-
Notifications
You must be signed in to change notification settings - Fork 257
/
Copy pathtensorboard_profiler_tutorial.html
1158 lines (987 loc) ยท 93.5 KB
/
tensorboard_profiler_tutorial.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="ko" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="ko" > <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta property="og:title" content="ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ" />
<meta property="og:type" content="article" />
<meta property="og:url" content="https://tutorials.pytorch.kr/intermediate/tensorboard_profiler_tutorial.html" />
<meta property="og:site_name" content="PyTorch Tutorials KR" />
<meta property="og:description" content="๋ฒ์ญ: ์๋์ฐ ์ด ํํ ๋ฆฌ์ผ์์๋ ํ์ดํ ์น(PyTorch) ํ๋กํ์ผ๋ฌ(profiler)์ ํจ๊ป ํ
์๋ณด๋(TensorBoard) ํ๋ฌ๊ทธ์ธ(plugin)์ ์ฌ์ฉํ์ฌ ๋ชจ๋ธ์ ์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ํ์งํ๋ ๋ฐฉ๋ฒ์ ๋ณด์ฌ ์ค๋๋ค. ์๊ฐ: ํ์ดํ ์น(PyTorch) 1.8๋ถํฐ GPU์์ CUDA ์ปค๋(kernel) ์คํ ๋ฟ๋ง ์๋๋ผ CPU ์์
์ ๊ธฐ๋กํ ์ ์๋ ์
๋ฐ์ดํธ๋ ํ๋กํ์ผ๋ฌ API๊ฐ ํฌํจ๋์ด ์์ต๋๋ค. ํ๋กํ์ผ๋ฌ๋ ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์์ ์ด๋ฐ ์ ๋ณด๋ฅผ ์๊ฐํํ๊ณ ์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ๋ํ ๋ถ์์ ์ ๊ณตํ ์ ์์ต๋๋ค. ์ด ํํ ๋ฆฌ์ผ์์๋ ๊ฐ๋จํ R..." />
<meta property="og:image" content="https://tutorials.pytorch.kr/_static/logos/logo-kr-sm-dark.png" />
<meta property="og:image:alt" content="PyTorch Tutorials KR" />
<meta name="description" content="๋ฒ์ญ: ์๋์ฐ ์ด ํํ ๋ฆฌ์ผ์์๋ ํ์ดํ ์น(PyTorch) ํ๋กํ์ผ๋ฌ(profiler)์ ํจ๊ป ํ
์๋ณด๋(TensorBoard) ํ๋ฌ๊ทธ์ธ(plugin)์ ์ฌ์ฉํ์ฌ ๋ชจ๋ธ์ ์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ํ์งํ๋ ๋ฐฉ๋ฒ์ ๋ณด์ฌ ์ค๋๋ค. ์๊ฐ: ํ์ดํ ์น(PyTorch) 1.8๋ถํฐ GPU์์ CUDA ์ปค๋(kernel) ์คํ ๋ฟ๋ง ์๋๋ผ CPU ์์
์ ๊ธฐ๋กํ ์ ์๋ ์
๋ฐ์ดํธ๋ ํ๋กํ์ผ๋ฌ API๊ฐ ํฌํจ๋์ด ์์ต๋๋ค. ํ๋กํ์ผ๋ฌ๋ ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์์ ์ด๋ฐ ์ ๋ณด๋ฅผ ์๊ฐํํ๊ณ ์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ๋ํ ๋ถ์์ ์ ๊ณตํ ์ ์์ต๋๋ค. ์ด ํํ ๋ฆฌ์ผ์์๋ ๊ฐ๋จํ R..." />
<meta property="og:ignore_canonical" content="true" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ — ํ์ดํ ์น ํ๊ตญ์ด ํํ ๋ฆฌ์ผ (PyTorch tutorials in Korean)</title>
<link rel="shortcut icon" href="../_static/favicon.ico"/>
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="../_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="../_static/copybutton.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.16.10/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="../_static/katex-math.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-binder.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-dataframe.css" type="text/css" />
<link rel="stylesheet" href="../_static/sg_gallery-rendered-html.css" type="text/css" />
<link rel="stylesheet" href="../_static/sphinx-design.5ea377869091fd0449014c60fc090103.min.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.10.0-beta/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/custom.css" type="text/css" />
<link rel="stylesheet" href="../_static/css/custom2.css" type="text/css" />
<link rel="index" title="์์ธ" href="../genindex.html" />
<link rel="search" title="๊ฒ์" href="../search.html" />
<link rel="next" title="Ray Tune์ ์ฌ์ฉํ ํ์ดํผํ๋ผ๋ฏธํฐ ํ๋" href="../beginner/hyperparameter_tuning_tutorial.html" />
<link rel="prev" title="Facilitating New Backend Integration by PrivateUse1" href="../advanced/privateuseone.html" />
<script src="../_static/js/modernizr.min.js"></script>
<!-- Preload the theme fonts -->
<link rel="preload" href="../_static/fonts/FreightSans/freight-sans-book.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="../_static/fonts/FreightSans/freight-sans-medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="../_static/fonts/IBMPlexMono/IBMPlexMono-Medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="../_static/fonts/FreightSans/freight-sans-bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="../_static/fonts/FreightSans/freight-sans-medium-italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="../_static/fonts/IBMPlexMono/IBMPlexMono-SemiBold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<!-- Preload the katex fonts -->
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/katex@0.10.0/dist/fonts/KaTeX_Caligraphic-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.15.2/css/all.css" integrity="sha384-vSIIfh2YWi9wW0r9iZe7RJPrKwp6bG+s9QZMoITbCckVJqGCCRhc+ccxNcdpHuYu" crossorigin="anonymous">
</head>
<div class="container-fluid header-holder tutorials-header" id="header-holder">
<div class="container">
<div class="header-container">
<a class="header-logo" href="https://pytorch.kr/" aria-label="PyTorch"></a>
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.kr/get-started">์์ํ๊ธฐ</a>
</li>
<li class="active">
<a href="https://tutorials.pytorch.kr/">ํํ ๋ฆฌ์ผ</a>
</li>
<li>
<a href="https://pytorch.kr/hub">ํ๋ธ</a>
</li>
<li>
<a href="https://discuss.pytorch.kr/">์ปค๋ฎค๋ํฐ</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">
2.3.1+cu121
</div>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="../search.html" method="get">
<input type="text" name="q" placeholder="Search Tutorials" />
<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">ํ์ดํ ์น(PyTorch) ๋ ์ํผ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../recipes/recipes_index.html">๋ชจ๋ ๋ ์ํผ ๋ณด๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../prototype/prototype_index.html">๋ชจ๋ ํ๋กํ ํ์
๋ ์ํผ ๋ณด๊ธฐ</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">ํ์ดํ ์น(PyTorch) ์์ํ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/intro.html">ํ์ดํ ์น(PyTorch) ๊ธฐ๋ณธ ์ตํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/quickstart_tutorial.html">๋น ๋ฅธ ์์(Quickstart)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/tensorqs_tutorial.html">ํ
์(Tensor)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/data_tutorial.html">Dataset๊ณผ DataLoader</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/transforms_tutorial.html">๋ณํ(Transform)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/buildmodel_tutorial.html">์ ๊ฒฝ๋ง ๋ชจ๋ธ ๊ตฌ์ฑํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/autogradqs_tutorial.html"><code class="docutils literal notranslate"><span class="pre">torch.autograd</span></code>๋ฅผ ์ฌ์ฉํ ์๋ ๋ฏธ๋ถ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/optimization_tutorial.html">๋ชจ๋ธ ๋งค๊ฐ๋ณ์ ์ต์ ํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/basics/saveloadrun_tutorial.html">๋ชจ๋ธ ์ ์ฅํ๊ณ ๋ถ๋ฌ์ค๊ธฐ</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Introduction to PyTorch on YouTube</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt.html">PyTorch ์๊ฐ - YouTube ์๋ฆฌ์ฆ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/introyt1_tutorial.html">PyTorch ์๊ฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/tensors_deeper_tutorial.html">Pytorch Tensor ์๊ฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/autogradyt_tutorial.html">The Fundamentals of Autograd</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/modelsyt_tutorial.html">Building Models with PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/tensorboardyt_tutorial.html">PyTorch TensorBoard Support</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/trainingyt.html">Training with PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/captumyt.html">Model Understanding with Captum</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">ํ์ดํ ์น(PyTorch) ๋ฐฐ์ฐ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/deep_learning_60min_blitz.html">PyTorch๋ก ๋ฅ๋ฌ๋ํ๊ธฐ: 60๋ถ๋ง์ ๋์ฅ๋ด๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/pytorch_with_examples.html">์์ ๋ก ๋ฐฐ์ฐ๋ ํ์ดํ ์น(PyTorch)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/nn_tutorial.html"><cite>torch.nn</cite> ์ด <em>์ค์ ๋ก</em> ๋ฌด์์ธ๊ฐ์?</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensorboard_tutorial.html">TensorBoard๋ก ๋ชจ๋ธ, ๋ฐ์ดํฐ, ํ์ต ์๊ฐํํ๊ธฐ</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">์ด๋ฏธ์ง/๋น๋์ค</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="torchvision_tutorial.html">TorchVision Object Detection Finetuning Tutorial</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/transfer_learning_tutorial.html">์ปดํจํฐ ๋น์ (Vision)์ ์ํ ์ ์ดํ์ต(Transfer Learning)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/fgsm_tutorial.html">์ ๋์ ์์ ์์ฑ(Adversarial Example Generation)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/dcgan_faces_tutorial.html">DCGAN ํํ ๋ฆฌ์ผ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/vt_tutorial.html">๋ฐฐํฌ๋ฅผ ์ํด ๋น์ ํธ๋์คํฌ๋จธ(Vision Transformer) ๋ชจ๋ธ ์ต์ ํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="tiatoolbox_tutorial.html">Whole Slide Image Classification Using PyTorch and TIAToolbox</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">์ค๋์ค</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_io_tutorial.html">Audio I/O</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_resampling_tutorial.html">Audio Resampling</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_data_augmentation_tutorial.html">Audio Data Augmentation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_feature_extractions_tutorial.html">Audio Feature Extractions</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_feature_augmentation_tutorial.html">Audio Feature Augmentation</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_datasets_tutorial.html">Audio Datasets</a></li>
<li class="toctree-l1"><a class="reference internal" href="speech_recognition_pipeline_tutorial.html">Speech Recognition with Wav2Vec2</a></li>
<li class="toctree-l1"><a class="reference internal" href="text_to_speech_with_torchaudio.html">Text-to-speech with Tacotron2</a></li>
<li class="toctree-l1"><a class="reference internal" href="forced_alignment_with_torchaudio_tutorial.html">wav2vec2์ ์ด์ฉํ ๊ฐ์ ์ ๋ ฌ</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">ํ
์คํธ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/bettertransformer_tutorial.html">Fast Transformer Inference with Better Transformer</a></li>
<li class="toctree-l1"><a class="reference internal" href="char_rnn_classification_tutorial.html">๊ธฐ์ด๋ถํฐ ์์ํ๋ NLP: ๋ฌธ์-๋จ์ RNN์ผ๋ก ์ด๋ฆ ๋ถ๋ฅํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="char_rnn_generation_tutorial.html">๊ธฐ์ด๋ถํฐ ์์ํ๋ NLP: ๋ฌธ์-๋จ์ RNN์ผ๋ก ์ด๋ฆ ์์ฑํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="seq2seq_translation_tutorial.html">๊ธฐ์ด๋ถํฐ ์์ํ๋ NLP: Sequence to Sequence ๋คํธ์ํฌ์ Attention์ ์ด์ฉํ ๋ฒ์ญ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/text_sentiment_ngrams_tutorial.html">torchtext ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ก ํ
์คํธ ๋ถ๋ฅํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/translation_transformer.html"><code class="docutils literal notranslate"><span class="pre">nn.Transformer</span></code> ์ torchtext๋ก ์ธ์ด ๋ฒ์ญํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/torchtext_custom_dataset_tutorial.html">Preprocess custom text dataset using Torchtext</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">๋ฐฑ์๋</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/onnx/intro_onnx.html">Introduction to ONNX</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">๊ฐํํ์ต</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="reinforcement_q_learning.html">๊ฐํ ํ์ต (DQN) ํํ ๋ฆฌ์ผ</a></li>
<li class="toctree-l1"><a class="reference internal" href="reinforcement_ppo.html">Reinforcement Learning (PPO) with TorchRL Tutorial</a></li>
<li class="toctree-l1"><a class="reference internal" href="mario_rl_tutorial.html">๋ง๋ฆฌ์ค ๊ฒ์ RL ์์ด์ ํธ๋ก ํ์ตํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/pendulum.html">Pendulum: Writing your environment and transforms with TorchRL</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">PyTorch ๋ชจ๋ธ์ ํ๋ก๋์
ํ๊ฒฝ์ ๋ฐฐํฌํ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/onnx/intro_onnx.html">Introduction to ONNX</a></li>
<li class="toctree-l1"><a class="reference internal" href="flask_rest_api_tutorial.html">Flask๋ฅผ ์ฌ์ฉํ์ฌ Python์์ PyTorch๋ฅผ REST API๋ก ๋ฐฐํฌํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/Intro_to_TorchScript_tutorial.html">TorchScript ์๊ฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/cpp_export.html">C++์์ TorchScript ๋ชจ๋ธ ๋ก๋ฉํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/super_resolution_with_onnxruntime.html">(์ ํ) PyTorch ๋ชจ๋ธ์ ONNX์ผ๋ก ๋ณํํ๊ณ ONNX ๋ฐํ์์์ ์คํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="realtime_rpi.html">Raspberry Pi 4 ์์ ์ค์๊ฐ ์ถ๋ก (Inference) (30fps!)</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">PyTorch ํ๋กํ์ผ๋ง</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/profiler.html">PyTorch ๋ชจ๋ ํ๋กํ์ผ๋งํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/hta_intro_tutorial.html">Introduction to Holistic Trace Analysis</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/hta_trace_diff_tutorial.html">Trace Diff using Holistic Trace Analysis</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Code Transforms with FX</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="fx_conv_bn_fuser.html">(๋ฒ ํ) FX์์ ํฉ์ฑ๊ณฑ/๋ฐฐ์น ์ ๊ทํ(Convolution/Batch Norm) ๊ฒฐํฉ๊ธฐ(Fuser) ๋ง๋ค๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="fx_profiling_tutorial.html">(beta) Building a Simple CPU Performance Profiler with FX</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">ํ๋ก ํธ์๋ API</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="memory_format_tutorial.html">(๋ฒ ํ) PyTorch๋ฅผ ์ฌ์ฉํ Channels Last ๋ฉ๋ชจ๋ฆฌ ํ์</a></li>
<li class="toctree-l1"><a class="reference internal" href="forward_ad_usage.html">Forward-mode Automatic Differentiation (Beta)</a></li>
<li class="toctree-l1"><a class="reference internal" href="jacobians_hessians.html">Jacobians, Hessians, hvp, vhp, and more: composing function transforms</a></li>
<li class="toctree-l1"><a class="reference internal" href="ensembling.html">๋ชจ๋ธ ์์๋ธ</a></li>
<li class="toctree-l1"><a class="reference internal" href="per_sample_grads.html">Per-sample-gradients</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/cpp_frontend.html">PyTorch C++ ํ๋ก ํธ์๋ ์ฌ์ฉํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/torch-script-parallelism.html">TorchScript์ ๋์ ๋ณ๋ ฌ ์ฒ๋ฆฌ(Dynamic Parallelism)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/cpp_autograd.html">C++ ํ๋ก ํธ์๋์ ์๋ ๋ฏธ๋ถ (autograd)</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">PyTorch ํ์ฅํ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="custom_function_double_backward_tutorial.html">Double Backward with Custom Functions</a></li>
<li class="toctree-l1"><a class="reference internal" href="custom_function_conv_bn_tutorial.html">Fusing Convolution and Batch Norm using Custom Function</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/cpp_extension.html">Custom C++ and CUDA Extensions</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/torch_script_custom_ops.html">Extending TorchScript with Custom C++ Operators</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/torch_script_custom_classes.html">์ปค์คํ
C++ ํด๋์ค๋ก TorchScript ํ์ฅํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/dispatcher.html">Registering a Dispatched Operator in C++</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/extend_dispatcher.html">Extending dispatcher for a new backend in C++</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/privateuseone.html">Facilitating New Backend Integration by PrivateUse1</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">๋ชจ๋ธ ์ต์ ํ</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="../beginner/profiler.html">PyTorch ๋ชจ๋ ํ๋กํ์ผ๋งํ๊ธฐ</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/hyperparameter_tuning_tutorial.html">Ray Tune์ ์ฌ์ฉํ ํ์ดํผํ๋ผ๋ฏธํฐ ํ๋</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/vt_tutorial.html">๋ฐฐํฌ๋ฅผ ์ํด ๋น์ ํธ๋์คํฌ๋จธ(Vision Transformer) ๋ชจ๋ธ ์ต์ ํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="parametrizations.html">Parametrizations Tutorial</a></li>
<li class="toctree-l1"><a class="reference internal" href="pruning_tutorial.html">๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ(Pruning) ํํ ๋ฆฌ์ผ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/dynamic_quantization_tutorial.html">(๋ฒ ํ) LSTM ๊ธฐ๋ฐ ๋จ์ด ๋จ์ ์ธ์ด ๋ชจ๋ธ์ ๋์ ์์ํ</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamic_quantization_bert_tutorial.html">(๋ฒ ํ) BERT ๋ชจ๋ธ ๋์ ์์ํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="quantized_transfer_learning_tutorial.html">(๋ฒ ํ) ์ปดํจํฐ ๋น์ ํํ ๋ฆฌ์ผ์ ์ํ ์์ํ๋ ์ ์ดํ์ต(Quantized Transfer Learning)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/static_quantization_tutorial.html">(๋ฒ ํ) PyTorch์์ Eager Mode๋ฅผ ์ด์ฉํ ์ ์ ์์ํ</a></li>
<li class="toctree-l1"><a class="reference internal" href="torchserve_with_ipex.html">Grokking PyTorch Intel CPU performance from first principles</a></li>
<li class="toctree-l1"><a class="reference internal" href="torchserve_with_ipex_2.html">Grokking PyTorch Intel CPU performance from first principles (Part 2)</a></li>
<li class="toctree-l1"><a class="reference internal" href="nvfuser_intro_tutorial.html">Getting Started - Accelerate Your Scripts with nvFuser</a></li>
<li class="toctree-l1"><a class="reference internal" href="ax_multiobjective_nas_tutorial.html">Multi-Objective NAS with Ax</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_compile_tutorial.html">Introduction to <code class="docutils literal notranslate"><span class="pre">torch.compile</span></code></a></li>
<li class="toctree-l1"><a class="reference internal" href="inductor_debug_cpu.html">Inductor CPU backend debugging and profiling</a></li>
<li class="toctree-l1"><a class="reference internal" href="scaled_dot_product_attention_tutorial.html">(Beta) Scaled Dot Product Attention (SDPA)๋ก ๊ณ ์ฑ๋ฅ ํธ๋์คํฌ๋จธ(Transformers) ๊ตฌํํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="scaled_dot_product_attention_tutorial.html#torch-compile-sdpa"><code class="docutils literal notranslate"><span class="pre">torch.compile</span></code> ๊ณผ ํจ๊ป SDPA ์ฌ์ฉํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="scaled_dot_product_attention_tutorial.html#sdpa-atteition-bias">SDPA๋ฅผ <code class="docutils literal notranslate"><span class="pre">atteition.bias</span></code> ํ์ ํด๋์ค์ ์ฌ์ฉํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="scaled_dot_product_attention_tutorial.html#id8">๊ฒฐ๋ก </a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/knowledge_distillation_tutorial.html">Knowledge Distillation Tutorial</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">๋ณ๋ ฌ ๋ฐ ๋ถ์ฐ ํ์ต</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../distributed/home.html">Distributed and Parallel Training Tutorials</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/dist_overview.html">PyTorch Distributed Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/ddp_series_intro.html">Distributed Data Parallel in PyTorch - Video Tutorials</a></li>
<li class="toctree-l1"><a class="reference internal" href="model_parallel_tutorial.html">๋จ์ผ ๋จธ์ ์ ์ฌ์ฉํ ๋ชจ๋ธ ๋ณ๋ ฌํ ๋ชจ๋ฒ ์ฌ๋ก</a></li>
<li class="toctree-l1"><a class="reference internal" href="ddp_tutorial.html">๋ถ์ฐ ๋ฐ์ดํฐ ๋ณ๋ ฌ ์ฒ๋ฆฌ ์์ํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="dist_tuto.html">PyTorch๋ก ๋ถ์ฐ ์ดํ๋ฆฌ์ผ์ด์
๊ฐ๋ฐํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="FSDP_tutorial.html">Getting Started with Fully Sharded Data Parallel(FSDP)</a></li>
<li class="toctree-l1"><a class="reference internal" href="FSDP_adavnced_tutorial.html">Advanced Model Training with Fully Sharded Data Parallel (FSDP)</a></li>
<li class="toctree-l1"><a class="reference internal" href="TP_tutorial.html">Large Scale Transformer model training with Tensor Parallel (TP)</a></li>
<li class="toctree-l1"><a class="reference internal" href="process_group_cpp_extension_tutorial.html">Cpp ํ์ฅ์ ์ฌ์ฉํ ํ๋ก์ธ์ค ๊ทธ๋ฃน ๋ฐฑ์๋ ์ฌ์ฉ์ ์ ์</a></li>
<li class="toctree-l1"><a class="reference internal" href="rpc_tutorial.html">Getting Started with Distributed RPC Framework</a></li>
<li class="toctree-l1"><a class="reference internal" href="rpc_param_server_tutorial.html">Implementing a Parameter Server Using Distributed RPC Framework</a></li>
<li class="toctree-l1"><a class="reference internal" href="dist_pipeline_parallel_tutorial.html">Distributed Pipeline Parallelism Using RPC</a></li>
<li class="toctree-l1"><a class="reference internal" href="rpc_async_execution.html">Implementing Batch RPC Processing Using Asynchronous Executions</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/rpc_ddp_tutorial.html">๋ถ์ฐ ๋ฐ์ดํฐ ๋ณ๋ ฌ(DDP)๊ณผ ๋ถ์ฐ RPC ํ๋ ์์ํฌ ๊ฒฐํฉ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/ddp_pipeline.html">๋ถ์ฐ ๋ฐ์ดํฐ ๋ณ๋ ฌ ์ฒ๋ฆฌ์ ๋ณ๋ ฌ ์ฒ๋ฆฌ ํ์ดํ๋ผ์ธ์ ์ฌ์ฉํ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/generic_join.html">Distributed Training with Uneven Inputs Using the Join Context Manager</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Edge with ExecuTorch</span></p>
<ul>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/executorch/stable/tutorials/export-to-executorch-tutorial.html">Exporting to ExecuTorch Tutorial</a></li>
<li class="toctree-l1"><a class="reference external" href=" https://pytorch.org/executorch/stable/running-a-model-cpp-tutorial.html">Running an ExecuTorch Model in C++ Tutorial</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/executorch/stable/tutorials/sdk-integration-tutorial.html">Using the ExecuTorch SDK to Profile a Model</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/executorch/stable/demo-apps-ios.html">Building an ExecuTorch iOS Demo App</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/executorch/stable/demo-apps-android.html">Building an ExecuTorch Android Demo App</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/executorch/stable/examples-end-to-end-to-lower-model-to-delegate.html">Lowering a Model as a Delegate</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">์ถ์ฒ ์์คํ
</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="torchrec_tutorial.html">TorchRec ์๊ฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../advanced/sharding.html">Exploring TorchRec sharding</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Multimodality</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/flava_finetuning_tutorial.html">TorchMultimodal ํํ ๋ฆฌ์ผ: FLAVA ๋ฏธ์ธ์กฐ์ </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">
Tutorials
</a> >
</li>
<li>ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/intermediate/tensorboard_profiler_tutorial.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="pytorch-call-to-action-links">
<div id="tutorial-type">intermediate/tensorboard_profiler_tutorial</div>
<div id="google-colab-link">
<img class="call-to-action-img" src="../_static/images/pytorch-colab.svg"/>
<div class="call-to-action-desktop-view">Run in Google Colab</div>
<div class="call-to-action-mobile-view">Colab</div>
</div>
<div id="download-notebook-link">
<img class="call-to-action-notebook-img" src="../_static/images/pytorch-download.svg"/>
<div class="call-to-action-desktop-view">Download Notebook</div>
<div class="call-to-action-mobile-view">Notebook</div>
</div>
<div id="github-view-link">
<img class="call-to-action-img" src="../_static/images/pytorch-github.svg"/>
<div class="call-to-action-desktop-view">View on GitHub</div>
<div class="call-to-action-mobile-view">GitHub</div>
</div>
</div>
<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="sphx-glr-download-link-note admonition note">
<p class="admonition-title">์ฐธ๊ณ </p>
<p>Click <a class="reference internal" href="#sphx-glr-download-intermediate-tensorboard-profiler-tutorial-py"><span class="std std-ref">here</span></a>
to download the full example code</p>
</div>
<div class="sphx-glr-example-title section" id="sphx-glr-intermediate-tensorboard-profiler-tutorial-py">
<span id="id1"></span><h1>ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ<a class="headerlink" href="#sphx-glr-intermediate-tensorboard-profiler-tutorial-py" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h1>
<p><strong>๋ฒ์ญ</strong>: <a class="reference external" href="https://github.com/dnd-qodqks">์๋์ฐ</a></p>
<p>์ด ํํ ๋ฆฌ์ผ์์๋ ํ์ดํ ์น(PyTorch) ํ๋กํ์ผ๋ฌ(profiler)์ ํจ๊ป ํ
์๋ณด๋(TensorBoard) ํ๋ฌ๊ทธ์ธ(plugin)์ ์ฌ์ฉํ์ฌ
๋ชจ๋ธ์ ์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ํ์งํ๋ ๋ฐฉ๋ฒ์ ๋ณด์ฌ ์ค๋๋ค.</p>
<div class="section" id="id2">
<h2>์๊ฐ<a class="headerlink" href="#id2" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>ํ์ดํ ์น(PyTorch) 1.8๋ถํฐ GPU์์ CUDA ์ปค๋(kernel) ์คํ ๋ฟ๋ง ์๋๋ผ
CPU ์์
์ ๊ธฐ๋กํ ์ ์๋ ์
๋ฐ์ดํธ๋ ํ๋กํ์ผ๋ฌ API๊ฐ ํฌํจ๋์ด ์์ต๋๋ค.
ํ๋กํ์ผ๋ฌ๋ ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์์ ์ด๋ฐ ์ ๋ณด๋ฅผ ์๊ฐํํ๊ณ
์ฑ๋ฅ ๋ณ๋ชฉ ํ์์ ๋ํ ๋ถ์์ ์ ๊ณตํ ์ ์์ต๋๋ค.</p>
<p>์ด ํํ ๋ฆฌ์ผ์์๋ ๊ฐ๋จํ Resnet ๋ชจ๋ธ์ ์ฌ์ฉํ์ฌ
ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์ ํ์ฉํ ๋ชจ๋ธ ์ฑ๋ฅ ๋ถ์ ๋ฐฉ๋ฒ์ ๋ณด์ฌ๋๋ฆฌ๊ฒ ์ต๋๋ค.</p>
</div>
<div class="section" id="id3">
<h2>์ค๋น<a class="headerlink" href="#id3" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>์๋ ๋ช
๋ น์ด๋ฅผ ์คํํ์ฌ <a href="#id4"><span class="problematic" id="id5">``</span></a>torch``์ <a href="#id6"><span class="problematic" id="id7">``</span></a>torchvision``์ ์ค์นํฉ๋๋ค:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">torch</span> <span class="n">torchvision</span>
</pre></div>
</div>
</div>
<div class="section" id="id8">
<h2>๊ณผ์ <a class="headerlink" href="#id8" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<ol class="arabic simple">
<li><p>๋ฐ์ดํฐ ๋ฐ ๋ชจ๋ธ ์ค๋น</p></li>
<li><p>ํ๋กํ์ผ๋ฌ๋ฅผ ์ฌ์ฉํ์ฌ ์คํ ์ด๋ฒคํธ(execution events) ๊ธฐ๋ก</p></li>
<li><p>ํ๋กํ์ผ๋ฌ ์คํ</p></li>
<li><p>ํ
์๋ณด๋๋ฅผ ์ฌ์ฉํ์ฌ ๊ฒฐ๊ณผ ํ์ธ ๋ฐ ๋ชจ๋ธ ์ฑ๋ฅ ๋ถ์</p></li>
<li><p>ํ๋กํ์ผ๋ฌ์ ๋์์ผ๋ก ์ฑ๋ฅ ๊ฐ์ </p></li>
<li><p>๋ค๋ฅธ ๊ณ ๊ธ ๊ธฐ๋ฅ์ผ๋ก ์ฑ๋ฅ ๋ถ์</p></li>
<li><p>์ถ๊ฐ ์ฐ์ต: AMD GPU์์ PyTorch ํ๋กํ์ผ๋ง</p></li>
</ol>
<div class="section" id="id9">
<h3>1. ๋ฐ์ดํฐ ๋ฐ ๋ชจ๋ธ ์ค๋น<a class="headerlink" href="#id9" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p>๋จผ์ ํ์ํ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ฅผ ๋ชจ๋ ๋ถ๋ฌ์ต๋๋ค:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span>
<span class="kn">import</span> <span class="nn">torch.optim</span>
<span class="kn">import</span> <span class="nn">torch.profiler</span>
<span class="kn">import</span> <span class="nn">torch.utils.data</span>
<span class="kn">import</span> <span class="nn">torchvision.datasets</span>
<span class="kn">import</span> <span class="nn">torchvision.models</span>
<span class="kn">import</span> <span class="nn">torchvision.transforms</span> <span class="k">as</span> <span class="nn">T</span>
</pre></div>
</div>
<p>์ดํ ์
๋ ฅ ๋ฐ์ดํฐ๋ฅผ ์ค๋นํฉ๋๋ค. ์ด ํํ ๋ฆฌ์ผ์ ๊ฒฝ์ฐ CIFAR10 ๋ฐ์ดํฐ์
์ ์ฌ์ฉํฉ๋๋ค.
์ํ๋ ํ์์ผ๋ก ๋ณํํ๊ณ <a href="#id10"><span class="problematic" id="id11">``</span></a>DataLoader``๋ฅผ ์ฌ์ฉํ์ฌ ๊ฐ ๋ฐฐ์น(batch)๋ฅผ ๋ก๋ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">transform</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">Compose</span><span class="p">(</span>
<span class="p">[</span><span class="n">T</span><span class="o">.</span><span class="n">Resize</span><span class="p">(</span><span class="mi">224</span><span class="p">),</span>
<span class="n">T</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
<span class="n">T</span><span class="o">.</span><span class="n">Normalize</span><span class="p">((</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">),</span> <span class="p">(</span><span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">,</span> <span class="mf">0.5</span><span class="p">))])</span>
<span class="n">train_set</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">datasets</span><span class="o">.</span><span class="n">CIFAR10</span><span class="p">(</span><span class="n">root</span><span class="o">=</span><span class="s1">'./data'</span><span class="p">,</span> <span class="n">train</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">download</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">transform</span><span class="o">=</span><span class="n">transform</span><span class="p">)</span>
<span class="n">train_loader</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span><span class="p">(</span><span class="n">train_set</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="mi">32</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
</pre></div>
</div>
<p>๊ทธ๋ฐ ๋ค์ Resnet ๋ชจ๋ธ, ์์ค ํจ์ ๋ฐ ์ตํฐ๋ง์ด์ ๊ฐ์ฒด๋ฅผ ์์ฑํฉ๋๋ค.
GPU์์ ์คํํ๊ธฐ ์ํด ๋ชจ๋ธ ๋ฐ ์์ค์ GPU ์ฅ์น๋ก ์ด๋ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">device</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s2">"cuda:0"</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">models</span><span class="o">.</span><span class="n">resnet18</span><span class="p">(</span><span class="n">weights</span><span class="o">=</span><span class="s1">'IMAGENET1K_V1'</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
<span class="n">criterion</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">CrossEntropyLoss</span><span class="p">()</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
<span class="n">optimizer</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">SGD</span><span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">lr</span><span class="o">=</span><span class="mf">0.001</span><span class="p">,</span> <span class="n">momentum</span><span class="o">=</span><span class="mf">0.9</span><span class="p">)</span>
<span class="n">model</span><span class="o">.</span><span class="n">train</span><span class="p">()</span>
</pre></div>
</div>
<p>๊ฐ ์
๋ ฅ ๋ฐ์ดํฐ ๋ฐฐ์น์ ๋ํ ํ์ต ๋จ๊ณ๋ฅผ ์ ์ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train</span><span class="p">(</span><span class="n">data</span><span class="p">):</span>
<span class="n">inputs</span><span class="p">,</span> <span class="n">labels</span> <span class="o">=</span> <span class="n">data</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">),</span> <span class="n">data</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
<span class="n">outputs</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="n">inputs</span><span class="p">)</span>
<span class="n">loss</span> <span class="o">=</span> <span class="n">criterion</span><span class="p">(</span><span class="n">outputs</span><span class="p">,</span> <span class="n">labels</span><span class="p">)</span>
<span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
<span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
<span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
</pre></div>
</div>
</div>
<div class="section" id="id12">
<h3>2. ํ๋กํ์ผ๋ฌ๋ฅผ ์ฌ์ฉํ์ฌ ์คํ ์ด๋ฒคํธ ๊ธฐ๋ก<a class="headerlink" href="#id12" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p>ํ๋กํ์ผ๋ฌ๋ ์ปจํ
์คํธ(context) ๊ด๋ฆฌ์๋ฅผ ํตํด ํ์ฑํ๋๊ณ ๋ช ๊ฐ์ง ๋งค๊ฐ๋ณ์๋ฅผ ์ฌ์ฉํ ์ ์์ผ๋ฉฐ,
๊ฐ์ฅ ์ ์ฉํ ๊ฒ์ ์๋์ ๊ฐ์ต๋๋ค:</p>
<ul>
<li><p><code class="docutils literal notranslate"><span class="pre">schedule</span></code> - step (int)์ ๋จ์ผ ๋งค๊ฐ๋ณ์๋ก ๋ฐ์๋ค์ด๊ณ ,
๊ฐ ๋จ๊ณ์์ ์ํํ ํ๋กํ์ผ๋ฌ ์์
์ ๋ฐํํ๋ ํธ์ถ ๊ฐ๋ฅํ ํจ์์
๋๋ค.</p>
<p>์ด ์์์์๋ <code class="docutils literal notranslate"><span class="pre">wait=1,</span> <span class="pre">warmup=1,</span> <span class="pre">active=3,</span> <span class="pre">repeat=1</span></code> ๋ก ์ค์ ๋์ด ์์ผ๋ฉฐ,
ํ๋กํ์ผ๋ฌ๋ ์ฒซ ๋ฒ์งธ ๋จ๊ณ/๋ฐ๋ณต(step/iteration)์ ๊ฑด๋๋๋๋ค.
๋ ๋ฒ์งธ๋ถํฐ ์๋ฐ์
(warming up)์ ์์ํ๋ฉด,
๋ค์ ์ธ ๋ฒ์ ๋ฐ๋ณต์ ๊ธฐ๋กํ๊ณ ,
๊ทธ ํ ์ถ์ (trace)์ ์ฌ์ฉํ ์ ์๊ฒ ๋๊ณ on_trace_ready (์ค์ ๋ ๊ฒฝ์ฐ)๊ฐ ํธ์ถ๋ฉ๋๋ค.
์ ์ฒด์ ์ผ๋ก ์ด ์ฃผ๊ธฐ๊ฐ ํ ๋ฒ ๋ฐ๋ณต๋ฉ๋๋ค. ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์์ ๊ฐ ์ฃผ๊ธฐ๋ ใspanใ์ด๋ผ๊ณ ํฉ๋๋ค.</p>
<p><code class="docutils literal notranslate"><span class="pre">wait</span></code> ๋จ๊ณ์ธ ๋์ ํ๋กํ์ผ๋ฌ๋ ๋นํ์ฑํ๋ฉ๋๋ค.
<code class="docutils literal notranslate"><span class="pre">warmup</span></code> ๋จ๊ณ์ธ ๋์์ ํ๋กํ์ผ๋ฌ๊ฐ ์ถ์ (tracing)์ ์์ํ์ง๋ง ๊ฒฐ๊ณผ๋ ๋ฌด์๋ฉ๋๋ค.
์ด๋ ํ๋กํ์ผ๋ง ๊ณผ๋ถํ(overhead)๋ฅผ ์ค์ด๊ธฐ ์ํจ์
๋๋ค.
ํ๋กํ์ผ๋ง์ ์์ํ ๋ ๊ณผ๋ถํ๋ ํฌ๊ณ ํ๋กํ์ผ๋ง ๊ฒฐ๊ณผ์ ์๊ณก์ ๊ฐ์ ธ์ค๊ธฐ ์ฝ์ต๋๋ค.
<code class="docutils literal notranslate"><span class="pre">active</span></code> ๋จ๊ณ์์ ํ๋กํ์ผ๋ฌ๊ฐ ์๋ํ๋ฉฐ ์ด๋ฒคํธ๋ฅผ ๊ธฐ๋กํฉ๋๋ค.</p>
</li>
<li><p><code class="docutils literal notranslate"><span class="pre">on_trace_ready</span></code> - ๊ฐ ์ฃผ๊ธฐ ๋ง์ง๋ง์ ํธ์ถ๋๋ ํจ์์
๋๋ค;
์ด ์์์์๋ <code class="docutils literal notranslate"><span class="pre">torch.profiler.tensorboard_trace_handler``๋ฅผ</span> <span class="pre">์ฌ์ฉํ์ฌ</span> <span class="pre">ํ
์๋ณด๋์</span> <span class="pre">๊ฒฐ๊ณผ</span> <span class="pre">ํ์ผ์</span> <span class="pre">์์ฑํฉ๋๋ค.</span>
<span class="pre">ํ๋กํ์ผ๋ง</span> <span class="pre">ํ</span> <span class="pre">๊ฒฐ๊ณผ</span> <span class="pre">ํ์ผ์</span> <span class="pre">``./log/resnet18</span></code> ๋๋ ํ ๋ฆฌ์ ์ ์ฅ๋ฉ๋๋ค.
ํ
์๋ณด๋์์ ํ๋กํ์ผ(profile)์ ๋ถ์ํ๋ ค๋ฉด ์ด ๋๋ ํ ๋ฆฌ๋ฅผ <code class="docutils literal notranslate"><span class="pre">logdir</span></code> ๋งค๊ฐ๋ณ์๋ก ์ง์ ํด์ผ ํฉ๋๋ค.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">record_shapes</span></code> - ์ฐ์ฐ์ ์
๋ ฅ์ shape์ ๊ธฐ๋กํ ์ง ์ฌ๋ถ๋ฅผ ๋ํ๋
๋๋ค.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">profile_memory</span></code> - Track tensor memory ํ ๋น/ํ ๋น ํด์ ์ฌ๋ถ๋ฅผ ๋ํ๋
๋๋ค. ์ฃผ์, 1.10 ์ด์ ๋ฒ์ ์ ํ์ดํ ์น๋ฅผ ์ฌ์ฉํ๋ ๊ฒฝ์ฐ
ํ๋กํ์ผ๋ง ์๊ฐ์ด ๊ธธ๋ค๋ฉด ์ด ๊ธฐ๋ฅ์ ๋นํ์ฑํํ๊ฑฐ๋ ์ ๋ฒ์ ์ผ๋ก ์
๊ทธ๋ ์ด๋ํด ์ฃผ์ธ์.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">with_stack</span></code> - ops์ ๋ํ ์์ค ์ ๋ณด(ํ์ผ ๋ฐ ๋ผ์ธ ๋ฒํธ)๋ฅผ ๊ธฐ๋ก ์ฌ๋ถ๋ฅผ ๋ํ๋
๋๋ค.
๋ง์ฝ VS Code์์ ํ
์๋ณด๋๋ฅผ ์คํํ๋ ๊ฒฝ์ฐ (<a class="reference external" href="https://code.visualstudio.com/docs/datascience/pytorch-support#_tensorboard-integration">์ฐธ๊ณ </a>),
์คํ ํ๋ ์(stack frame)์ ํด๋ฆญํ๋ฉด ํน์ ์ฝ๋ ๋ผ์ธ์ผ๋ก ์ด๋ํฉ๋๋ค.</p></li>
</ul>
<div class="highlight-default 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">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">3</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">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">tensorboard_trace_handler</span><span class="p">(</span><span class="s1">'./log/resnet18'</span><span class="p">),</span>
<span class="n">record_shapes</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">profile_memory</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">with_stack</span><span class="o">=</span><span class="kc">True</span>
<span class="p">)</span> <span class="k">as</span> <span class="n">prof</span><span class="p">:</span>
<span class="k">for</span> <span class="n">step</span><span class="p">,</span> <span class="n">batch_data</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">train_loader</span><span class="p">):</span>
<span class="n">prof</span><span class="o">.</span><span class="n">step</span><span class="p">()</span> <span class="c1"># ๊ฐ ๋จ๊ณ์์ ํธ์ถํ์ฌ ํ๋กํ์ผ๋ฌ์๊ฒ ๋จ๊ณ์ ๊ฒฝ๊ณ๋ฅผ ์๋ ค์ผ ํฉ๋๋ค.</span>
<span class="k">if</span> <span class="n">step</span> <span class="o">>=</span> <span class="mi">1</span> <span class="o">+</span> <span class="mi">1</span> <span class="o">+</span> <span class="mi">3</span><span class="p">:</span>
<span class="k">break</span>
<span class="n">train</span><span class="p">(</span><span class="n">batch_data</span><span class="p">)</span>
</pre></div>
</div>
<p>๋ํ, ๋ค์์ non-context ๊ด๋ฆฌ์(manager)๋ ์์(start)/์ ์ง(stop) ๊ธฐ๋ฅ๋ ์ง์๋ฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">prof</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">profile</span><span class="p">(</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">3</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">torch</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">tensorboard_trace_handler</span><span class="p">(</span><span class="s1">'./log/resnet18'</span><span class="p">),</span>
<span class="n">record_shapes</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">with_stack</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">prof</span><span class="o">.</span><span class="n">start</span><span class="p">()</span>
<span class="k">for</span> <span class="n">step</span><span class="p">,</span> <span class="n">batch_data</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">train_loader</span><span class="p">):</span>
<span class="n">prof</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
<span class="k">if</span> <span class="n">step</span> <span class="o">>=</span> <span class="mi">1</span> <span class="o">+</span> <span class="mi">1</span> <span class="o">+</span> <span class="mi">3</span><span class="p">:</span>
<span class="k">break</span>
<span class="n">train</span><span class="p">(</span><span class="n">batch_data</span><span class="p">)</span>
<span class="n">prof</span><span class="o">.</span><span class="n">stop</span><span class="p">()</span>
</pre></div>
</div>
</div>
<div class="section" id="id13">
<h3>3. ํ๋กํ์ผ๋ฌ ์คํ<a class="headerlink" href="#id13" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p>์ ์ฝ๋๋ฅผ ์คํํฉ๋๋ค. ํ๋กํ์ผ๋ง ๊ฒฐ๊ณผ๋ <code class="docutils literal notranslate"><span class="pre">./log/resnet18</span></code> ๋๋ ํ ๋ฆฌ์ ์ ์ฅ๋ฉ๋๋ค.</p>
</div>
<div class="section" id="id14">
<h3>4. ํ
์๋ณด๋๋ฅผ ์ฌ์ฉํ์ฌ ๊ฒฐ๊ณผ ํ์ธ ๋ฐ ๋ชจ๋ธ ์ฑ๋ฅ ๋ถ์<a class="headerlink" href="#id14" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<div class="admonition note">
<p class="admonition-title">์ฐธ๊ณ </p>
<p>ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ(Tensorboard Plugin) ์ง์์ด ์ค๋จ๋์์ผ๋ฏ๋ก, ์๋ ๊ธฐ๋ฅ๋ค ์ค ์ผ๋ถ๋
์ด์ ์ฒ๋ผ ๋์ํ์ง ์์ ์ ์์ต๋๋ค. ์ด์ ๋ํ ๋์์ผ๋ก <a class="reference external" href="https://github.com/pytorch/kineto/tree/main#holistic-trace-analysis">HTA</a>
๋ฅผ ์ฌ์ฉํ ์ ์์ต๋๋ค.</p>
</div>
<p>ํ์ดํ ์น ํ๋กํ์ผ๋ฌ ํ
์๋ณด๋ ํ๋ฌ๊ทธ์ธ์ ์ค์นํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">torch_tb_profiler</span>
</pre></div>
</div>
<p>ํ
์๋ณด๋๋ฅผ ์คํํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">tensorboard</span> <span class="o">--</span><span class="n">logdir</span><span class="o">=./</span><span class="n">log</span>
</pre></div>
</div>
<p>๊ตฌ๊ธ ํฌ๋กฌ(Google Chrome) ๋ธ๋ผ์ฐ์ ๋๋ ๋ง์ดํฌ๋ก์ํํธ ์ฃ์ง(Microsoft Edge) ๋ธ๋ผ์ฐ์ ์์ ํ
์๋ณด๋ ํ๋กํ์ผ(profile) URL์ ์ ์ํฉ๋๋ค. (<strong>Safari ๋ธ๋ผ์ฐ์ ๋ ์ง์ํ์ง ์์ต๋๋ค.</strong>)</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">http</span><span class="p">:</span><span class="o">//</span><span class="n">localhost</span><span class="p">:</span><span class="mi">6006</span><span class="o">/</span><span class="c1">#pytorch_profiler</span>
</pre></div>
</div>
<p>์๋์ ๊ฐ์ด ํ๋กํ์ผ๋ฌ ํ๋ฌ๊ทธ์ธ ํ์ด์ง๋ฅผ ๋ณผ ์ ์์ต๋๋ค.</p>
<ul class="simple">
<li><p>๊ฐ์(Overview)</p></li>
</ul>
<a class="reference internal image-reference" href="../_static/img/profiler_overview1.png"><img alt="../_static/img/profiler_overview1.png" src="../_static/img/profiler_overview1.png" /></a>
<p>๊ฐ์ ํ์ด์ง์๋ ๋ชจ๋ธ ์ฑ๋ฅ์ ๋ํ ๋๋ต์ ์ธ ์์ฝ์ด ํ์๋ฉ๋๋ค.</p>
<p>ใGPU ์์ฝ(GPU Summary)ใ ํจ๋์๋ GPU ๊ตฌ์ฑ, GPU ์ฌ์ฉ๋ ๋ฐ Tensor ์ฝ์ด ์ฌ์ฉ๋์ด ํ์๋ฉ๋๋ค.
์ด ์์ ์์๋ GPU ์ฌ์ฉ๋์ด ๋ฎ์ต๋๋ค.
์ด๋ฌํ ์ธก์ ์งํ(metrics)์ ๋ํ ์์ธํ ๋ด์ฉ์ <a class="reference external" href="https://github.com/pytorch/kineto/blob/main/tb_plugin/docs/gpu_utilization.md">์ฌ๊ธฐ</a> ์์ ํ์ธํด์ฃผ์ธ์.</p>
<p>ใ๋จ๊ณ ์๊ฐ ์ธ๋ถํ(Step Time Breakdown)ใ๋ ๊ฐ ๋จ๊ณ์์ ์ํ๋ ์๊ฐ์ ๋ถํฌ๋ฅผ ๋ณด์ฌ์ค๋๋ค.
์ด ์์ ์์๋ <code class="docutils literal notranslate"><span class="pre">DataLoader</span></code> ๊ณผ๋ถํ๊ฐ ์๋นํ ๊ฒ์ ๋ณผ ์ ์์ต๋๋ค.</p>
<p>ํ๋จ์ ใ์ฑ๋ฅ ๊ถ์ฅ์ฌํญ(Performance Recommendation)ใ์ ํ๋กํ์ผ๋ง ๋ฐ์ดํฐ๋ฅผ ์ฌ์ฉํ์ฌ
๋ฐ์ ๊ฐ๋ฅํ ๋ณ๋ชฉ ํ์์ ์๋์ผ๋ก ๊ฐ์กฐํ๊ณ ,
์คํ ๊ฐ๋ฅํ ์ต์ ํ ์ ์์ ์ ๊ณตํฉ๋๋ค.</p>
<p>์ผ์ชฝ ใ๋ณด๊ธฐ(Views)ใ ๋๋กญ๋ค์ด(dropdown) ๋ชฉ๋ก์์ ๋ณด๊ธฐ ํ์ด์ง๋ฅผ ๋ณ๊ฒฝํ ์ ์์ต๋๋ค.</p>
<img alt="" src="../_static/img/profiler_views_list.png" />
<ul class="simple">
<li><p>์ฐ์ฐ ๋ณด๊ธฐ(Operator view)</p></li>
</ul>
<p>์ฐ์ฐ ๋ณด๊ธฐ๋ ํธ์คํธ ๋๋ ์ฅ์น์์ ์คํ๋๋
๋ชจ๋ ํ์ดํ ์น ์ฐ์ฐ์์ ์ฑ๋ฅ์ ํ์ํฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_operator_view.png"><img alt="../_static/img/profiler_operator_view.png" src="../_static/img/profiler_operator_view.png" /></a>
<p>ใ์
ํ(Self)ใ ๊ธฐ๊ฐ์๋ ํ์ ์ฐ์ฐ์ ์๊ฐ์ด ํฌํจ๋์ง ์์ต๋๋ค.
ใ์ ์ฒด(Total)ใ ๊ธฐ๊ฐ์๋ ํ์ ์ฐ์ฐ์ ์๊ฐ์ด ํฌํจ๋ฉ๋๋ค.</p>
<ul class="simple">
<li><p>ํธ์ถ ์คํ ๋ณด๊ธฐ(View call stack)</p></li>
</ul>
<p>์ฐ์ฐ์์ <a href="#id15"><span class="problematic" id="id16">``</span></a>View Callstack``๋ฅผ ํด๋ฆญํ๋ฉด, ์ด๋ฆ์ ๊ฐ์ง๋ง ์๋ก ๋ค๋ฅธ ์ฐ์ฐ์๊ฐ ํ์๋ฉ๋๋ค.
ํ์ ํ
์ด๋ธ์ <a href="#id17"><span class="problematic" id="id18">``</span></a>View Callstack``๋ฅผ ํด๋ฆญํ๋ฉด, ํธ์ถ ์คํ ํ๋ ์(call stack frames)์ด ํ์๋ฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_callstack.png"><img alt="../_static/img/profiler_callstack.png" src="../_static/img/profiler_callstack.png" /></a>
<p>VS Code ๋ด๋ถ์์ ํ
์๋ณด๋๊ฐ ์คํ๋๋ ๊ฒฝ์ฐ
(<a class="reference external" href="https://devblogs.microsoft.com/python/python-in-visual-studio-code-february-2021-release/#tensorboard-integration">์คํ ๊ฐ์ด๋</a>),
ํธ์ถ ์คํ ํ๋ ์(call stack frame)์ ํด๋ฆญํ๋ฉด ํน์ ์ฝ๋ ๋ผ์ธ์ผ๋ก ์ด๋ํฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_vscode.png"><img alt="../_static/img/profiler_vscode.png" src="../_static/img/profiler_vscode.png" /></a>
<ul class="simple">
<li><p>์ปค๋ ๋ณด๊ธฐ(Kernel view)</p></li>
</ul>
<p>GPU ์ปค๋ ๋ณด๊ธฐ(GPU kernel view)๋ ๋ชจ๋ ์ปค๋(kernel)์ด GPU์ ์๋นํ ์๊ฐ์ ๋ณด์ฌ์ค๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_kernel_view.png"><img alt="../_static/img/profiler_kernel_view.png" src="../_static/img/profiler_kernel_view.png" /></a>
<p>์ฌ์ฉ๋ Tensor ์ฝ์ด:
์ด ์ปค๋(kernel)์ด tensor ์ฝ์ด๋ฅผ ์ฌ์ฉํ๋์ง ์ฌ๋ถ๋ฃฐ ๋ํ๋
๋๋ค.</p>
<p>SM๋น ํ๊ท ๋ธ๋ญ ์:
SM๋น ๋ธ๋ญ ์ = ์ปค๋(kernel)์ ๋ธ๋ญ / GPU์ SM ์.
์ด ์์น๊ฐ 1๋ณด๋ค ์์ผ๋ฉด GPU ๋ฉํฐํ๋ก์ธ์๊ฐ ์์ ํ ์ฌ์ฉ๋์ง ์์์ ๋ํ๋
๋๋ค.
ใSM๋น ํ๊ท ๋ธ๋ญ ์(Mean Blocks per SM)ใ๋ ์ด ์ปค๋ ์ด๋ฆ์ ๋ชจ๋ ์คํ์ ๋ํ ๊ฐ์ค ํ๊ท ์ด๊ณ , ๊ฐ ์คํ ๊ธฐ๊ฐ์ ๊ฐ์ค์น๋ก ์ฌ์ฉํ์์ต๋๋ค.</p>
<p>ํ๊ท ์์ ๋ฌ์ฑ ์ ์ ์จ(Mean Est. Achieved Occupancy):
์์ ๋ฌ์ฑ ์ ์ ์จ(Est. Achieved Occupancy)์ ์ด์ ํดํ(columnโs tooltip)์ ์ ์๋์ด ์์ต๋๋ค.
๋ฉ๋ชจ๋ฆฌ ๋์ญํญ ๊ฒฝ๊ณ ์ปค๋๊ณผ ๊ฐ์ ๋๋ถ๋ถ์ ๊ฒฝ์ฐ, ๋์์๋ก ์ข์ต๋๋ค.
ใํ๊ท ์์ ๋ฌ์ฑ ์ ์ ์จ(Mean Est. Achieved Occupancy)ใ์ ์ปค๋ ์ด๋ฆ์ ๋ชจ๋ ์คํ์ ๋ํ ๊ฐ์ค ํ๊ท ์ด๋ฉฐ,
๊ฐ ์คํ์ ์ง์ ์๊ฐ์ ๊ฐ์ค์น๋ก ์ฌ์ฉํฉ๋๋ค.</p>
<ul class="simple">
<li><p>์ถ์ ๋ณด๊ธฐ(Trace view)</p></li>
</ul>
<p>์ถ์ ๋ณด๊ธฐ๋ ํ๋กํ์ผ๋ ์ฐ์ฐ์์ GPU ์ปค๋์ ํ์๋ผ์ธ์ ๋ณด์ฌ์ค๋๋ค.
์๋์ ๊ฐ์ด ์ ํํ์ฌ ์ธ๋ถ ์ ๋ณด๋ฅผ ํ์ธํ ์ ์์ต๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_trace_view1.png"><img alt="../_static/img/profiler_trace_view1.png" src="../_static/img/profiler_trace_view1.png" /></a>
<p>์ค๋ฅธ์ชฝ ๋๊ตฌ ๋ชจ์์ ์ฌ์ฉํ์ฌ ๊ทธ๋ํ๋ฅผ ์ด๋ํ๊ณ ํ๋/์ถ์ํ ์ ์์ต๋๋ค.
๋ํ ํค๋ณด๋๋ฅผ ์ฌ์ฉํ์ฌ ํ์๋ผ์ธ ์์์ ํ๋/์ด๋ํ ์ ์์ต๋๋ค.
ใwโ๋ฐ ใsใ ํค๋ ๋ง์ฐ์ค ์ค์ฌ์ผ๋ก ํ๋๋๋ฉฐ,
ใaโ์ ใdใ ํค๋ ํ์๋ผ์ธ์ ์ข์ฐ๋ก ์ด๋ํฉ๋๋ค.
์ฝ์ ์ ์๋ ํํ์ด ๋ณด์ผ ๋๊น์ง ์ด ํค๋ฅผ ์ฌ๋ฌ ๋ฒ ๋๋ฅผ ์ ์์ต๋๋ค.</p>
<p>์ญ๋ฐฉํฅ ์ฐ์ฐ์(backward operator)์ ใIncoming Flowใ ํ๋๊ฐ ใforward correspond to backwardใ ๊ฐ์ธ ๊ฒฝ์ฐ,
ํ
์คํธ๋ฅผ ํด๋ฆญํ์ฌ ์์๋๋ ์ ์ง ์ฐ์ฐ์(forward operator)๋ฅผ ๊ฐ์ ธ์ฌ ์ ์์ต๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_trace_view_fwd_bwd.png"><img alt="../_static/img/profiler_trace_view_fwd_bwd.png" src="../_static/img/profiler_trace_view_fwd_bwd.png" /></a>
<p>์ด ์์์์๋ <a href="#id19"><span class="problematic" id="id20">``</span></a>enumerate(DataLoader)``๋ก ์ ๋์ฌ๊ฐ ๋ถ์ ์ด๋ฒคํธ์ ๋ง์ ์๊ฐ์ด ์์๋๋ ๊ฒ์ ํ์ธํ ์ ์์ต๋๋ค.
๊ทธ๋ฆฌ๊ณ ๋๋ถ๋ถ์ ๊ธฐ๊ฐ ๋์ GPU๋ ์ฌ๋ ์ํ์
๋๋ค.
์ด ๊ธฐ๋ฅ์ ํธ์คํธ ์ธก์์ ๋ฐ์ดํฐ๋ฅผ ๋ก๋ํ๊ณ ๋ฐ์ดํฐ๋ฅผ ๋ณํํ๋ ๊ธฐ๋ฅ์ด๊ธฐ ๋๋ฌธ์,
GPU ๋ฆฌ์์ค๊ฐ ๋ญ๋น๋ฉ๋๋ค.</p>
</div>
<div class="section" id="id21">
<h3>5. ํ๋กํ์ผ๋ฌ์ ๋์์ผ๋ก ์ฑ๋ฅ ๊ฐ์ <a class="headerlink" href="#id21" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p>ใ๊ฐ์(Overview)ใ ํ์ด์ง ํ๋จ์ ใ์ฑ๋ฅ ์ถ์ฒ(Performance Recommendation)ใ ์ ์์ ๋ณ๋ชฉ ํ์์ด ใ <a href="#id22"><span class="problematic" id="id23">``</span></a>DataLoader``์์ ์์ํฉ๋๋ค.
ํ์ดํ ์น <a href="#id24"><span class="problematic" id="id25">``</span></a>DataLoader``๋ ๊ธฐ๋ณธ์ ์ผ๋ก ๋จ์ผ ํ๋ก์ธ์ค๋ฅผ ์ฌ์ฉํฉ๋๋ค.
์ฌ์ฉ์๋ ๋งค๊ฐ๋ณ์ <a href="#id26"><span class="problematic" id="id27">``</span></a>num_workers``๋ฅผ ์ค์ ํ์ฌ ๋ค์ค ํ๋ก์ธ์ค ๋ฐ์ดํฐ ๋ก๋๋ฅผ ํ์ฑํํ ์ ์์ต๋๋ค.
์์ธํ ๋ด์ฉ์ <a class="reference external" href="https://pytorch.org/docs/stable/data.html#single-and-multi-process-data-loading">์ฌ๊ธฐ</a> ์ ์์ต๋๋ค.</p>
<p>์ด ์์์์ ใ์ฑ๋ฅ ๊ถ์ฅ์ฌํญ(Performance Recommendation)ใ์ ๋ฐ๋ผ ์๋์ ๊ฐ์ด <a href="#id28"><span class="problematic" id="id29">``</span></a>num_workers``๋ฅผ ์ค์ ํ๊ณ ,
<a href="#id30"><span class="problematic" id="id31">``</span></a>./log/resnet18_4workers``์ ๊ฐ์ ๋ค๋ฅธ ์ด๋ฆ์ <a href="#id32"><span class="problematic" id="id33">``</span></a>tensorboard_trace_handler``๋ก ์ ๋ฌํ ํ ๋ค์ ์คํํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">train_loader</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span><span class="p">(</span><span class="n">train_set</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="mi">32</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span> <span class="n">num_workers</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span>
</pre></div>
</div>
<p>๊ทธ๋ฐ ๋ค์ ์ผ์ชฝ ใ์คํ(Runs)ใ ๋๋กญ๋ค์ด(dropdown) ๋ชฉ๋ก์์ ์ต๊ทผ ํ๋กํ์ผ๋ ์คํ์ ์ ํํฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_overview2.png"><img alt="../_static/img/profiler_overview2.png" src="../_static/img/profiler_overview2.png" /></a>
<p>์์ ๋ณด๊ธฐ(view)์์ ์ด์ ์คํ์ธ 132ms์ ๋นํด ๋จ๊ณ(step) ์๊ฐ์ด ์ฝ 76ms๋ก ๊ฐ์ํ๊ณ ,
<a href="#id34"><span class="problematic" id="id35">``</span></a>DataLoader``์ ์๊ฐ ๊ฐ์๊ฐ ์ฃผ๋ก ๊ธฐ์ฌํ๋ค๋ ๊ฒ์ ์ ์ ์์ต๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_trace_view2.png"><img alt="../_static/img/profiler_trace_view2.png" src="../_static/img/profiler_trace_view2.png" /></a>
<p>์์ ๋ณด๊ธฐ(view)์์ <a href="#id36"><span class="problematic" id="id37">``</span></a>enumerate(DataLoader)``์ ๋ฐํ์์ด ๊ฐ์ํ๊ณ ,
GPU ํ์ฉ๋๊ฐ ์ฆ๊ฐํ๋ ๊ฒ์ ์ ์ ์์ต๋๋ค.</p>
</div>
<div class="section" id="id38">
<h3>6. ๋ค๋ฅธ ๊ณ ๊ธ ๊ธฐ๋ฅ์ผ๋ก ์ฑ๋ฅ ๋ถ์<a class="headerlink" href="#id38" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<ul class="simple">
<li><p>๋ฉ๋ชจ๋ฆฌ ๋ณด๊ธฐ(Memory view)</p></li>
</ul>
<p>๋ฉ๋ชจ๋ฆฌ ํ๋กํ์ผ์ ์ค์ ํ๋ ค๋ฉด <code class="docutils literal notranslate"><span class="pre">torch.profiler.profile</span></code> ์ธ์์์ <a href="#id39"><span class="problematic" id="id40">``</span></a>profile_memory``๋ฅผ <a href="#id41"><span class="problematic" id="id42">``</span></a>True``๋ก ์ค์ ํด์ผ ํฉ๋๋ค.</p>
<p>Azure์ ๊ธฐ์กด ์์ ๋ฅผ ์ฌ์ฉํด ๋ณผ ์ ์์ต๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">azure</span><span class="o">-</span><span class="n">storage</span><span class="o">-</span><span class="n">blob</span>
<span class="n">tensorboard</span> <span class="o">--</span><span class="n">logdir</span><span class="o">=</span><span class="n">https</span><span class="p">:</span><span class="o">//</span><span class="n">torchtbprofiler</span><span class="o">.</span><span class="n">blob</span><span class="o">.</span><span class="n">core</span><span class="o">.</span><span class="n">windows</span><span class="o">.</span><span class="n">net</span><span class="o">/</span><span class="n">torchtbprofiler</span><span class="o">/</span><span class="n">demo</span><span class="o">/</span><span class="n">memory_demo_1_10</span>
</pre></div>
</div>
<p>ํ๋กํ์ผ๋ฌ๋ ํ๋กํ์ผ๋ง ์ค์ ๋ชจ๋ ๋ฉ๋ชจ๋ฆฌ ํ ๋น/ํด์ ์ด๋ฒคํธ์ ํ ๋น์์ ๋ด๋ถ ์ํ๋ฅผ ๊ธฐ๋กํฉ๋๋ค.
๋ฉ๋ชจ๋ฆฌ ๋ณด๊ธฐ(memory view)๋ ๋ค์๊ณผ ๊ฐ์ด ์ธ ๊ฐ์ง ์์๋ก ๊ตฌ์ฑ๋ฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_memory_view.png"><img alt="../_static/img/profiler_memory_view.png" src="../_static/img/profiler_memory_view.png" /></a>
<p>๊ตฌ์ฑ ์์๋ ๊ฐ๊ฐ ๋ฉ๋ชจ๋ฆฌ ๊ณก์ ๊ทธ๋ํ, ๋ฉ๋ชจ๋ฆฌ ์ด๋ฒคํธ ํ
์ด๋ธ ๋ฐ ๋ฉ๋ชจ๋ฆฌ ํต๊ณ ํ
์ด๋ธ์
๋๋ค.</p>
<p>๋ฉ๋ชจ๋ฆฌ ์ ํ์ ใ์ฅ์น(Device)ใ ์ ํ ์์์์ ์ ํํ ์ ์์ต๋๋ค.
์๋ฅผ ๋ค๋ฉด, ๋ค์ ํ์์ ใGPU0ใ์ GPU 0์์์ ๊ฐ ์ฐ์ฐ์์ ๋ฉ๋ก๋ฆฌ ์ฌ์ฉ๋๋ง ๋ณด์ฌ์ฃผ๊ณ , CPU ๋๋ ๋ค๋ฅธ GPU๋ฅผ ํฌํจํ์ง ์๋๋ค๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค.</p>
<p>๋ฉ๋ชจ๋ฆฌ ๊ณก์ ์ ๋ฉ๋ชจ๋ฆฌ ์๋น์ ์ถ์ธ๋ฅผ ๋ณด์ฌ์ค๋๋ค. ใAllocatedใ ๊ณก์ ์ ์ค์ ์ฌ์ฉ ์ค์ธ ์ด ๋ฉ๋ชจ๋ฆฌ,
์๋ฅผ ๋ค๋ฉด tensor๋ฅผ ๋ณด์ฌ์ค๋๋ค. ํ์ดํ ์น์์ ์บ์ฑ ๋ฉ์ปค๋์ฆ(caching mechanism)์ CUDA ํ ๋น๊ธฐ ๋ฐ ์ผ๋ถ ๋ค๋ฅธ ํ ๋น๊ธฐ์ ์ฌ์ฉ๋ฉ๋๋ค.
ใReservedใ ๊ณก์ ์ ํ ๋น์์ ์ํด ์์ฝ๋ ์ด ๋ฉ๋ชจ๋ฆฌ๋ฅผ ๋ณด์ฌ์ค๋๋ค. ๊ทธ๋ํ๋ฅผ ์ขํด๋ฆญํ๊ณ ๋์ด์
์ํ๋ ๋ฒ์์ ์ด๋ฒคํธ๋ฅผ ์ ํํ ์ ์์ต๋๋ค:</p>
<a class="reference internal image-reference" href="../_static/img/profiler_memory_curve_selecting.png"><img alt="../_static/img/profiler_memory_curve_selecting.png" src="../_static/img/profiler_memory_curve_selecting.png" /></a>
<p>์ ํํ ํ์๋ ์ธ ๊ฐ์ง ๊ตฌ์ฑ ์์๊ฐ ์ ํ๋ ๋ฒ์์ ๋ง๊ฒ ์
๋ฐ์ดํธ๋์ด
์์ธํ ์ ๋ณด๋ฅผ ์ป์ ์ ์์ต๋๋ค. ์ด ํ๋ก์ธ์ค๋ฅผ ๋ฐ๋ณตํ๋ฉด, ๋งค์ฐ ์ธ๋ถํ๋ ์ธ๋ถ ์ ๋ณด๋ฅผ ํ๋ํ ์ ์์ต๋๋ค. ๊ทธ๋ํ๋ฅผ ์ฐํด๋ฆญํ๋ฉด
๊ทธ๋ํ๊ฐ ์ด๊ธฐ ์ํ๋ก ์ฌ์ค์ ๋ฉ๋๋ค.</p>
<a class="reference internal image-reference" href="../_static/img/profiler_memory_curve_single.png"><img alt="../_static/img/profiler_memory_curve_single.png" src="../_static/img/profiler_memory_curve_single.png" /></a>
<p>๋ฉ๋ชจ๋ฆฌ ์ด๋ฒคํธ ํ
์ด๋ธ์์ ํ ๋น ๋ฐ ํด์ ์ด๋ฒคํธ๋ ํ๋์ ํญ๋ชฉ์ผ๋ก ์์ผ๋ก ๊ตฌ์ฑ๋ฉ๋๋ค. ใoperatorใ ์ด์๋
ํ ๋น์ ๋ฐ์์ํค๋ ์ฆ์ ATen ์ฐ์ฐ์๊ฐ ํ์๋ฉ๋๋ค. ํ์ดํ ์น์์ ATen ์ฐ์ฐ์๋ ์ผ๋ฐ์ ์ผ๋ก
<code class="docutils literal notranslate"><span class="pre">aten::empty``๋ฅผ</span> <span class="pre">์ฌ์ฉํ์ฌ</span> <span class="pre">๋ฉ๋ชจ๋ฆฌ๋ฅผ</span> <span class="pre">ํ ๋นํฉ๋๋ค.</span> <span class="pre">์๋ฅผ</span> <span class="pre">๋ค์ด,</span> <span class="pre">``aten::ones``์</span> <span class="pre">``aten::empty</span></code> ๋ค์์
<a href="#id43"><span class="problematic" id="id44">``</span></a>aten::fill_``๋ก ๊ตฌํ๋ฉ๋๋ค. ์ฐ์ฐ์ ์ด๋ฆ๋ง <a href="#id45"><span class="problematic" id="id46">``</span></a>aten::empty``๋ก ํ์ํด๋ ๋ณ ๋์์ด ๋์ง ์์ต๋๋ค. ์ด ํน์ํ ๊ฒฝ์ฐ์๋
<a href="#id47"><span class="problematic" id="id48">``</span></a>aten::ones (aten::empty)``๋ก ํ์๋ฉ๋๋ค. ใํ ๋น ์๊ฐ(Allocation Time)ใ, ใํด์ ์๊ฐ(Release Time)ใ ๋ฐ ใ๊ธฐ๊ฐ(Duration)ใ์
์ด๋ฒคํธ๊ฐ ์๊ฐ ๋ฒ์๋ฅผ ๋ฒ์ด๋๋ ๊ฒฝ์ฐ ์ด์ ๋ฐ์ดํฐ๊ฐ ๋๋ฝ๋ ์ ์์ต๋๋ค.</p>
<p>๋ฉ๋ชจ๋ฆฌ ํต๊ณ ํ
์ด๋ธ์์, ใํฌ๊ธฐ ์ฆ๊ฐ(Size Increase)ใ ์ด์ ๋ชจ๋ ํ ๋น ํฌ๊ธฐ๋ฅผ ํฉ์ฐํ๊ณ ๋ชจ๋ ๋ฉ๋ชจ๋ฆฌ ๋ฆด๋ฆฌ์ค(release)
ํฌ๊ธฐ๋ฅผ ๋บ ๊ฐ, ์ฆ, ์ด ์ฐ์ฐ์ ์ดํ์ ๋ฉ๋ชจ๋ฆฌ ์ฌ์ฉ๋ ์ ์ฆ๊ฐ ๊ฐ์
๋๋ค. ใ์์ฒด ํฌ๊ธฐ ์ฆ๊ฐ(Self Size Increase)ใ ์ด์
ใํฌ๊ธฐ ์ฆ๊ฐ(Size Increase)ใ์ ์ ์ฌ ํ์ง๋ง, ํ์ ์ฐ์ฐ์์ ํ ๋น์ ๊ณ์ฐํ์ง ์์ต๋๋ค. ATen ์ฐ์ฐ์์ ๊ตฌํ ์ธ๋ถ ์ฌํญ๊ณผ
๊ด๋ จํ์ฌ, ์ผ๋ถ ์ฐ์ฐ์๋ ๋ค๋ฅธ ์ฐ์ฐ์๋ฅผ ํธ์ถํ ์ ์์ผ๋ฏ๋ก, ๋ฉ๋ชจ๋ฆฌ ํ ๋น์ ์ฝ ์คํ์ ๋ชจ๋ ์์ค์์
๋ฐ์ํ ์ ์์ต๋๋ค. ์ฆ, ใ์์ฒด ํฌ๊ธฐ ์ฆ๊ฐ(Self Size Increase)ใ๋ ํ์ฌ ์์ค์ ์ฝ ์คํ์์ ๋ฉ๋ชจ๋ฆฌ ์ฌ์ฉ๋ ์ฆ๊ฐ๋ง์ ๊ณ์ฐํฉ๋๋ค.
๋ง์ง๋ง์ผ๋ก, ใํ ๋น ํฌ๊ธฐ(Allocation Size)ใ ์ด์ ๋ฉ๋ชจ๋ฆฌ ๋ฆด๋ฆฌ์ค๋ฅผ ๊ณ ๋ คํ์ง ์๊ณ ๋ชจ๋ ํ ๋น์ ํฉ์ฐํฉ๋๋ค.</p>
<ul class="simple">
<li><p>๋ถ์ฐ ๋ณด๊ธฐ(Distributed view)</p></li>
</ul>
<p>์ด์ ํ๋ฌ๊ทธ์ธ์ NCCL/GLOO๋ฅผ ๋ฐฑ์๋๋ก ์ฌ์ฉํ๋ DDP ํ๋กํ์ผ๋ง์ ๋ํ ๋ถ์ฐ ๋ณด๊ธฐ๋ฅผ ์ง์ํฉ๋๋ค.</p>
<p>Azure์ ๊ธฐ์กด ์์ ๋ฅผ ์ฌ์ฉํด ๋ณผ ์ ์์ต๋๋ค:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">azure</span><span class="o">-</span><span class="n">storage</span><span class="o">-</span><span class="n">blob</span>
<span class="n">tensorboard</span> <span class="o">--</span><span class="n">logdir</span><span class="o">=</span><span class="n">https</span><span class="p">:</span><span class="o">//</span><span class="n">torchtbprofiler</span><span class="o">.</span><span class="n">blob</span><span class="o">.</span><span class="n">core</span><span class="o">.</span><span class="n">windows</span><span class="o">.</span><span class="n">net</span><span class="o">/</span><span class="n">torchtbprofiler</span><span class="o">/</span><span class="n">demo</span><span class="o">/</span><span class="n">distributed_bert</span>
</pre></div>
</div>
<a class="reference internal image-reference" href="../_static/img/profiler_distributed_view.png"><img alt="../_static/img/profiler_distributed_view.png" src="../_static/img/profiler_distributed_view.png" /></a>
<p>ใ์ปดํจํ
/์ปค๋ฎค๋์ผ์ด์
๊ฐ์(Computation/Communication Overview)ใ์๋ ์ปดํจํ
/์ปค๋ฎค๋์ผ์ด์
๋น์จ๊ณผ ์ค๋ณต ์ ๋๊ฐ ํ์๋ฉ๋๋ค.
์ด ๋ณด๊ธฐ์์, ์ฌ์ฉ์๋ ์์
์ ๊ฐ์ ๋ก๋ ๋ฐธ๋ฐ์ฑ ๋ฌธ์ ๋ฅผ ํ์
ํ ์ ์์ต๋๋ค.
์๋ฅผ ๋ค์ด, ํ ์์
์์ ์ฐ์ฐ + ์ค๋ณต ์๊ฐ์ด ๋ค๋ฅธ ์์
์๋ณด๋ค ํจ์ฌ ํฐ ๊ฒฝ์ฐ,
๋ก๋ ๋ฐธ๋ฐ์ฑ์ ๋ฌธ์ ๊ฐ ์๊ฑฐ๋ ์ด ์์
์๊ฐ ๋์ค์(straggler)์ผ ์ ์์ต๋๋ค.</p>
<p>ใ๋๊ธฐํ/์ปค๋ฎค๋์ผ์ด์
๊ฐ์(Synchronizing/Communication Overview)ใ๋ ํต์ ์ ํจ์จ์ฑ์ ๋ณด์ฌ์ค๋๋ค.
ใ๋ฐ์ดํฐ ๊ตํ ์๊ฐ(Data Transfer Time)ใ์ ์ค์ ๋ฐ์ดํฐ๋ฅผ ๊ตํํ๋ ์๊ฐ์
๋๋ค.
ใ๋๊ธฐํ ์๊ฐ(Synchronizing Time)ใ์ ๋ค๋ฅธ ์์
์์ ๋๊ธฐ ๋ฐ ๋๊ธฐํํ๋ ์๊ฐ์
๋๋ค.</p>
<p>ํ ์์
์์ ใ๋๊ธฐํ ์๊ฐใ์ด ๋ค๋ฅธ ์์
์ ๋ณด๋ค ํจ์ฌ ์งง๋ค๋ฉดโ,
์ด ์์
์๋ ๋ค๋ฅธ ์์
์๋ณด๋ค ๋ ๋ง์ ๊ณ์ฐ ์์
๋์ ๊ฐ์ง ์ ์๋ ๋์ค์(straggler)์ผ ์ ์์ต๋๋คโ.</p>
<p>ใ์ปค๋ฎค๋์ผ์ด์
์์
ํต๊ณ(Communication Operations Stats)ใ๋ ๊ฐ ์์
์์ ๋ชจ๋ ํต์ ์์
์ ๋ํ ์ธ๋ถ ํต๊ณ๋ฅผ ์์ฝํฉ๋๋ค.</p>
</div>
<div class="section" id="amd-gpu-pytorch">
<h3>7. ์ถ๊ฐ ์ฐ์ต: AMD GPU์์ PyTorch ํ๋กํ์ผ๋ง<a class="headerlink" href="#amd-gpu-pytorch" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p>The AMD ROCm Platform is an open-source software stack designed for GPU computation, consisting of drivers, development tools, and APIs.
We can run the above mentioned steps on AMD GPUs. In this section, we will use Docker to install the ROCm base development image
before installing PyTorch.</p>
<p>For the purpose of example, letโs create a directory called <code class="docutils literal notranslate"><span class="pre">profiler_tutorial</span></code>, and save the code in <strong>Step 1</strong> as <code class="docutils literal notranslate"><span class="pre">test_cifar10.py</span></code> in this directory.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">mkdir</span> <span class="o">~/</span><span class="n">profiler_tutorial</span>
<span class="n">cd</span> <span class="n">profiler_tutorial</span>
<span class="n">vi</span> <span class="n">test_cifar10</span><span class="o">.</span><span class="n">py</span>
</pre></div>
</div>
<p>At the time of this writing, the Stable(<code class="docutils literal notranslate"><span class="pre">2.1.1</span></code>) Linux version of PyTorch on ROCm Platform is <a class="reference external" href="https://pytorch.org/get-started/locally/">ROCm 5.6</a>.</p>
<ul class="simple">
<li><p>Obtain a base Docker image with the correct user-space ROCm version installed from <a class="reference external" href="https://hub.docker.com/repository/docker/rocm/dev-ubuntu-20.04">Docker Hub</a>.</p></li>
</ul>
<p>It is <code class="docutils literal notranslate"><span class="pre">rocm/dev-ubuntu-20.04:5.6</span></code>.</p>
<ul class="simple">
<li><p>Start the ROCm base Docker container:</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">docker</span> <span class="n">run</span> <span class="o">-</span><span class="n">it</span> <span class="o">--</span><span class="n">network</span><span class="o">=</span><span class="n">host</span> <span class="o">--</span><span class="n">device</span><span class="o">=/</span><span class="n">dev</span><span class="o">/</span><span class="n">kfd</span> <span class="o">--</span><span class="n">device</span><span class="o">=/</span><span class="n">dev</span><span class="o">/</span><span class="n">dri</span> <span class="o">--</span><span class="n">group</span><span class="o">-</span><span class="n">add</span><span class="o">=</span><span class="n">video</span> <span class="o">--</span><span class="n">ipc</span><span class="o">=</span><span class="n">host</span> <span class="o">--</span><span class="n">cap</span><span class="o">-</span><span class="n">add</span><span class="o">=</span><span class="n">SYS_PTRACE</span> <span class="o">--</span><span class="n">security</span><span class="o">-</span><span class="n">opt</span> <span class="n">seccomp</span><span class="o">=</span><span class="n">unconfined</span> <span class="o">--</span><span class="n">shm</span><span class="o">-</span><span class="n">size</span> <span class="mi">8</span><span class="n">G</span> <span class="o">-</span><span class="n">v</span> <span class="o">~/</span><span class="n">profiler_tutorial</span><span class="p">:</span><span class="o">/</span><span class="n">profiler_tutorial</span> <span class="n">rocm</span><span class="o">/</span><span class="n">dev</span><span class="o">-</span><span class="n">ubuntu</span><span class="o">-</span><span class="mf">20.04</span><span class="p">:</span><span class="mf">5.6</span>
</pre></div>
</div>
<ul class="simple">
<li><p>Inside the container, install any dependencies needed for installing the wheels package.</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">sudo</span> <span class="n">apt</span> <span class="n">update</span>
<span class="n">sudo</span> <span class="n">apt</span> <span class="n">install</span> <span class="n">libjpeg</span><span class="o">-</span><span class="n">dev</span> <span class="n">python3</span><span class="o">-</span><span class="n">dev</span> <span class="o">-</span><span class="n">y</span>
<span class="n">pip3</span> <span class="n">install</span> <span class="n">wheel</span> <span class="n">setuptools</span>
<span class="n">sudo</span> <span class="n">apt</span> <span class="n">install</span> <span class="n">python</span><span class="o">-</span><span class="ow">is</span><span class="o">-</span><span class="n">python3</span>
</pre></div>
</div>
<ul class="simple">
<li><p>Install the wheels:</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip3</span> <span class="n">install</span> <span class="n">torch</span> <span class="n">torchvision</span> <span class="n">torchaudio</span> <span class="o">--</span><span class="n">index</span><span class="o">-</span><span class="n">url</span> <span class="n">https</span><span class="p">:</span><span class="o">//</span><span class="n">download</span><span class="o">.</span><span class="n">pytorch</span><span class="o">.</span><span class="n">org</span><span class="o">/</span><span class="n">whl</span><span class="o">/</span><span class="n">rocm5</span><span class="mf">.6</span>
</pre></div>
</div>
<ul class="simple">
<li><p>Install the <code class="docutils literal notranslate"><span class="pre">torch_tb_profiler</span></code>, and then, run the Python file <code class="docutils literal notranslate"><span class="pre">test_cifar10.py</span></code>:</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">torch_tb_profiler</span>
<span class="n">cd</span> <span class="o">/</span><span class="n">profiler_tutorial</span>
<span class="n">python</span> <span class="n">test_cifar10</span><span class="o">.</span><span class="n">py</span>
</pre></div>
</div>
<p>Now, we have all the data needed to view in TensorBoard:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">tensorboard</span> <span class="o">--</span><span class="n">logdir</span><span class="o">=./</span><span class="n">log</span>
</pre></div>
</div>
<p>Choose different views as described in <strong>Step 4</strong>. For example, below is the <strong>Operator</strong> View:</p>
<a class="reference internal image-reference" href="../_static/img/profiler_rocm_tensorboard_operartor_view.png"><img alt="../_static/img/profiler_rocm_tensorboard_operartor_view.png" src="../_static/img/profiler_rocm_tensorboard_operartor_view.png" /></a>
<p>At the time this section is written, <strong>Trace</strong> view does not work and it displays nothing. You can work around by typing <code class="docutils literal notranslate"><span class="pre">chrome://tracing</span></code> in your Chrome Browser.</p>
<ul class="simple">
<li><p>Copy the <code class="docutils literal notranslate"><span class="pre">trace.json</span></code> file under <code class="docutils literal notranslate"><span class="pre">~/profiler_tutorial/log/resnet18</span></code> directory to the Windows.</p></li>
</ul>
<p>You may need to copy the file by using <code class="docutils literal notranslate"><span class="pre">scp</span></code> if the file is located in a remote location.</p>
<ul class="simple">
<li><p>Click <strong>Load</strong> button to load the trace JSON file from the <code class="docutils literal notranslate"><span class="pre">chrome://tracing</span></code> page in the browser.</p></li>
</ul>
<a class="reference internal image-reference" href="../_static/img/profiler_rocm_chrome_trace_view.png"><img alt="../_static/img/profiler_rocm_chrome_trace_view.png" src="../_static/img/profiler_rocm_chrome_trace_view.png" /></a>
<p>As mentioned previously, you can move the graph and zoom in and out.
You can also use keyboard to zoom and move around inside the timeline.
The <code class="docutils literal notranslate"><span class="pre">w</span></code> and <code class="docutils literal notranslate"><span class="pre">s</span></code> keys zoom in centered around the mouse,
and the <code class="docutils literal notranslate"><span class="pre">a</span></code> and <code class="docutils literal notranslate"><span class="pre">d</span></code> keys move the timeline left and right.
You can hit these keys multiple times until you see a readable representation.</p>
</div>
</div>
<div class="section" id="id49">
<h2>๋ ์์๋ณด๊ธฐ<a class="headerlink" href="#id49" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>ํ์ต์ ๊ณ์ํ๋ ค๋ฉด ๋ค์ ๋ฌธ์๋ฅผ ์ฐธ์กฐํ์๊ณ ,
<a class="reference external" href="https://github.com/pytorch/kineto/issues">์ฌ๊ธฐ</a> ์์ ์์ ๋กญ๊ฒ ์ด์๋ฅผ ์ด์ด๋ณด์ธ์.</p>
<ul class="simple">
<li><p><a class="reference external" href="https://github.com/pytorch/kineto/tree/master/tb_plugin">PyTorch TensorBoard Profiler Github</a></p></li>
<li><p><a class="reference external" href="https://pytorch.org/docs/master/profiler.html">torch.profiler API</a></p></li>
<li><p><a class="reference external" href="https://github.com/pytorch/kineto/tree/main#holistic-trace-analysis">HTA</a></p></li>
</ul>
<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 0 minutes 0.000 seconds)</p>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-intermediate-tensorboard-profiler-tutorial-py">
<div class="sphx-glr-download sphx-glr-download-python docutils container">
<p><a class="reference download internal" download="" href="../_downloads/67e47b6d6793c700666471b688068f72/tensorboard_profiler_tutorial.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">tensorboard_profiler_tutorial.py</span></code></a></p>
</div>
<div class="sphx-glr-download sphx-glr-download-jupyter docutils container">
<p><a class="reference download internal" download="" href="../_downloads/0aec568a42e89122e5ca293c86289287/tensorboard_profiler_tutorial.ipynb"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Jupyter</span> <span class="pre">notebook:</span> <span class="pre">tensorboard_profiler_tutorial.ipynb</span></code></a></p>
</div>
</div>
<p class="sphx-glr-signature"><a class="reference external" href="https://sphinx-gallery.github.io">Gallery generated by Sphinx-Gallery</a></p>
</div>
</div>
</article>
</div>
<footer>
<div class="rst-footer-buttons" role="navigation" aria-label="footer navigation">
<a href="../beginner/hyperparameter_tuning_tutorial.html" class="btn btn-neutral float-right" title="Ray Tune์ ์ฌ์ฉํ ํ์ดํผํ๋ผ๋ฏธํฐ ํ๋" accesskey="n" rel="next">Next <img src="../_static/images/chevron-right-orange.svg" class="next-page"></a>
<a href="../advanced/privateuseone.html" class="btn btn-neutral" title="Facilitating New Backend Integration by PrivateUse1" accesskey="p" rel="prev"><img src="../_static/images/chevron-right-orange.svg" class="previous-page"> Previous</a>
</div>
<hr class="community-hr hr-top" />
<div class="community-container">
<div class="community-prompt">๋ ๊ถ๊ธํ์๊ฑฐ๋ ๊ฐ์ ํ ๋ด์ฉ์ด ์์ผ์ ๊ฐ์? ์ปค๋ฎค๋ํฐ์ ์ฐธ์ฌํด๋ณด์ธ์!</div>
<div class="community-link"><a href="https://discuss.pytorch.kr/" aria-label="PyTorchKoreaCommunity">ํ๊ตญ์ด ์ปค๋ฎค๋ํฐ ๋ฐ๋ก๊ฐ๊ธฐ</a></div>
</div>
<hr class="community-hr hr-bottom"/>
<hr class="rating-hr hr-top" />
<div class="rating-container">
<div class="rating-prompt">์ด ํํ ๋ฆฌ์ผ์ด ์ด๋ ์
จ๋์? ํ๊ฐํด์ฃผ์๋ฉด ์ดํ ๊ฐ์ ์ ์ฐธ๊ณ ํ๊ฒ ์ต๋๋ค! :)</div>
<div class="stars-outer">
<i class="far fa-star" title="1 Star" data-behavior="tutorial-rating" data-count="1"></i>
<i class="far fa-star" title="2 Stars" data-behavior="tutorial-rating" data-count="2"></i>
<i class="far fa-star" title="3 Stars" data-behavior="tutorial-rating" data-count="3"></i>
<i class="far fa-star" title="4 Stars" data-behavior="tutorial-rating" data-count="4"></i>
<i class="far fa-star" title="5 Stars" data-behavior="tutorial-rating" data-count="5"></i>
</div>
</div>
<hr class="rating-hr hr-bottom"/>
<div role="contentinfo">
<p>
© Copyright 2018-2024, PyTorch & ํ์ดํ ์น ํ๊ตญ ์ฌ์ฉ์ ๋ชจ์(PyTorch Korea User Group).
</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>
if((window.location.href.indexOf("/prototype/")!= -1) && (window.location.href.indexOf("/prototype/prototype_index")< 1))
{
var div = '<div class="admonition note"><p class="admonition-title">Note</p><p><i class="fa fa-flask" aria-hidden="true"> </i> ์ด ํํ ๋ฆฌ์ผ์ ํ๋กํ ํ์
(prototype) ๊ธฐ๋ฅ๋ค์ ๋ํด์ ์ค๋ช
ํ๊ณ ์์ต๋๋ค. ํ๋กํ ํ์
๊ธฐ๋ฅ์ ์ผ๋ฐ์ ์ผ๋ก ํผ๋๋ฐฑ ๋ฐ ํ
์คํธ์ฉ์ผ๋ก, ๋ฐํ์ ํ๋๊ทธ ์์ด๋ PyPI๋ Conda๋ก ๋ฐฐํฌ๋๋ ๋ฐ์ด๋๋ฆฌ์์๋ ์ฌ์ฉํ ์ ์์ต๋๋ค.</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="#">ํ
์๋ณด๋๋ฅผ ์ด์ฉํ ํ์ดํ ์น ํ๋กํ์ผ๋ฌ</a><ul>
<li><a class="reference internal" href="#id2">์๊ฐ</a></li>
<li><a class="reference internal" href="#id3">์ค๋น</a></li>
<li><a class="reference internal" href="#id8">๊ณผ์ </a><ul>
<li><a class="reference internal" href="#id9">1. ๋ฐ์ดํฐ ๋ฐ ๋ชจ๋ธ ์ค๋น</a></li>
<li><a class="reference internal" href="#id12">2. ํ๋กํ์ผ๋ฌ๋ฅผ ์ฌ์ฉํ์ฌ ์คํ ์ด๋ฒคํธ ๊ธฐ๋ก</a></li>
<li><a class="reference internal" href="#id13">3. ํ๋กํ์ผ๋ฌ ์คํ</a></li>
<li><a class="reference internal" href="#id14">4. ํ
์๋ณด๋๋ฅผ ์ฌ์ฉํ์ฌ ๊ฒฐ๊ณผ ํ์ธ ๋ฐ ๋ชจ๋ธ ์ฑ๋ฅ ๋ถ์</a></li>
<li><a class="reference internal" href="#id21">5. ํ๋กํ์ผ๋ฌ์ ๋์์ผ๋ก ์ฑ๋ฅ ๊ฐ์ </a></li>
<li><a class="reference internal" href="#id38">6. ๋ค๋ฅธ ๊ณ ๊ธ ๊ธฐ๋ฅ์ผ๋ก ์ฑ๋ฅ ๋ถ์</a></li>
<li><a class="reference internal" href="#amd-gpu-pytorch">7. ์ถ๊ฐ ์ฐ์ต: AMD GPU์์ PyTorch ํ๋กํ์ผ๋ง</a></li>
</ul>
</li>
<li><a class="reference internal" href="#id49">๋ ์์๋ณด๊ธฐ</a></li>
</ul>
</li>
</ul>
</div>
</div>
</div>
</section>
</div>
<script type="text/javascript" id="documentation_options" data-url_root="../" src="../_static/documentation_options.js"></script>
<script data-url_root="../" id="documentation_options" src="../_static/documentation_options.js"></script>
<script src="../_static/jquery.js"></script>
<script src="../_static/underscore.js"></script>
<script src="../_static/_sphinx_javascript_frameworks_compat.js"></script>
<script src="../_static/doctools.js"></script>
<script src="../_static/clipboard.min.js"></script>
<script src="../_static/copybutton.js"></script>
<script src="../_static/translations.js"></script>
<script src="../_static/katex.min.js"></script>
<script src="../_static/auto-render.min.js"></script>
<script src="../_static/katex_autorenderer.js"></script>
<script src="../_static/design-tabs.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>
// Helper function to make it easier to call dataLayer.push()
function gtag(){window.dataLayer.push(arguments);}
//add microsoft link
if(window.location.href.indexOf("/beginner/basics/")!= -1)
{
var url="https://docs.microsoft.com/learn/paths/pytorch-fundamentals/?wt.mc_id=aiml-7486-cxa";
switch(window.location.pathname.split("/").pop().replace('.html',''))
{
case"quickstart_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/9-quickstart?WT.mc_id=aiml-7486-cxa";
break;
case"tensorqs_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/2-tensors?WT.mc_id=aiml-7486-cxa";
break;
case"data_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/3-data?WT.mc_id=aiml-7486-cxa";
break;
case"transforms_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/4-transforms?WT.mc_id=aiml-7486-cxa";
break;
case"buildmodel_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/5-model?WT.mc_id=aiml-7486-cxa";
break;
case"autogradqs_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/6-autograd?WT.mc_id=aiml-7486-cxa";
break;
case"optimization_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/7-optimization?WT.mc_id=aiml-7486-cxa";
break;
case"saveloadrun_tutorial":
url="https://docs.microsoft.com/learn/modules/intro-machine-learning-pytorch/8-inference?WT.mc_id=aiml-7486-cxa";
}
$(".pytorch-call-to-action-links").children().first().before("<a href="+url+' data-behavior="call-to-action-event" data-response="Run in Microsoft Learn" target="_blank"><div id="microsoft-learn-link" style="padding-bottom: 0.625rem;border-bottom: 1px solid #f3f4f7;padding-right: 2.5rem;display: -webkit-box; display: -ms-flexbox; display: flex; -webkit-box-align: center;-ms-flex-align: center;align-items: center;"><img class="call-to-action-img" src="../../_static/images/microsoft-logo.svg"/><div class="call-to-action-desktop-view">Run in Microsoft Learn</div><div class="call-to-action-mobile-view">Learn</div></div></a>')
}
</script>
<script async src="https://www.googletagmanager.com/gtag/js?id=G-LZRD6GXDLF"></script>