><em>str</em></a><em>, </em><em>optional</em>) – Name of the file to create. Defaults to “dump_snapshot.pickle”.</p>
</dd>
</dl>
</dd></dl>
</div>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="mps.html" class="btn btn-neutral float-right" title="torch.mps" accesskey="n" rel="next">Next <img src="_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="cuda._sanitizer.html" class="btn btn-neutral" title="CUDA Stream Sanitizer" accesskey="p" rel="prev"><img src="_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr>
<div role="contentinfo">
<p>
© Copyright 2023, 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>
<script>
var match = window.location.href.match(/\/_[a-zA-Z0-9_]*.html|_dynamo/gi);
var url = window.location.href.lastIndexOf(match[match.length-1]);
if (url)
{
var div = '<div class="admonition note"><p class="admonition-title">Note</p><p><i class="fa fa-exclamation-circle" aria-hidden="true"> </i> This page describes an internal API which is not intended to be used outside of the PyTorch codebase and can be modified or removed without notice.</p></div>'
document.getElementById("pytorch-article").insertAdjacentHTML('afterBegin', div)
}
</script>
</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="#">Understanding CUDA Memory Usage</a></li>
<li><a class="reference internal" href="#generating-a-snapshot">Generating a Snapshot</a></li>
<li><a class="reference internal" href="#using-the-visualizer">Using the visualizer</a><ul>
<li><a class="reference internal" href="#active-memory-timeline">Active Memory Timeline</a></li>
<li><a class="reference internal" href="#allocator-state-history">Allocator State History</a></li>
</ul>
</li>
<li><a class="reference internal" href="#snapshot-api-reference">Snapshot API Reference</a><ul>
<li><a class="reference internal" href="#torch.cuda.memory._record_memory_history"><code class="docutils literal notranslate"><span class="pre">_record_memory_history()</span></code></a></li>
<li><a class="reference internal" href="#torch.cuda.memory._snapshot"><code class="docutils literal notranslate"><span class="pre">_snapshot()</span></code></a></li>
<li><a class="reference internal" href="#torch.cuda.memory._dump_snapshot"><code class="docutils literal notranslate"><span class="pre">_dump_snapshot()</span></code></a></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/sphinx_highlight.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="https://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="https://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="">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>