forked from pytorch/pytorch.github.io
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbenchmark_utils.html
1063 lines (856 loc) · 61.1 KB
/
benchmark_utils.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>Benchmark Utils - torch.utils.benchmark — PyTorch 1.11.0 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/benchmark_utils.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/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/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.utils.bottleneck" href="bottleneck.html" />
<link rel="prev" title="torch.testing" href="testing.html" />
<!-- Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-117752657-2"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-117752657-2');
</script>
<!-- End Google Analytics -->
<script src="_static/js/modernizr.min.js"></script>
<!-- Preload the theme fonts -->
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-book.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-Medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium-italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-SemiBold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<!-- Preload the katex fonts -->
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Caligraphic-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.15.2/css/all.css" integrity="sha384-vSIIfh2YWi9wW0r9iZe7RJPrKwp6bG+s9QZMoITbCckVJqGCCRhc+ccxNcdpHuYu" crossorigin="anonymous">
</head>
<div class="container-fluid header-holder tutorials-header" id="header-holder">
<div class="container">
<div class="header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</li>
<li>
<a href="https://pytorch.org/mobile">Mobile</a>
</li>
<li>
<a href="https://pytorch.org/blog/">Blog</a>
</li>
<li>
<a href="https://pytorch.org/tutorials">Tutorials</a>
</li>
<li class="active docs-active">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-orange-arrow">
Docs
</a>
<div class="resources-dropdown-menu">
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/docs/stable/index.html">
<span class="dropdown-title">PyTorch</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/audio/stable/index.html">
<span class="dropdown-title">torchaudio</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/text/stable/index.html">
<span class="dropdown-title">torchtext</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/vision/stable/index.html">
<span class="dropdown-title">torchvision</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torchrec">
<span class="dropdown-title">TorchRec</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/serve/">
<span class="dropdown-title">TorchServe</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/#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/resources">
<span class="dropdown-title">Developer Resources</span>
<p>Find resources and get 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/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'>1.11.0 ▼</a>
</div>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="search.html" method="get">
<input type="text" name="q" placeholder="Search Docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div>
<p class="caption"><span class="caption-text">Notes</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="notes/amp_examples.html">Automatic Mixed Precision examples</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/autograd.html">Autograd mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/broadcasting.html">Broadcasting semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cpu_threading_torchscript_inference.html">CPU threading and TorchScript inference</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cuda.html">CUDA semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/ddp.html">Distributed Data Parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/extending.html">Extending PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/faq.html">Frequently Asked Questions</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/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/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"><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"><span class="caption-text">Python API</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="torch.html">torch</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.html">torch.nn</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.functional.html">torch.nn.functional</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensors.html">torch.Tensor</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_attributes.html">Tensor Attributes</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_view.html">Tensor Views</a></li>
<li class="toctree-l1"><a class="reference internal" href="autograd.html">torch.autograd</a></li>
<li class="toctree-l1"><a class="reference internal" href="cuda.html">torch.cuda</a></li>
<li class="toctree-l1"><a class="reference internal" href="amp.html">torch.cuda.amp</a></li>
<li class="toctree-l1"><a class="reference internal" href="backends.html">torch.backends</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.html">torch.distributed</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.algorithms.join.html">torch.distributed.algorithms.join</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.elastic.html">torch.distributed.elastic</a></li>
<li class="toctree-l1"><a class="reference internal" href="fsdp.html">torch.distributed.fsdp</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.optim.html">torch.distributed.optim</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributions.html">torch.distributions</a></li>
<li class="toctree-l1"><a class="reference internal" href="fft.html">torch.fft</a></li>
<li class="toctree-l1"><a class="reference internal" href="futures.html">torch.futures</a></li>
<li class="toctree-l1"><a class="reference internal" href="fx.html">torch.fx</a></li>
<li class="toctree-l1"><a class="reference internal" href="hub.html">torch.hub</a></li>
<li class="toctree-l1"><a class="reference internal" href="jit.html">torch.jit</a></li>
<li class="toctree-l1"><a class="reference internal" href="linalg.html">torch.linalg</a></li>
<li class="toctree-l1"><a class="reference internal" href="monitor.html">torch.monitor</a></li>
<li class="toctree-l1"><a class="reference internal" href="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="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 current"><a class="current reference internal" href="#">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="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__.html">torch.__config__</a></li>
</ul>
<p class="caption"><span class="caption-text">Libraries</span></p>
<ul>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/audio/stable">torchaudio</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/serve">TorchServe</a></li>
<li class="toctree-l1"><a class="reference external" href="http://pytorch.org/xla/">PyTorch on XLA Devices</a></li>
</ul>
<p class="caption"><span class="caption-text">Community</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="community/contribution_guide.html">PyTorch Contribution Guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/governance.html">PyTorch Governance</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/persons_of_interest.html">PyTorch Governance | Persons of Interest</a></li>
</ul>
</div>
</div>
</nav>
<div class="pytorch-container">
<div class="pytorch-page-level-bar" id="pytorch-page-level-bar">
<div class="pytorch-breadcrumbs-wrapper">
<div role="navigation" aria-label="breadcrumbs navigation">
<ul class="pytorch-breadcrumbs">
<li>
<a href="index.html">
Docs
</a> >
</li>
<li>Benchmark Utils - torch.utils.benchmark</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/benchmark_utils.rst.txt" rel="nofollow"><img src="_static/images/view-page-source-icon.svg"></a>
</li>
</ul>
</div>
</div>
<div class="pytorch-shortcuts-wrapper" id="pytorch-shortcuts-wrapper">
Shortcuts
</div>
</div>
<section data-toggle="wy-nav-shift" id="pytorch-content-wrap" class="pytorch-content-wrap">
<div class="pytorch-content-left">
<div class="rst-content">
<div role="main" class="main-content" itemscope="itemscope" itemtype="http://schema.org/Article">
<article itemprop="articleBody" id="pytorch-article" class="pytorch-article">
<div class="section" id="module-torch.utils.benchmark">
<span id="benchmark-utils-torch-utils-benchmark"></span><h1>Benchmark Utils - torch.utils.benchmark<a class="headerlink" href="#module-torch.utils.benchmark" title="Permalink to this headline">¶</a></h1>
<dl class="py class">
<dt id="torch.utils.benchmark.Timer">
<em class="property"><span class="pre">class</span> </em><code class="sig-prename descclassname"><span class="pre">torch.utils.benchmark.</span></code><code class="sig-name descname"><span class="pre">Timer</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="pre">stmt='pass'</span></em>, <em class="sig-param"><span class="pre">setup='pass'</span></em>, <em class="sig-param"><span class="pre">global_setup=''</span></em>, <em class="sig-param"><span class="pre">timer=<built-in</span> <span class="pre">function</span> <span class="pre">perf_counter></span></em>, <em class="sig-param"><span class="pre">globals=None</span></em>, <em class="sig-param"><span class="pre">label=None</span></em>, <em class="sig-param"><span class="pre">sub_label=None</span></em>, <em class="sig-param"><span class="pre">description=None</span></em>, <em class="sig-param"><span class="pre">env=None</span></em>, <em class="sig-param"><span class="pre">num_threads=1</span></em>, <em class="sig-param"><span class="pre">language=<Language.PYTHON:</span> <span class="pre">0></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/timer.html#Timer"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Timer" title="Permalink to this definition">¶</a></dt>
<dd><p>Helper class for measuring execution time of PyTorch statements.</p>
<p>For a full tutorial on how to use this class, see:
<a class="reference external" href="https://pytorch.org/tutorials/recipes/recipes/benchmark.html">https://pytorch.org/tutorials/recipes/recipes/benchmark.html</a></p>
<p>The PyTorch Timer is based on <cite>timeit.Timer</cite> (and in fact uses
<cite>timeit.Timer</cite> internally), but with several key differences:</p>
<ol class="arabic simple">
<li><dl class="simple">
<dt>Runtime aware:</dt><dd><p>Timer will perform warmups (important as some elements of PyTorch are
lazily initialized), set threadpool size so that comparisons are
apples-to-apples, and synchronize asynchronous CUDA functions when
necessary.</p>
</dd>
</dl>
</li>
<li><dl class="simple">
<dt>Focus on replicates:</dt><dd><p>When measuring code, and particularly complex kernels / models,
run-to-run variation is a significant confounding factor. It is
expected that all measurements should include replicates to quantify
noise and allow median computation, which is more robust than mean.
To that effect, this class deviates from the <cite>timeit</cite> API by
conceptually merging <cite>timeit.Timer.repeat</cite> and <cite>timeit.Timer.autorange</cite>.
(Exact algorithms are discussed in method docstrings.) The <cite>timeit</cite>
method is replicated for cases where an adaptive strategy is not
desired.</p>
</dd>
</dl>
</li>
<li><dl class="simple">
<dt>Optional metadata:</dt><dd><p>When defining a Timer, one can optionally specify <cite>label</cite>, <cite>sub_label</cite>,
<cite>description</cite>, and <cite>env</cite>. (Defined later) These fields are included in
the representation of result object and by the <cite>Compare</cite> class to group
and display results for comparison.</p>
</dd>
</dl>
</li>
<li><dl class="simple">
<dt>Instruction counts</dt><dd><p>In addition to wall times, Timer can run a statement under Callgrind
and report instructions executed.</p>
</dd>
</dl>
</li>
</ol>
<p>Directly analogous to <cite>timeit.Timer</cite> constructor arguments:</p>
<blockquote>
<div><p><cite>stmt</cite>, <cite>setup</cite>, <cite>timer</cite>, <cite>globals</cite></p>
</div></blockquote>
<p>PyTorch Timer specific constructor arguments:</p>
<blockquote>
<div><p><cite>label</cite>, <cite>sub_label</cite>, <cite>description</cite>, <cite>env</cite>, <cite>num_threads</cite></p>
</div></blockquote>
<dl class="field-list simple">
<dt class="field-odd">Parameters</dt>
<dd class="field-odd"><ul class="simple">
<li><p><strong>stmt</strong> – Code snippet to be run in a loop and timed.</p></li>
<li><p><strong>setup</strong> – Optional setup code. Used to define variables used in <cite>stmt</cite></p></li>
<li><p><strong>global_setup</strong> – (C++ only)
Code which is placed at the top level of the file for things like
<cite>#include</cite> statements.</p></li>
<li><p><strong>timer</strong> – Callable which returns the current time. If PyTorch was built
without CUDA or there is no GPU present, this defaults to
<cite>timeit.default_timer</cite>; otherwise it will synchronize CUDA before
measuring the time.</p></li>
<li><p><strong>globals</strong> – A dict which defines the global variables when <cite>stmt</cite> is being
executed. This is the other method for providing variables which
<cite>stmt</cite> needs.</p></li>
<li><p><strong>label</strong> – String which summarizes <cite>stmt</cite>. For instance, if <cite>stmt</cite> is
“torch.nn.functional.relu(torch.add(x, 1, out=out))”
one might set label to “ReLU(x + 1)” to improve readability.</p></li>
<li><p><strong>sub_label</strong> – <p>Provide supplemental information to disambiguate measurements
with identical stmt or label. For instance, in our example
above sub_label might be “float” or “int”, so that it is easy
to differentiate:
“ReLU(x + 1): (float)”</p>
<p>”ReLU(x + 1): (int)”
when printing Measurements or summarizing using <cite>Compare</cite>.</p>
</p></li>
<li><p><strong>description</strong> – <p>String to distinguish measurements with identical label and
sub_label. The principal use of <cite>description</cite> is to signal to
<cite>Compare</cite> the columns of data. For instance one might set it
based on the input size to create a table of the form:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span> <span class="o">|</span> <span class="n">n</span><span class="o">=</span><span class="mi">1</span> <span class="o">|</span> <span class="n">n</span><span class="o">=</span><span class="mi">4</span> <span class="o">|</span> <span class="o">...</span>
<span class="o">-------------</span> <span class="o">...</span>
<span class="n">ReLU</span><span class="p">(</span><span class="n">x</span> <span class="o">+</span> <span class="mi">1</span><span class="p">):</span> <span class="p">(</span><span class="nb">float</span><span class="p">)</span> <span class="o">|</span> <span class="o">...</span> <span class="o">|</span> <span class="o">...</span> <span class="o">|</span> <span class="o">...</span>
<span class="n">ReLU</span><span class="p">(</span><span class="n">x</span> <span class="o">+</span> <span class="mi">1</span><span class="p">):</span> <span class="p">(</span><span class="nb">int</span><span class="p">)</span> <span class="o">|</span> <span class="o">...</span> <span class="o">|</span> <span class="o">...</span> <span class="o">|</span> <span class="o">...</span>
</pre></div>
</div>
<p>using <cite>Compare</cite>. It is also included when printing a Measurement.</p>
</p></li>
<li><p><strong>env</strong> – This tag indicates that otherwise identical tasks were run in
different environments, and are therefore not equivilent, for
instance when A/B testing a change to a kernel. <cite>Compare</cite> will
treat Measurements with different <cite>env</cite> specification as distinct
when merging replicate runs.</p></li>
<li><p><strong>num_threads</strong> – The size of the PyTorch threadpool when executing <cite>stmt</cite>. Single
threaded performace is important as both a key inference workload
and a good indicator of intrinsic algorithmic efficiency, so the
default is set to one. This is in contrast to the default PyTorch
threadpool size which tries to utilize all cores.</p></li>
</ul>
</dd>
</dl>
<dl class="py method">
<dt id="torch.utils.benchmark.Timer.blocked_autorange">
<code class="sig-name descname"><span class="pre">blocked_autorange</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">callback</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">min_run_time</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.2</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/timer.html#Timer.blocked_autorange"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Timer.blocked_autorange" title="Permalink to this definition">¶</a></dt>
<dd><p>Measure many replicates while keeping timer overhead to a minimum.</p>
<p>At a high level, blocked_autorange executes the following pseudo-code:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span>`setup`
total_time = 0
while total_time < min_run_time
start = timer()
for _ in range(block_size):
`stmt`
total_time += (timer() - start)
</pre></div>
</div>
<p>Note the variable <cite>block_size</cite> in the inner loop. The choice of block
size is important to measurement quality, and must balance two
competing objectives:</p>
<blockquote>
<div><ol class="arabic simple">
<li><p>A small block size results in more replicates and generally
better statistics.</p></li>
<li><p>A large block size better amortizes the cost of <cite>timer</cite>
invocation, and results in a less biased measurement. This is
important because CUDA syncronization time is non-trivial
(order single to low double digit microseconds) and would
otherwise bias the measurement.</p></li>
</ol>
</div></blockquote>
<p>blocked_autorange sets block_size by running a warmup period,
increasing block size until timer overhead is less than 0.1% of
the overall computation. This value is then used for the main
measurement loop.</p>
<dl class="field-list simple">
<dt class="field-odd">Returns</dt>
<dd class="field-odd"><p>A <cite>Measurement</cite> object that contains measured runtimes and
repetition counts, and can be used to compute statistics.
(mean, median, etc.)</p>
</dd>
</dl>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.Timer.collect_callgrind">
<code class="sig-name descname"><span class="pre">collect_callgrind</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">number</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">100</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">*</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">repeats</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">collect_baseline</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">retain_out_file</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/utils/benchmark/utils/timer.html#Timer.collect_callgrind"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Timer.collect_callgrind" title="Permalink to this definition">¶</a></dt>
<dd><p>Collect instruction counts using Callgrind.</p>
<p>Unlike wall times, instruction counts are deterministic
(modulo non-determinism in the program itself and small amounts of
jitter from the Python interpreter.) This makes them ideal for detailed
performance analysis. This method runs <cite>stmt</cite> in a separate process
so that Valgrind can instrument the program. Performance is severely
degraded due to the instrumentation, however this is ameliorated by
the fact that a small number of iterations is generally sufficient to
obtain good measurements.</p>
<p>In order to to use this method <cite>valgrind</cite>, <cite>callgrind_control</cite>, and
<cite>callgrind_annotate</cite> must be installed.</p>
<p>Because there is a process boundary between the caller (this process)
and the <cite>stmt</cite> execution, <cite>globals</cite> cannot contain arbitrary in-memory
data structures. (Unlike timing methods) Instead, globals are
restricted to builtins, <cite>nn.Modules</cite>’s, and TorchScripted functions/modules
to reduce the surprise factor from serialization and subsequent
deserialization. The <cite>GlobalsBridge</cite> class provides more detail on this
subject. Take particular care with nn.Modules: they rely on pickle and
you may need to add an import to <cite>setup</cite> for them to transfer properly.</p>
<p>By default, a profile for an empty statement will be collected and
cached to indicate how many instructions are from the Python loop which
drives <cite>stmt</cite>.</p>
<dl class="field-list simple">
<dt class="field-odd">Returns</dt>
<dd class="field-odd"><p>A <cite>CallgrindStats</cite> object which provides instruction counts and
some basic facilities for analyzing and manipulating results.</p>
</dd>
</dl>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.Timer.timeit">
<code class="sig-name descname"><span class="pre">timeit</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">number</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1000000</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/timer.html#Timer.timeit"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Timer.timeit" title="Permalink to this definition">¶</a></dt>
<dd><p>Mirrors the semantics of timeit.Timer.timeit().</p>
<p>Execute the main statement (<cite>stmt</cite>) <cite>number</cite> times.
<a class="reference external" href="https://docs.python.org/3/library/timeit.html#timeit.Timer.timeit">https://docs.python.org/3/library/timeit.html#timeit.Timer.timeit</a></p>
</dd></dl>
</dd></dl>
<dl class="py class">
<dt id="torch.utils.benchmark.Measurement">
<em class="property"><span class="pre">class</span> </em><code class="sig-prename descclassname"><span class="pre">torch.utils.benchmark.</span></code><code class="sig-name descname"><span class="pre">Measurement</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">number_per_run</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">raw_times</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">task_spec</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">metadata</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/utils/benchmark/utils/common.html#Measurement"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Measurement" title="Permalink to this definition">¶</a></dt>
<dd><p>The result of a Timer measurement.</p>
<p>This class stores one or more measurements of a given statement. It is
serializable and provides several convenience methods
(including a detailed __repr__) for downstream consumers.</p>
<dl class="py method">
<dt id="torch.utils.benchmark.Measurement.merge">
<em class="property"><span class="pre">static</span> </em><code class="sig-name descname"><span class="pre">merge</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">measurements</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/common.html#Measurement.merge"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.Measurement.merge" title="Permalink to this definition">¶</a></dt>
<dd><p>Convenience method for merging replicates.</p>
<p>Merge will extrapolate times to <cite>number_per_run=1</cite> and will not
transfer any metadata. (Since it might differ between replicates)</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.Measurement.significant_figures">
<em class="property"><span class="pre">property</span> </em><code class="sig-name descname"><span class="pre">significant_figures</span></code><a class="headerlink" href="#torch.utils.benchmark.Measurement.significant_figures" title="Permalink to this definition">¶</a></dt>
<dd><p>Approximate significant figure estimate.</p>
<p>This property is intended to give a convenient way to estimate the
precision of a measurement. It only uses the interquartile region to
estimate statistics to try to mitigate skew from the tails, and
uses a static z value of 1.645 since it is not expected to be used
for small values of <cite>n</cite>, so z can approximate <cite>t</cite>.</p>
<p>The significant figure estimation used in conjunction with the
<cite>trim_sigfig</cite> method to provide a more human interpretable data
summary. __repr__ does not use this method; it simply displays raw
values. Significant figure estimation is intended for <cite>Compare</cite>.</p>
</dd></dl>
</dd></dl>
<dl class="py class">
<dt id="torch.utils.benchmark.CallgrindStats">
<em class="property"><span class="pre">class</span> </em><code class="sig-prename descclassname"><span class="pre">torch.utils.benchmark.</span></code><code class="sig-name descname"><span class="pre">CallgrindStats</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task_spec</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">number_per_run</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">built_with_debug_symbols</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">baseline_inclusive_stats</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">baseline_exclusive_stats</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stmt_inclusive_stats</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stmt_exclusive_stats</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">stmt_callgrind_out</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#CallgrindStats"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.CallgrindStats" title="Permalink to this definition">¶</a></dt>
<dd><p>Top level container for Callgrind results collected by Timer.</p>
<p>Manipulation is generally done using the FunctionCounts class, which is
obtained by calling <cite>CallgrindStats.stats(…)</cite>. Several convenience
methods are provided as well; the most significant is
<cite>CallgrindStats.as_standardized()</cite>.</p>
<dl class="py method">
<dt id="torch.utils.benchmark.CallgrindStats.as_standardized">
<code class="sig-name descname"><span class="pre">as_standardized</span></code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#CallgrindStats.as_standardized"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.CallgrindStats.as_standardized" title="Permalink to this definition">¶</a></dt>
<dd><p>Strip library names and some prefixes from function strings.</p>
<p>When comparing two different sets of instruction counts, on stumbling
block can be path prefixes. Callgrind includes the full filepath
when reporting a function (as it should). However, this can cause
issues when diffing profiles. If a key component such as Python
or PyTorch was built in separate locations in the two profiles, which
can result in something resembling:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="mi">23234231</span> <span class="o">/</span><span class="n">tmp</span><span class="o">/</span><span class="n">first_build_dir</span><span class="o">/</span><span class="n">thing</span><span class="o">.</span><span class="n">c</span><span class="p">:</span><span class="n">foo</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="mi">9823794</span> <span class="o">/</span><span class="n">tmp</span><span class="o">/</span><span class="n">first_build_dir</span><span class="o">/</span><span class="n">thing</span><span class="o">.</span><span class="n">c</span><span class="p">:</span><span class="n">bar</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="o">...</span>
<span class="mi">53453</span> <span class="o">.../</span><span class="n">aten</span><span class="o">/</span><span class="n">src</span><span class="o">/</span><span class="n">Aten</span><span class="o">/...</span><span class="p">:</span><span class="n">function_that_actually_changed</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="o">...</span>
<span class="o">-</span><span class="mi">9823794</span> <span class="o">/</span><span class="n">tmp</span><span class="o">/</span><span class="n">second_build_dir</span><span class="o">/</span><span class="n">thing</span><span class="o">.</span><span class="n">c</span><span class="p">:</span><span class="n">bar</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
<span class="o">-</span><span class="mi">23234231</span> <span class="o">/</span><span class="n">tmp</span><span class="o">/</span><span class="n">second_build_dir</span><span class="o">/</span><span class="n">thing</span><span class="o">.</span><span class="n">c</span><span class="p">:</span><span class="n">foo</span><span class="p">(</span><span class="o">...</span><span class="p">)</span>
</pre></div>
</div>
<p>Stripping prefixes can ameliorate this issue by regularizing the
strings and causing better cancellation of equivilent call sites
when diffing.</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.CallgrindStats.counts">
<code class="sig-name descname"><span class="pre">counts</span></code><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">denoise</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/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#CallgrindStats.counts"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.CallgrindStats.counts" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns the total number of instructions executed.</p>
<p>See <cite>FunctionCounts.denoise()</cite> for an explation of the <cite>denoise</cite> arg.</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.CallgrindStats.delta">
<code class="sig-name descname"><span class="pre">delta</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">other</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">inclusive</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/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#CallgrindStats.delta"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.CallgrindStats.delta" title="Permalink to this definition">¶</a></dt>
<dd><p>Diff two sets of counts.</p>
<p>One common reason to collect instruction counts is to determine the
the effect that a particular change will have on the number of instructions
needed to perform some unit of work. If a change increases that number, the
next logical question is “why”. This generally involves looking at what part
if the code increased in instruction count. This function automates that
process so that one can easily diff counts on both an inclusive and
exclusive basis.</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.CallgrindStats.stats">
<code class="sig-name descname"><span class="pre">stats</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">inclusive</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/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#CallgrindStats.stats"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.CallgrindStats.stats" title="Permalink to this definition">¶</a></dt>
<dd><p>Returns detailed function counts.</p>
<p>Conceptually, the FunctionCounts returned can be thought of as a tuple
of (count, path_and_function_name) tuples.</p>
<p><cite>inclusive</cite> matches the semantics of callgrind. If True, the counts
include instructions executed by children. <cite>inclusive=True</cite> is useful
for identifying hot spots in code; <cite>inclusive=False</cite> is useful for
reducing noise when diffing counts from two different runs. (See
CallgrindStats.delta(…) for more details)</p>
</dd></dl>
</dd></dl>
<dl class="py class">
<dt id="torch.utils.benchmark.FunctionCounts">
<em class="property"><span class="pre">class</span> </em><code class="sig-prename descclassname"><span class="pre">torch.utils.benchmark.</span></code><code class="sig-name descname"><span class="pre">FunctionCounts</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">_data</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">inclusive</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">truncate_rows</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">_linewidth</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/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#FunctionCounts"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.FunctionCounts" title="Permalink to this definition">¶</a></dt>
<dd><p>Container for manipulating Callgrind results.</p>
<dl class="simple">
<dt>It supports:</dt><dd><ol class="arabic simple">
<li><p>Addition and subtraction to combine or diff results.</p></li>
<li><p>Tuple-like indexing.</p></li>
<li><p>A <cite>denoise</cite> function which strips CPython calls which are known to
be non-deterministic and quite noisy.</p></li>
<li><p>Two higher order methods (<cite>filter</cite> and <cite>transform</cite>) for custom
manipulation.</p></li>
</ol>
</dd>
</dl>
<dl class="py method">
<dt id="torch.utils.benchmark.FunctionCounts.denoise">
<code class="sig-name descname"><span class="pre">denoise</span></code><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#FunctionCounts.denoise"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.FunctionCounts.denoise" title="Permalink to this definition">¶</a></dt>
<dd><p>Remove known noisy instructions.</p>
<p>Several instructions in the CPython interpreter are rather noisy. These
instructions involve unicode to dictionary lookups which Python uses to
map variable names. FunctionCounts is generally a content agnostic
container, however this is sufficiently important for obtaining
reliable results to warrant an exception.</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.FunctionCounts.filter">
<code class="sig-name descname"><span class="pre">filter</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">filter_fn</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#FunctionCounts.filter"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.FunctionCounts.filter" title="Permalink to this definition">¶</a></dt>
<dd><p>Keep only the elements where <cite>filter_fn</cite> applied to function name returns True.</p>
</dd></dl>
<dl class="py method">
<dt id="torch.utils.benchmark.FunctionCounts.transform">
<code class="sig-name descname"><span class="pre">transform</span></code><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">map_fn</span></span></em><span class="sig-paren">)</span><a class="reference internal" href="_modules/torch/utils/benchmark/utils/valgrind_wrapper/timer_interface.html#FunctionCounts.transform"><span class="viewcode-link"><span class="pre">[source]</span></span></a><a class="headerlink" href="#torch.utils.benchmark.FunctionCounts.transform" title="Permalink to this definition">¶</a></dt>
<dd><p>Apply <cite>map_fn</cite> to all of the function names.</p>
<p>This can be used to regularize function names (e.g. stripping irrelevant
parts of the file path), coalesce entries by mapping multiple functions
to the same name (in which case the counts are added together), etc.</p>
</dd></dl>
</dd></dl>
</div>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="bottleneck.html" class="btn btn-neutral float-right" title="torch.utils.bottleneck" accesskey="n" rel="next">Next <img src="_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="testing.html" class="btn btn-neutral" title="torch.testing" accesskey="p" rel="prev"><img src="_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr>
<div role="contentinfo">
<p>
© Copyright 2019, Torch Contributors.
</p>
</div>
<div>
Built with <a href="http://sphinx-doc.org/">Sphinx</a> using a <a href="https://github.com/rtfd/sphinx_rtd_theme">theme</a> provided by <a href="https://readthedocs.org">Read the Docs</a>.
</div>
</footer>
</div>
</div>
<div class="pytorch-content-right" id="pytorch-content-right">
<div class="pytorch-right-menu" id="pytorch-right-menu">
<div class="pytorch-side-scroll" id="pytorch-side-scroll-right">
<ul>
<li><a class="reference internal" href="#">Benchmark Utils - torch.utils.benchmark</a></li>
</ul>
</div>
</div>
</div>
</section>
</div>
<script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></script>
<script src="_static/jquery.js"></script>
<script src="_static/underscore.js"></script>
<script src="_static/doctools.js"></script>
<script src="_static/clipboard.min.js"></script>
<script src="_static/copybutton.js"></script>
<script type="text/javascript" src="_static/js/vendor/popper.min.js"></script>
<script type="text/javascript" src="_static/js/vendor/bootstrap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/list.js/1.5.0/list.min.js"></script>
<script type="text/javascript" src="_static/js/theme.js"></script>
<script type="text/javascript">
jQuery(function () {
SphinxRtdTheme.Navigation.enable(true);
});
</script>
<script script type="text/javascript">
var collapsedSections = ['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 follow-us-col">
<ul>
<li class="list-title">Stay Connected</li>
<li>
<div id="mc_embed_signup">
<form
action="https://twitter.us14.list-manage.com/subscribe/post?u=75419c71fe0a935e53dfa4a3f&id=91d0dccd39"
method="post"
id="mc-embedded-subscribe-form"
name="mc-embedded-subscribe-form"
class="email-subscribe-form validate"
target="_blank"
novalidate>
<div id="mc_embed_signup_scroll" class="email-subscribe-form-fields-wrapper">
<div class="mc-field-group">
<label for="mce-EMAIL" style="display:none;">Email Address</label>
<input type="email" value="" name="EMAIL" class="required email" id="mce-EMAIL" placeholder="Email Address">
</div>
<div id="mce-responses" class="clear">
<div class="response" id="mce-error-response" style="display:none"></div>
<div class="response" id="mce-success-response" style="display:none"></div>
</div> <!-- real people should not fill this in and expect good things - do not remove this or risk form bot signups-->
<div style="position: absolute; left: -5000px;" aria-hidden="true"><input type="text" name="b_75419c71fe0a935e53dfa4a3f_91d0dccd39" tabindex="-1" value=""></div>
<div class="clear">
<input type="submit" value="" name="subscribe" id="mc-embedded-subscribe" class="button email-subscribe-button">
</div>
</div>
</form>
</div>
</li>
</ul>
<div class="footer-social-icons">
<a href="https://www.facebook.com/pytorch" target="_blank" class="facebook"></a>
<a href="https://twitter.com/pytorch" target="_blank" class="twitter"></a>
<a href="https://www.youtube.com/pytorch" target="_blank" class="youtube"></a>
</div>
</div>
</div>
</div>
</footer>
<div class="cookie-banner-wrapper">
<div class="container">
<p class="gdpr-notice">To analyze traffic and optimize your experience, we serve cookies on this site. By clicking or navigating, you agree to allow our usage of cookies. As the current maintainers of this site, Facebook’s Cookies Policy applies. Learn more, including about available controls: <a href="https://www.facebook.com/policies/cookies/">Cookies Policy</a>.</p>
<img class="close-button" src="_static/images/pytorch-x.svg">
</div>
</div>
<!-- End Footer -->
<!-- Begin Mobile Menu -->
<div class="mobile-main-menu">
<div class="container-fluid">
<div class="container">
<div class="mobile-main-menu-header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<a class="main-menu-close-button" href="#" data-behavior="close-mobile-menu"></a>
</div>
</div>
</div>
<div class="mobile-main-menu-links-container">
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</li>
<li>
<a href="https://pytorch.org/mobile">Mobile</a>
</li>
<li>
<a href="https://pytorch.org/hub">PyTorch Hub</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/elastic/">TorchElastic</a>
</li>
<li>
<a href="https://pytorch.org/serve/">TorchServe</a>
</li>