-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathprofiler.html
1318 lines (1114 loc) · 93.6 KB
/
profiler.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta name="robots" content="noindex">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>torch.profiler — PyTorch 2.5 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/profiler.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="torch.nn.init" href="nn.init.html" />
<link rel="prev" title="torch.package" href="package.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/[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 class="main-menu-item">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Learn
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/get-started">
<span class=dropdown-title>Get Started</span>
<p>Run PyTorch locally or get started quickly with one of the supported cloud platforms</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials">
<span class="dropdown-title">Tutorials</span>
<p>Whats new in PyTorch tutorials</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/beginner/basics/intro.html">
<span class="dropdown-title">Learn the Basics</span>
<p>Familiarize yourself with PyTorch concepts and modules</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/recipes/recipes_index.html">
<span class="dropdown-title">PyTorch Recipes</span>
<p>Bite-size, ready-to-deploy PyTorch code examples</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/tutorials/beginner/introyt.html">
<span class="dropdown-title">Intro to PyTorch - YouTube Series</span>
<p>Master PyTorch basics with our engaging YouTube tutorial series</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Ecosystem
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/ecosystem">
<span class="dropdown-title">Tools</span>
<p>Learn about the tools and frameworks in the PyTorch Ecosystem</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://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/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/ecosystem/contributor-awards-2023">
<span class="dropdown-title">Contributor Awards - 2023</span>
<p>Award winners announced at this year's PyTorch Conference</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="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>
<p>Build innovative and privacy-aware AI experiences for edge devices</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/executorch-overview">
<span class="dropdown-title">ExecuTorch</span>
<p>End-to-end solution for enabling on-device inference capabilities across mobile and edge devices</p>
</a>
</div>
</div>
</li>
<li class="main-menu-item">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Docs
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/docs/stable/index.html">
<span class="dropdown-title">PyTorch</span>
<p>Explore the documentation for comprehensive guidance on how to use PyTorch</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/pytorch-domains">
<span class="dropdown-title">PyTorch Domains</span>
<p>Read the PyTorch Domains documentation to learn more about domain-specific libraries</p>
</a>
</div>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
Blogs & News
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/blog/">
<span class="dropdown-title">PyTorch Blog</span>
<p>Catch up on the latest technical news and happenings</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/community-blog">
<span class="dropdown-title">Community Blog</span>
<p>Stories from the PyTorch ecosystem</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/videos">
<span class="dropdown-title">Videos</span>
<p>Learn about the latest PyTorch tutorials, new, and more </p>
<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/events">
<span class="dropdown-title">Events</span>
<p>Find events, webinars, and podcasts</p>
</a>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="with-down-arrow">
About
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/foundation">
<span class="dropdown-title">PyTorch Foundation</span>
<p>Learn more about the PyTorch Foundation</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/governing-board">
<span class="dropdown-title">Governing Board</span>
<p></p>
</a>
</div>
</div>
</li>
<li class="main-menu-item">
<div class="no-dropdown">
<a href="https://pytorch.org/join" data-cta="join">
Become a Member
</a>
</div>
</li>
<li>
<div class="main-menu-item">
<a href="https://github.com/pytorch/pytorch" class="github-icon">
</a>
</div>
</li>
<!--- TODO: This block adds the search icon to the nav bar. We will enable it later.
<li>
<div class="main-menu-item">
<a href="https://github.com/pytorch/pytorch" class="search-icon">
</a>
</div>
</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'>2.5 ▼</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">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/custom_operators.html">PyTorch Custom Operators Landing Page</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/fsdp.html">FSDP Notes</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/get_start_xpu.html">Getting Started on Intel GPU</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"><a class="reference internal" href="torch_cuda_memory.html">Understanding CUDA Memory Usage</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_cuda_memory.html#generating-a-snapshot">Generating a Snapshot</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_cuda_memory.html#using-the-visualizer">Using the visualizer</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_cuda_memory.html#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="xpu.html">torch.xpu</a></li>
<li class="toctree-l1"><a class="reference internal" href="mtia.html">torch.mtia</a></li>
<li class="toctree-l1"><a class="reference internal" href="meta.html">Meta device</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.tensor.html">torch.distributed.tensor</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.tensor.parallel.html">torch.distributed.tensor.parallel</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.pipelining.html">torch.distributed.pipelining</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="fx.experimental.html">torch.fx.experimental</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 current"><a class="current reference internal" href="#">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="nn.attention.html">torch.nn.attention</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="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="size.html">torch.Size</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="module_tracker.html">torch.utils.module_tracker</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="future_mod.html">torch.__future__</a></li>
<li class="toctree-l1"><a class="reference internal" href="logging.html">torch._logging</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_environment_variables.html">Torch Environment Variables</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>torch.profiler</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/profiler.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="torch-profiler">
<h1>torch.profiler<a class="headerlink" href="#torch-profiler" title="Permalink to this heading">¶</a></h1>
<div class="section" id="module-torch.profiler">
<span id="overview"></span><h2>Overview<a class="headerlink" href="#module-torch.profiler" title="Permalink to this heading">¶</a></h2>
<p>PyTorch Profiler is a tool that allows the collection of performance metrics during training and inference.
Profiler’s context manager API can be used to better understand what model operators are the most expensive,
examine their input shapes and stack traces, study device kernel activity and visualize the execution trace.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>An earlier version of the API in <a class="reference internal" href="autograd.html#module-torch.autograd" title="torch.autograd"><code class="xref py py-mod docutils literal notranslate"><span class="pre">torch.autograd</span></code></a> module is considered legacy and will be deprecated.</p>
</div>
</div>
<div class="section" id="api-reference">
<h2>API Reference<a class="headerlink" href="#api-reference" title="Permalink to this heading">¶</a></h2>
<dl class="py class">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile">
<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">_KinetoProfile</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">activities</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">record_shapes</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">profile_memory</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_stack</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_flops</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_modules</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">experimental_config</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">execution_trace_observer</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">acc_events</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile" title="Permalink to this definition">¶</a></dt>
<dd><p>Low-level profiler wrap the autograd profile</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>activities</strong> (<em>iterable</em>) – list of activity groups (CPU, CUDA) to use in profiling, supported values:
<code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CPU</span></code>, <code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CUDA</span></code>,
<code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.XPU</span></code>.
Default value: ProfilerActivity.CPU and (when available) ProfilerActivity.CUDA
or (when available) ProfilerActivity.XPU.</p></li>
<li><p><strong>record_shapes</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – save information about operator’s input shapes.</p></li>
<li><p><strong>profile_memory</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – track tensor memory allocation/deallocation (see <code class="docutils literal notranslate"><span class="pre">export_memory_timeline</span></code>
for more details).</p></li>
<li><p><strong>with_stack</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – record source information (file and line number) for the ops.</p></li>
<li><p><strong>with_flops</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – use formula to estimate the FLOPS of specific operators
(matrix multiplication and 2D convolution).</p></li>
<li><p><strong>with_modules</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – record module hierarchy (including function names)
corresponding to the callstack of the op. e.g. If module A’s forward call’s
module B’s forward which contains an aten::add op,
then aten::add’s module hierarchy is A.B
Note that this support exist, at the moment, only for TorchScript models
and not eager mode models.</p></li>
<li><p><strong>experimental_config</strong> (<em>_ExperimentalConfig</em>) – A set of experimental options
used by profiler libraries like Kineto. Note, backward compatibility is not guaranteed.</p></li>
<li><p><strong>execution_trace_observer</strong> (<em>ExecutionTraceObserver</em>) – A PyTorch Execution Trace Observer object.
<a class="reference external" href="https://arxiv.org/pdf/2305.14516.pdf">PyTorch Execution Traces</a> offer a graph based
representation of AI/ML workloads and enable replay benchmarks, simulators, and emulators.
When this argument is included the observer start() and stop() will be called for the
same time window as PyTorch profiler.</p></li>
<li><p><strong>acc_events</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – Enable the accumulation of FunctionEvents across multiple profiling cycles</p></li>
</ul>
</dd>
</dl>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>This API is experimental and subject to change in the future.</p>
<p>Enabling shape and stack tracing results in additional overhead.
When record_shapes=True is specified, profiler will temporarily hold references to the tensors;
that may further prevent certain optimizations that depend on the reference count and introduce
extra tensor copies.</p>
</div>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.add_metadata">
<span class="sig-name descname"><span class="pre">add_metadata</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">key</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.add_metadata"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.add_metadata" title="Permalink to this definition">¶</a></dt>
<dd><p>Adds a user defined metadata with a string key and a string value
into the trace file</p>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.add_metadata_json">
<span class="sig-name descname"><span class="pre">add_metadata_json</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">key</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.add_metadata_json"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.add_metadata_json" title="Permalink to this definition">¶</a></dt>
<dd><p>Adds a user defined metadata with a string key and a valid json value
into the trace file</p>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.events">
<span class="sig-name descname"><span class="pre">events</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.events"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.events" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns the list of unaggregated profiler events,
to be used in the trace callback or after the profiling is finished</p>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.export_chrome_trace">
<span class="sig-name descname"><span class="pre">export_chrome_trace</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">path</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.export_chrome_trace"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.export_chrome_trace" title="Permalink to this definition">¶</a></dt>
<dd><p>Exports the collected trace in Chrome JSON format. If kineto is enabled, only
last cycle in schedule is exported.</p>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.export_memory_timeline">
<span class="sig-name descname"><span class="pre">export_memory_timeline</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">path</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/profiler/profiler.html#_KinetoProfile.export_memory_timeline"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.export_memory_timeline" title="Permalink to this definition">¶</a></dt>
<dd><p>Export memory event information from the profiler collected
tree for a given device, and export a timeline plot. There are 3
exportable files using <code class="docutils literal notranslate"><span class="pre">export_memory_timeline</span></code>, each controlled by the
<code class="docutils literal notranslate"><span class="pre">path</span></code>’s suffix.</p>
<ul class="simple">
<li><p>For an HTML compatible plot, use the suffix <code class="docutils literal notranslate"><span class="pre">.html</span></code>, and a memory timeline
plot will be embedded as a PNG file in the HTML file.</p></li>
<li><p>For plot points consisting of <code class="docutils literal notranslate"><span class="pre">[times,</span> <span class="pre">[sizes</span> <span class="pre">by</span> <span class="pre">category]]</span></code>, where
<code class="docutils literal notranslate"><span class="pre">times</span></code> are timestamps and <code class="docutils literal notranslate"><span class="pre">sizes</span></code> are memory usage for each category.
The memory timeline plot will be saved a JSON (<code class="docutils literal notranslate"><span class="pre">.json</span></code>) or gzipped JSON
(<code class="docutils literal notranslate"><span class="pre">.json.gz</span></code>) depending on the suffix.</p></li>
<li><p>For raw memory points, use the suffix <code class="docutils literal notranslate"><span class="pre">.raw.json.gz</span></code>. Each raw memory
event will consist of <code class="docutils literal notranslate"><span class="pre">(timestamp,</span> <span class="pre">action,</span> <span class="pre">numbytes,</span> <span class="pre">category)</span></code>, where
<code class="docutils literal notranslate"><span class="pre">action</span></code> is one of <code class="docutils literal notranslate"><span class="pre">[PREEXISTING,</span> <span class="pre">CREATE,</span> <span class="pre">INCREMENT_VERSION,</span> <span class="pre">DESTROY]</span></code>,
and <code class="docutils literal notranslate"><span class="pre">category</span></code> is one of the enums from
<code class="docutils literal notranslate"><span class="pre">torch.profiler._memory_profiler.Category</span></code>.</p></li>
</ul>
<p>Output: Memory timeline written as gzipped JSON, JSON, or HTML.</p>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.export_stacks">
<span class="sig-name descname"><span class="pre">export_stacks</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">path</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metric</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'self_cpu_time_total'</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.export_stacks"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.export_stacks" title="Permalink to this definition">¶</a></dt>
<dd><p>Save stack traces to a file</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>path</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.13)"><em>str</em></a>) – save stacks file to this location;</p></li>
<li><p><strong>metric</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.13)"><em>str</em></a>) – metric to use: “self_cpu_time_total” or “self_cuda_time_total”</p></li>
</ul>
</dd>
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.key_averages">
<span class="sig-name descname"><span class="pre">key_averages</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">group_by_input_shape</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">group_by_stack_n</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.key_averages"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.key_averages" title="Permalink to this definition">¶</a></dt>
<dd><p>Averages events, grouping them by operator name and (optionally) input shapes and
stack.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>To use shape/stack functionality make sure to set record_shapes/with_stack
when creating profiler context manager.</p>
</div>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.preset_metadata_json">
<span class="sig-name descname"><span class="pre">preset_metadata_json</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">key</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.preset_metadata_json"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.preset_metadata_json" title="Permalink to this definition">¶</a></dt>
<dd><p>Preset a user defined metadata when the profiler is not started
and added into the trace file later.
Metadata is in the format of a string key and a valid json value</p>
<dl class="field-list simple">
</dl>
</dd></dl>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler._KinetoProfile.toggle_collection_dynamic">
<span class="sig-name descname"><span class="pre">toggle_collection_dynamic</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">enable</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">activities</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#_KinetoProfile.toggle_collection_dynamic"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler._KinetoProfile.toggle_collection_dynamic" title="Permalink to this definition">¶</a></dt>
<dd><p>Toggle collection of activities on/off at any point of collection. Currently supports toggling Torch Ops
(CPU) and CUDA activity supported in Kineto</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>activities</strong> (<em>iterable</em>) – list of activity groups to use in profiling, supported values:
<code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CPU</span></code>, <code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CUDA</span></code></p>
</dd>
</dl>
<p>Examples:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">profile</span><span class="p">(</span>
<span class="n">activities</span><span class="o">=</span><span class="p">[</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CPU</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CUDA</span><span class="p">,</span>
<span class="p">]</span>
<span class="p">)</span> <span class="k">as</span> <span class="n">p</span><span class="p">:</span>
<span class="n">code_to_profile_0</span><span class="p">()</span>
<span class="o">//</span> <span class="n">turn</span> <span class="n">off</span> <span class="n">collection</span> <span class="n">of</span> <span class="nb">all</span> <span class="n">CUDA</span> <span class="n">activity</span>
<span class="n">p</span><span class="o">.</span><span class="n">toggle_collection_dynamic</span><span class="p">(</span><span class="kc">False</span><span class="p">,</span> <span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CUDA</span><span class="p">])</span>
<span class="n">code_to_profile_1</span><span class="p">()</span>
<span class="o">//</span> <span class="n">turn</span> <span class="n">on</span> <span class="n">collection</span> <span class="n">of</span> <span class="nb">all</span> <span class="n">CUDA</span> <span class="n">activity</span>
<span class="n">p</span><span class="o">.</span><span class="n">toggle_collection_dynamic</span><span class="p">(</span><span class="kc">True</span><span class="p">,</span> <span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CUDA</span><span class="p">])</span>
<span class="n">code_to_profile_2</span><span class="p">()</span>
<span class="nb">print</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">key_averages</span><span class="p">()</span><span class="o">.</span><span class="n">table</span><span class="p">(</span>
<span class="n">sort_by</span><span class="o">=</span><span class="s2">"self_cuda_time_total"</span><span class="p">,</span> <span class="n">row_limit</span><span class="o">=-</span><span class="mi">1</span><span class="p">))</span>
</pre></div>
</div>
</dd></dl>
</dd></dl>
<dl class="py class">
<dt class="sig sig-object py" id="torch.profiler.profile">
<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">profile</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">activities</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">schedule</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">on_trace_ready</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">record_shapes</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">profile_memory</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_stack</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_flops</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">with_modules</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">experimental_config</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">execution_trace_observer</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">acc_events</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">use_cuda</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/profiler/profiler.html#profile"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.profile" title="Permalink to this definition">¶</a></dt>
<dd><p>Profiler context manager.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>activities</strong> (<em>iterable</em>) – list of activity groups (CPU, CUDA) to use in profiling, supported values:
<code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CPU</span></code>, <code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.CUDA</span></code>,
<code class="docutils literal notranslate"><span class="pre">torch.profiler.ProfilerActivity.XPU</span></code>.
Default value: ProfilerActivity.CPU and (when available) ProfilerActivity.CUDA
or (when available) ProfilerActivity.XPU.</p></li>
<li><p><strong>schedule</strong> (<em>Callable</em>) – callable that takes step (int) as a single parameter and returns
<code class="docutils literal notranslate"><span class="pre">ProfilerAction</span></code> value that specifies the profiler action to perform at each step.</p></li>
<li><p><strong>on_trace_ready</strong> (<em>Callable</em>) – callable that is called at each step when <code class="docutils literal notranslate"><span class="pre">schedule</span></code>
returns <code class="docutils literal notranslate"><span class="pre">ProfilerAction.RECORD_AND_SAVE</span></code> during the profiling.</p></li>
<li><p><strong>record_shapes</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – save information about operator’s input shapes.</p></li>
<li><p><strong>profile_memory</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – track tensor memory allocation/deallocation.</p></li>
<li><p><strong>with_stack</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – record source information (file and line number) for the ops.</p></li>
<li><p><strong>with_flops</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – use formula to estimate the FLOPs (floating point operations) of specific operators
(matrix multiplication and 2D convolution).</p></li>
<li><p><strong>with_modules</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – record module hierarchy (including function names)
corresponding to the callstack of the op. e.g. If module A’s forward call’s
module B’s forward which contains an aten::add op,
then aten::add’s module hierarchy is A.B
Note that this support exist, at the moment, only for TorchScript models
and not eager mode models.</p></li>
<li><p><strong>experimental_config</strong> (<em>_ExperimentalConfig</em>) – A set of experimental options
used for Kineto library features. Note, backward compatibility is not guaranteed.</p></li>
<li><p><strong>execution_trace_observer</strong> (<em>ExecutionTraceObserver</em>) – A PyTorch Execution Trace Observer object.
<a class="reference external" href="https://arxiv.org/pdf/2305.14516.pdf">PyTorch Execution Traces</a> offer a graph based
representation of AI/ML workloads and enable replay benchmarks, simulators, and emulators.
When this argument is included the observer start() and stop() will be called for the
same time window as PyTorch profiler. See the examples section below for a code sample.</p></li>
<li><p><strong>acc_events</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – Enable the accumulation of FunctionEvents across multiple profiling cycles</p></li>
<li><p><strong>use_cuda</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.13)"><em>bool</em></a>) – <div class="deprecated">
<p><span class="versionmodified deprecated">Deprecated since version 1.8.1: </span>use <code class="docutils literal notranslate"><span class="pre">activities</span></code> instead.</p>
</div>
</p></li>
</ul>
</dd>
</dl>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Use <a class="reference internal" href="#torch.profiler.schedule" title="torch.profiler.schedule"><code class="xref py py-func docutils literal notranslate"><span class="pre">schedule()</span></code></a> to generate the callable schedule.
Non-default schedules are useful when profiling long training jobs
and allow the user to obtain multiple traces at the different iterations
of the training process.
The default schedule simply records all the events continuously for the
duration of the context manager.</p>
</div>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Use <a class="reference internal" href="#torch.profiler.tensorboard_trace_handler" title="torch.profiler.tensorboard_trace_handler"><code class="xref py py-func docutils literal notranslate"><span class="pre">tensorboard_trace_handler()</span></code></a> to generate result files for TensorBoard:</p>
<p><code class="docutils literal notranslate"><span class="pre">on_trace_ready=torch.profiler.tensorboard_trace_handler(dir_name)</span></code></p>
<p>After profiling, result files can be found in the specified directory. Use the command:</p>
<p><code class="docutils literal notranslate"><span class="pre">tensorboard</span> <span class="pre">--logdir</span> <span class="pre">dir_name</span></code></p>
<p>to see the results in TensorBoard.
For more information, see
<a class="reference external" href="https://github.com/pytorch/kineto/tree/master/tb_plugin">PyTorch Profiler TensorBoard Plugin</a></p>
</div>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>Enabling shape and stack tracing results in additional overhead.
When record_shapes=True is specified, profiler will temporarily hold references to the tensors;
that may further prevent certain optimizations that depend on the reference count and introduce
extra tensor copies.</p>
</div>
<p>Examples:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">profile</span><span class="p">(</span>
<span class="n">activities</span><span class="o">=</span><span class="p">[</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CPU</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CUDA</span><span class="p">,</span>
<span class="p">]</span>
<span class="p">)</span> <span class="k">as</span> <span class="n">p</span><span class="p">:</span>
<span class="n">code_to_profile</span><span class="p">()</span>
<span class="nb">print</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">key_averages</span><span class="p">()</span><span class="o">.</span><span class="n">table</span><span class="p">(</span>
<span class="n">sort_by</span><span class="o">=</span><span class="s2">"self_cuda_time_total"</span><span class="p">,</span> <span class="n">row_limit</span><span class="o">=-</span><span class="mi">1</span><span class="p">))</span>
</pre></div>
</div>
<p>Using the profiler’s <code class="docutils literal notranslate"><span class="pre">schedule</span></code>, <code class="docutils literal notranslate"><span class="pre">on_trace_ready</span></code> and <code class="docutils literal notranslate"><span class="pre">step</span></code> functions:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="c1"># Non-default profiler schedule allows user to turn profiler on and off</span>
<span class="c1"># on different iterations of the training loop;</span>
<span class="c1"># trace_handler is called every time a new trace becomes available</span>
<span class="k">def</span> <span class="nf">trace_handler</span><span class="p">(</span><span class="n">prof</span><span class="p">):</span>
<span class="nb">print</span><span class="p">(</span><span class="n">prof</span><span class="o">.</span><span class="n">key_averages</span><span class="p">()</span><span class="o">.</span><span class="n">table</span><span class="p">(</span>
<span class="n">sort_by</span><span class="o">=</span><span class="s2">"self_cuda_time_total"</span><span class="p">,</span> <span class="n">row_limit</span><span class="o">=-</span><span class="mi">1</span><span class="p">))</span>
<span class="c1"># prof.export_chrome_trace("/tmp/test_trace_" + str(prof.step_num) + ".json")</span>
<span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">profile</span><span class="p">(</span>
<span class="n">activities</span><span class="o">=</span><span class="p">[</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CPU</span><span class="p">,</span>
<span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">ProfilerActivity</span><span class="o">.</span><span class="n">CUDA</span><span class="p">,</span>
<span class="p">],</span>
<span class="c1"># In this example with wait=1, warmup=1, active=2, repeat=1,</span>
<span class="c1"># profiler will skip the first step/iteration,</span>
<span class="c1"># start warming up on the second, record</span>
<span class="c1"># the third and the forth iterations,</span>
<span class="c1"># after which the trace will become available</span>
<span class="c1"># and on_trace_ready (when set) is called;</span>
<span class="c1"># the cycle repeats starting with the next step</span>
<span class="n">schedule</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">schedule</span><span class="p">(</span>
<span class="n">wait</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="n">warmup</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="n">active</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span>
<span class="n">repeat</span><span class="o">=</span><span class="mi">1</span><span class="p">),</span>
<span class="n">on_trace_ready</span><span class="o">=</span><span class="n">trace_handler</span>
<span class="c1"># on_trace_ready=torch.profiler.tensorboard_trace_handler('./log')</span>
<span class="c1"># used when outputting for tensorboard</span>
<span class="p">)</span> <span class="k">as</span> <span class="n">p</span><span class="p">:</span>
<span class="k">for</span> <span class="nb">iter</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">N</span><span class="p">):</span>
<span class="n">code_iteration_to_profile</span><span class="p">(</span><span class="nb">iter</span><span class="p">)</span>
<span class="c1"># send a signal to the profiler that the next iteration has started</span>
<span class="n">p</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
</pre></div>
</div>
<p>The following sample shows how to setup up an Execution Trace Observer (<cite>execution_trace_observer</cite>)</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">profile</span><span class="p">(</span>
<span class="o">...</span>
<span class="n">execution_trace_observer</span><span class="o">=</span><span class="p">(</span>
<span class="n">ExecutionTraceObserver</span><span class="p">()</span><span class="o">.</span><span class="n">register_callback</span><span class="p">(</span><span class="s2">"./execution_trace.json"</span><span class="p">)</span>
<span class="p">),</span>
<span class="p">)</span> <span class="k">as</span> <span class="n">p</span><span class="p">:</span>
<span class="k">for</span> <span class="nb">iter</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">N</span><span class="p">):</span>
<span class="n">code_iteration_to_profile</span><span class="p">(</span><span class="nb">iter</span><span class="p">)</span>
<span class="n">p</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
</pre></div>
</div>
<p>You can also refer to test_execution_trace_with_kineto() in tests/profiler/test_profiler.py.
Note: One can also pass any object satisfying the _ITraceObserver interface.</p>
<dl class="py method">
<dt class="sig sig-object py" id="torch.profiler.profile.step">
<span class="sig-name descname"><span class="pre">step</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#profile.step"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.profile.step" title="Permalink to this definition">¶</a></dt>
<dd><p>Signals the profiler that the next profiling step has started.</p>
</dd></dl>
</dd></dl>
<dl class="py class">
<dt class="sig sig-object py" id="torch.profiler.ProfilerAction">
<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">ProfilerAction</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">value</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#ProfilerAction"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.ProfilerAction" title="Permalink to this definition">¶</a></dt>
<dd><p>Profiler actions that can be taken at the specified intervals</p>
</dd></dl>
<dl class="py class">
<dt class="sig sig-object py" id="torch.profiler.ProfilerActivity">
<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">ProfilerActivity</span></span><a class="headerlink" href="#torch.profiler.ProfilerActivity" title="Permalink to this definition">¶</a></dt>
<dd><p>Members:</p>
<p>CPU</p>
<p>XPU</p>
<p>MTIA</p>
<p>CUDA</p>
<p>PrivateUse1</p>
<dl class="py property">
<dt class="sig sig-object py" id="torch.profiler.ProfilerActivity.name">
<em class="property"><span class="pre">property</span><span class="w"> </span></em><span class="sig-name descname"><span class="pre">name</span></span><a class="headerlink" href="#torch.profiler.ProfilerActivity.name" title="Permalink to this definition">¶</a></dt>
<dd></dd></dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.schedule">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">wait</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">warmup</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">active</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">repeat</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">skip_first</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#schedule"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.schedule" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns a callable that can be used as profiler <code class="docutils literal notranslate"><span class="pre">schedule</span></code> argument. The profiler will skip
the first <code class="docutils literal notranslate"><span class="pre">skip_first</span></code> steps, then wait for <code class="docutils literal notranslate"><span class="pre">wait</span></code> steps, then do the warmup for the next <code class="docutils literal notranslate"><span class="pre">warmup</span></code> steps,
then do the active recording for the next <code class="docutils literal notranslate"><span class="pre">active</span></code> steps and then repeat the cycle starting with <code class="docutils literal notranslate"><span class="pre">wait</span></code> steps.
The optional number of cycles is specified with the <code class="docutils literal notranslate"><span class="pre">repeat</span></code> parameter, the zero value means that
the cycles will continue until the profiling is finished.</p>
<dl class="field-list simple">
<dt class="field-odd">Return type</dt>
<dd class="field-odd"><p><a class="reference external" href="https://docs.python.org/3/library/typing.html#typing.Callable" title="(in Python v3.13)"><em>Callable</em></a></p>
</dd>
</dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.tensorboard_trace_handler">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.</span></span><span class="sig-name descname"><span class="pre">tensorboard_trace_handler</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">dir_name</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">worker_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">use_gzip</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/profiler.html#tensorboard_trace_handler"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.tensorboard_trace_handler" title="Permalink to this definition">¶</a></dt>
<dd><p>Outputs tracing files to directory of <code class="docutils literal notranslate"><span class="pre">dir_name</span></code>, then that directory can be
directly delivered to tensorboard as logdir.
<code class="docutils literal notranslate"><span class="pre">worker_name</span></code> should be unique for each worker in distributed scenario,
it will be set to ‘[hostname]_[pid]’ by default.</p>
<dl class="field-list simple">
</dl>
</dd></dl>
</div>
<div class="section" id="intel-instrumentation-and-tracing-technology-apis">
<h2>Intel Instrumentation and Tracing Technology APIs<a class="headerlink" href="#intel-instrumentation-and-tracing-technology-apis" title="Permalink to this heading">¶</a></h2>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.itt.is_available">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.itt.</span></span><span class="sig-name descname"><span class="pre">is_available</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/itt.html#is_available"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.itt.is_available" title="Permalink to this definition">¶</a></dt>
<dd><p>Check if ITT feature is available or not</p>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.itt.mark">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.itt.</span></span><span class="sig-name descname"><span class="pre">mark</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">msg</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/itt.html#mark"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.itt.mark" title="Permalink to this definition">¶</a></dt>
<dd><p>Describe an instantaneous event that occurred at some point.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>msg</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.13)"><em>str</em></a>) – ASCII message to associate with the event.</p>
</dd>
</dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.itt.range_push">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.itt.</span></span><span class="sig-name descname"><span class="pre">range_push</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">msg</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/itt.html#range_push"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.itt.range_push" title="Permalink to this definition">¶</a></dt>
<dd><p>Pushes a range onto a stack of nested range span. Returns zero-based
depth of the range that is started.</p>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><p><strong>msg</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.13)"><em>str</em></a>) – ASCII message to associate with range</p>
</dd>
</dl>
</dd></dl>
<dl class="py function">
<dt class="sig sig-object py" id="torch.profiler.itt.range_pop">
<span class="sig-prename descclassname"><span class="pre">torch.profiler.itt.</span></span><span class="sig-name descname"><span class="pre">range_pop</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/profiler/itt.html#range_pop"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.profiler.itt.range_pop" title="Permalink to this definition">¶</a></dt>
<dd><p>Pops a range off of a stack of nested range spans. Returns the
zero-based depth of the range that is ended.</p>
</dd></dl>
<span class="target" id="module-torch.profiler.itt"></span><span class="target" id="module-torch.profiler.profiler"></span><span class="target" id="module-torch.profiler.python_tracer"></span></div>
</div>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="nn.init.html" class="btn btn-neutral float-right" title="torch.nn.init" accesskey="n" rel="next">Next <img src="_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="package.html" class="btn btn-neutral" title="torch.package" 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="#">torch.profiler</a><ul>
<li><a class="reference internal" href="#module-torch.profiler">Overview</a></li>
<li><a class="reference internal" href="#api-reference">API Reference</a><ul>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile</span></code></a><ul>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.add_metadata"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.add_metadata()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.add_metadata_json"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.add_metadata_json()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.events"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.events()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.export_chrome_trace"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.export_chrome_trace()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.export_memory_timeline"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.export_memory_timeline()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.export_stacks"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.export_stacks()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.key_averages"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.key_averages()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.preset_metadata_json"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.preset_metadata_json()</span></code></a></li>
<li><a class="reference internal" href="#torch.profiler._KinetoProfile.toggle_collection_dynamic"><code class="docutils literal notranslate"><span class="pre">_KinetoProfile.toggle_collection_dynamic()</span></code></a></li>
</ul>
</li>