-
Notifications
You must be signed in to change notification settings - Fork 258
/
Copy pathpipeline_tutorial.html
1244 lines (1044 loc) ยท 126 KB
/
pipeline_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="ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ" />
<meta property="og:type" content="article" />
<meta property="og:url" content="https://tutorials.pytorch.kr/intermediate/pipeline_tutorial.html" />
<meta property="og:site_name" content="PyTorch Tutorials KR" />
<meta property="og:description" content="Author: Pritam Damania, ๋ฒ์ญ: ๋ฐฑ์ ํฌ,. ์ด ํํ ๋ฆฌ์ผ์ ํ์ดํ๋ผ์ธ(pipeline) ๋ณ๋ ฌํ(parallelism)๋ฅผ ์ฌ์ฉํ์ฌ ์ฌ๋ฌ GPU์ ๊ฑธ์น ๊ฑฐ๋ํ ํธ๋์คํฌ๋จธ(transformer) ๋ชจ๋ธ์ ์ด๋ป๊ฒ ํ์ต์ํค๋์ง ๋ณด์ฌ์ค๋๋ค. NN.TRANSFORMER ์ TORCHTEXT ๋ก ์ํ์ค-ํฌ-์ํ์ค(SEQUENCE-TO-SEQUENCE) ๋ชจ๋ธ๋งํ๊ธฐ ํํ ๋ฆฌ์ผ์ ํ์ฅํ์ด๋ฉฐ ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๊ฐ ์ด๋ป๊ฒ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ ์ฐ์ด๋์ง ์ฆ๋ช
ํ๊ธฐ ์ํด ์ด์ ํํ ๋ฆฌ์ผ์์์ ๋ชจ๋ธ ๊ท๋ชจ๋ฅผ ์ฆ๊ฐ์์ผฐ์ต๋๋ค. ์ ์๊ณผ๋ชฉ(Prerequi..." />
<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="Author: Pritam Damania, ๋ฒ์ญ: ๋ฐฑ์ ํฌ,. ์ด ํํ ๋ฆฌ์ผ์ ํ์ดํ๋ผ์ธ(pipeline) ๋ณ๋ ฌํ(parallelism)๋ฅผ ์ฌ์ฉํ์ฌ ์ฌ๋ฌ GPU์ ๊ฑธ์น ๊ฑฐ๋ํ ํธ๋์คํฌ๋จธ(transformer) ๋ชจ๋ธ์ ์ด๋ป๊ฒ ํ์ต์ํค๋์ง ๋ณด์ฌ์ค๋๋ค. NN.TRANSFORMER ์ TORCHTEXT ๋ก ์ํ์ค-ํฌ-์ํ์ค(SEQUENCE-TO-SEQUENCE) ๋ชจ๋ธ๋งํ๊ธฐ ํํ ๋ฆฌ์ผ์ ํ์ฅํ์ด๋ฉฐ ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๊ฐ ์ด๋ป๊ฒ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ ์ฐ์ด๋์ง ์ฆ๋ช
ํ๊ธฐ ์ํด ์ด์ ํํ ๋ฆฌ์ผ์์์ ๋ชจ๋ธ ๊ท๋ชจ๋ฅผ ์ฆ๊ฐ์์ผฐ์ต๋๋ค. ์ ์๊ณผ๋ชฉ(Prerequi..." />
<meta property="og:ignore_canonical" content="true" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ — ํ์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" />
<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>
<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="tensorboard_profiler_tutorial.html">ํ
์๋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>ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/intermediate/pipeline_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/pipeline_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-pipeline-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-pipeline-tutorial-py">
<span id="id1"></span><h1>ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ<a class="headerlink" href="#sphx-glr-intermediate-pipeline-tutorial-py" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h1>
<dl class="simple">
<dt><strong>Author</strong>: <a class="reference external" href="https://github.com/pritamdamania87">Pritam Damania</a></dt><dd><p><strong>๋ฒ์ญ</strong>: <a class="reference external" href="https://github.com/spongebob03">๋ฐฑ์ ํฌ</a></p>
</dd>
</dl>
<p>์ด ํํ ๋ฆฌ์ผ์ ํ์ดํ๋ผ์ธ(pipeline) ๋ณ๋ ฌํ(parallelism)๋ฅผ ์ฌ์ฉํ์ฌ ์ฌ๋ฌ GPU์ ๊ฑธ์น ๊ฑฐ๋ํ ํธ๋์คํฌ๋จธ(transformer)
๋ชจ๋ธ์ ์ด๋ป๊ฒ ํ์ต์ํค๋์ง ๋ณด์ฌ์ค๋๋ค. <a class="reference external" href="https://tutorials.pytorch.kr/beginner/transformer_tutorial.html">NN.TRANSFORMER ์ TORCHTEXT ๋ก ์ํ์ค-ํฌ-์ํ์ค(SEQUENCE-TO-SEQUENCE) ๋ชจ๋ธ๋งํ๊ธฐ</a> ํํ ๋ฆฌ์ผ์
ํ์ฅํ์ด๋ฉฐ ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๊ฐ ์ด๋ป๊ฒ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ ์ฐ์ด๋์ง ์ฆ๋ช
ํ๊ธฐ ์ํด ์ด์ ํํ ๋ฆฌ์ผ์์์
๋ชจ๋ธ ๊ท๋ชจ๋ฅผ ์ฆ๊ฐ์์ผฐ์ต๋๋ค.</p>
<p>์ ์๊ณผ๋ชฉ(Prerequisites):</p>
<blockquote>
<div><ul class="simple">
<li><p><a class="reference external" href="https://pytorch.org/docs/stable/pipeline.html">Pipeline Parallelism</a></p></li>
<li><p><a class="reference external" href="https://tutorials.pytorch.kr/beginner/transformer_tutorial.html">NN.TRANSFORMER ์ TORCHTEXT ๋ก ์ํ์ค-ํฌ-์ํ์ค(SEQUENCE-TO-SEQUENCE) ๋ชจ๋ธ๋งํ๊ธฐ</a></p></li>
</ul>
</div></blockquote>
<div class="section" id="id3">
<h2>๋ชจ๋ธ ์ ์ํ๊ธฐ<a class="headerlink" href="#id3" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>์ด๋ฒ ํํ ๋ฆฌ์ผ์์๋, ํธ๋์คํฌ๋จธ ๋ชจ๋ธ์ ๋ ๊ฐ์ GPU์ ๊ฑธ์ณ์ ๋๋๊ณ ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํ์ต์์ผ ๋ณด๊ฒ ์ต๋๋ค.
๋ชจ๋ธ์ ๋ฐ๋ก <a class="reference external" href="https://tutorials.pytorch.kr/beginner/transformer_tutorial.html">NN.TRANSFORMER ์ TORCHTEXT ๋ก ์ํ์ค-ํฌ-์ํ์ค(SEQUENCE-TO-SEQUENCE) ๋ชจ๋ธ๋งํ๊ธฐ</a> ํํ ๋ฆฌ์ผ๊ณผ
๋๊ฐ์ ๋ชจ๋ธ์ด์ง๋ง ๋ ๋จ๊ณ๋ก ๋๋ฉ๋๋ค. ๋๋ถ๋ถ ํ๋ผ๋ฏธํฐ(parameter)๋ค์
<a class="reference external" href="https://pytorch.org/docs/stable/generated/torch.nn.TransformerEncoder.html">nn.TransformerEncoder</a> ๊ณ์ธต(layer)์ ํฌํจ๋ฉ๋๋ค.
<a class="reference external" href="https://pytorch.org/docs/stable/generated/torch.nn.TransformerEncoder.html">nn.TransformerEncoder</a> ๋
<a class="reference external" href="https://pytorch.org/docs/stable/generated/torch.nn.TransformerEncoderLayer.html">nn.TransformerEncoderLayer</a> ์ <code class="docutils literal notranslate"><span class="pre">nlayers</span></code> ๋ก ๊ตฌ์ฑ๋์ด ์์ต๋๋ค.
๊ฒฐ๊ณผ์ ์ผ๋ก, ์ฐ๋ฆฌ๋ <code class="docutils literal notranslate"><span class="pre">nn.TransformerEncoder</span></code> ์ ์ค์ ์ ๋๊ณ ์์ผ๋ฉฐ,
<code class="docutils literal notranslate"><span class="pre">nn.TransformerEncoderLayer</span></code> ์ ์ ๋ฐ์ ํ GPU์ ๋๊ณ
๋๋จธ์ง ์ ๋ฐ์ ๋ค๋ฅธ GPU์ ์๋๋ก ๋ชจ๋ธ์ ๋ถํ ํฉ๋๋ค. ์ด๋ฅผ ์ํด์ <code class="docutils literal notranslate"><span class="pre">Encoder</span></code> ์
<code class="docutils literal notranslate"><span class="pre">Decoder</span></code> ์น์
์ ๋ถ๋ฆฌ๋ ๋ชจ๋๋ก ๋นผ๋ธ ๋ค์, ์๋ณธ ํธ๋์คํฌ๋จธ ๋ชจ๋์
๋ํ๋ด๋ nn.Sequential์ ๋น๋ ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">sys</span>
<span class="kn">import</span> <span class="nn">math</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">import</span> <span class="nn">torch.nn</span> <span class="k">as</span> <span class="nn">nn</span>
<span class="kn">import</span> <span class="nn">torch.nn.functional</span> <span class="k">as</span> <span class="nn">F</span>
<span class="kn">import</span> <span class="nn">tempfile</span>
<span class="kn">from</span> <span class="nn">torch.nn</span> <span class="kn">import</span> <span class="n">TransformerEncoder</span><span class="p">,</span> <span class="n">TransformerEncoderLayer</span>
<span class="k">if</span> <span class="n">sys</span><span class="o">.</span><span class="n">platform</span> <span class="o">==</span> <span class="s1">'win32'</span><span class="p">:</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Windows platform is not supported for pipeline parallelism'</span><span class="p">)</span>
<span class="n">sys</span><span class="o">.</span><span class="n">exit</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="k">if</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">device_count</span><span class="p">()</span> <span class="o"><</span> <span class="mi">2</span><span class="p">:</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Need at least two GPU devices for this tutorial'</span><span class="p">)</span>
<span class="n">sys</span><span class="o">.</span><span class="n">exit</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span>
<span class="k">class</span> <span class="nc">Encoder</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">ntoken</span><span class="p">,</span> <span class="n">ninp</span><span class="p">,</span> <span class="n">dropout</span><span class="o">=</span><span class="mf">0.5</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">Encoder</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">pos_encoder</span> <span class="o">=</span> <span class="n">PositionalEncoding</span><span class="p">(</span><span class="n">ninp</span><span class="p">,</span> <span class="n">dropout</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">encoder</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Embedding</span><span class="p">(</span><span class="n">ntoken</span><span class="p">,</span> <span class="n">ninp</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">ninp</span> <span class="o">=</span> <span class="n">ninp</span>
<span class="bp">self</span><span class="o">.</span><span class="n">init_weights</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">init_weights</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="n">initrange</span> <span class="o">=</span> <span class="mf">0.1</span>
<span class="bp">self</span><span class="o">.</span><span class="n">encoder</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">uniform_</span><span class="p">(</span><span class="o">-</span><span class="n">initrange</span><span class="p">,</span> <span class="n">initrange</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">src</span><span class="p">):</span>
<span class="c1"># ์ธ์ฝ๋๋ก (S, N) ํฌ๋งท์ด ํ์ํฉ๋๋ค.</span>
<span class="n">src</span> <span class="o">=</span> <span class="n">src</span><span class="o">.</span><span class="n">t</span><span class="p">()</span>
<span class="n">src</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">encoder</span><span class="p">(</span><span class="n">src</span><span class="p">)</span> <span class="o">*</span> <span class="n">math</span><span class="o">.</span><span class="n">sqrt</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">ninp</span><span class="p">)</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">pos_encoder</span><span class="p">(</span><span class="n">src</span><span class="p">)</span>
<span class="k">class</span> <span class="nc">Decoder</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">ntoken</span><span class="p">,</span> <span class="n">ninp</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">Decoder</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">decoder</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="n">ninp</span><span class="p">,</span> <span class="n">ntoken</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">init_weights</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">init_weights</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="n">initrange</span> <span class="o">=</span> <span class="mf">0.1</span>
<span class="bp">self</span><span class="o">.</span><span class="n">decoder</span><span class="o">.</span><span class="n">bias</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">zero_</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">decoder</span><span class="o">.</span><span class="n">weight</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">uniform_</span><span class="p">(</span><span class="o">-</span><span class="n">initrange</span><span class="p">,</span> <span class="n">initrange</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">inp</span><span class="p">):</span>
<span class="c1"># ํ์ดํ๋ผ์ธ ๊ฒฐ๊ณผ๋ฌผ์ ์ํด ๋จผ์ ๋ฐฐ์น ์ฐจ์ ํ์ํฉ๋๋ค.</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">decoder</span><span class="p">(</span><span class="n">inp</span><span class="p">)</span><span class="o">.</span><span class="n">permute</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span>
</pre></div>
</div>
<p><code class="docutils literal notranslate"><span class="pre">PositionalEncoding</span></code>ย ๋ชจ๋์ ์ํ์ค ์์์ ํ ํฐ์ ์๋์ ์ธ ๋๋ ์ ๋์ ์ธ ํฌ์ง์
์ ๋ํ ์ ๋ณด๋ฅผ ์ฃผ์
ํฉ๋๋ค.
ํฌ์ง์
๋ ์ธ์ฝ๋ฉ์ ์๋ฒ ๋ฉ๊ณผ ํฉ์น ์ ์๋๋ก ๋๊ฐ์ ์ฐจ์์ ๊ฐ์ง๋๋ค. ์ฌ๊ธฐ์
๋ค๋ฅธ ์ฃผ๊ธฐ(frequency)์ย <code class="docutils literal notranslate"><span class="pre">sine</span></code>ย ๊ณผย <code class="docutils literal notranslate"><span class="pre">cosine</span></code>ย ํจ์๋ฅผ ์ฌ์ฉํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">PositionalEncoding</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="n">dropout</span><span class="o">=</span><span class="mf">0.1</span><span class="p">,</span> <span class="n">max_len</span><span class="o">=</span><span class="mi">5000</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">PositionalEncoding</span><span class="p">,</span> <span class="bp">self</span><span class="p">)</span><span class="o">.</span><span class="fm">__init__</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">dropout</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Dropout</span><span class="p">(</span><span class="n">p</span><span class="o">=</span><span class="n">dropout</span><span class="p">)</span>
<span class="n">pe</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">zeros</span><span class="p">(</span><span class="n">max_len</span><span class="p">,</span> <span class="n">d_model</span><span class="p">)</span>
<span class="n">position</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">max_len</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">float</span><span class="p">)</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">1</span><span class="p">)</span>
<span class="n">div_term</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">d_model</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span><span class="o">.</span><span class="n">float</span><span class="p">()</span> <span class="o">*</span> <span class="p">(</span><span class="o">-</span><span class="n">math</span><span class="o">.</span><span class="n">log</span><span class="p">(</span><span class="mf">10000.0</span><span class="p">)</span> <span class="o">/</span> <span class="n">d_model</span><span class="p">))</span>
<span class="n">pe</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">sin</span><span class="p">(</span><span class="n">position</span> <span class="o">*</span> <span class="n">div_term</span><span class="p">)</span>
<span class="n">pe</span><span class="p">[:,</span> <span class="mi">1</span><span class="p">::</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cos</span><span class="p">(</span><span class="n">position</span> <span class="o">*</span> <span class="n">div_term</span><span class="p">)</span>
<span class="n">pe</span> <span class="o">=</span> <span class="n">pe</span><span class="o">.</span><span class="n">unsqueeze</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span><span class="o">.</span><span class="n">transpose</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">register_buffer</span><span class="p">(</span><span class="s1">'pe'</span><span class="p">,</span> <span class="n">pe</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">forward</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">x</span><span class="p">):</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">x</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">pe</span><span class="p">[:</span><span class="n">x</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">),</span> <span class="p">:]</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">dropout</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
</pre></div>
</div>
</div>
<div class="section" id="id4">
<h2>๋ฐ์ดํฐ ๋ก๋ํ๊ณ ๋ฐฐ์น ๋ง๋ค๊ธฐ<a class="headerlink" href="#id4" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>ํ์ต ํ๋ก์ธ์ค๋ <code class="docutils literal notranslate"><span class="pre">torchtext</span></code> ์ Wikitext-2 ๋ฐ์ดํฐ์
์ ์ฌ์ฉํฉ๋๋ค.
torchtext ๋ฐ์ดํฐ์
์ ์ ๊ทผํ๊ธฐ ์ ์, <a class="reference external" href="https://github.com/pytorch/data">https://github.com/pytorch/data</a> ์ ์ฐธ๊ณ ํ์ฌ torchdata๋ฅผ ์ค์นํ์๊ธฐ ๋ฐ๋๋๋ค.</p>
<p>๋จ์ด ์ค๋ธ์ ํธ๋ ํ๋ จ ๋ฐ์ดํฐ์
์ผ๋ก ๋ง๋ค์ด์ง๊ณ , ํ ํฐ์ ํ
์(tensor)๋ก ์์นํํ๋๋ฐ ์ฌ์ฉ๋ฉ๋๋ค.
์ํ์ค ๋ฐ์ดํฐ๋ก๋ถํฐ ์์ํ์ฌ, <code class="docutils literal notranslate"><span class="pre">batchify()</span></code> ํจ์๋ ๋ฐ์ดํฐ์
์ ์ด(column)๋ค๋ก ์ ๋ฆฌํ๊ณ ,
<code class="docutils literal notranslate"><span class="pre">batch_size</span></code> ์ฌ์ด์ฆ์ ๋ฐฐ์น๋ค๋ก ๋๋ ํ์ ๋จ์ ๋ชจ๋ ํ ํฐ์ ๋ฒ๋ฆฝ๋๋ค.
์๋ฅผ ๋ค์ด, ์ํ๋ฒณ์ ์ํ์ค(์ด ๊ธธ์ด 26)๋ก ์๊ฐํ๊ณ ๋ฐฐ์น ์ฌ์ด์ฆ๋ฅผ 4๋ผ๊ณ ํ๋ค๋ฉด,
์ํ๋ฒณ์ ๊ธธ์ด๊ฐ 6์ธ 4๊ฐ์ ์ํ์ค๋ก ๋๋ ์ ์์ต๋๋ค:</p>
<div class="math">
\[\begin{bmatrix}
\text{A} & \text{B} & \text{C} & \ldots & \text{X} & \text{Y} & \text{Z}
\end{bmatrix}
\Rightarrow
\begin{bmatrix}
\begin{bmatrix}\text{A} \\ \text{B} \\ \text{C} \\ \text{D} \\ \text{E} \\ \text{F}\end{bmatrix} &
\begin{bmatrix}\text{G} \\ \text{H} \\ \text{I} \\ \text{J} \\ \text{K} \\ \text{L}\end{bmatrix} &
\begin{bmatrix}\text{M} \\ \text{N} \\ \text{O} \\ \text{P} \\ \text{Q} \\ \text{R}\end{bmatrix} &
\begin{bmatrix}\text{S} \\ \text{T} \\ \text{U} \\ \text{V} \\ \text{W} \\ \text{X}\end{bmatrix}
\end{bmatrix}\]</div>
<p>์ด ์ด๋ค์ ๋ชจ๋ธ์ ์ํด์ ๋
๋ฆฝ์ ์ผ๋ก ์ทจ๊ธ๋๋ฉฐ, ์ด๋
<code class="docutils literal notranslate"><span class="pre">G</span></code>ย ์ย <code class="docutils literal notranslate"><span class="pre">F</span></code>ย ์ ์์กด์ฑ์ด ํ์ต๋ ์ ์๋ค๋ ๊ฒ์ ์๋ฏธํ์ง๋ง, ๋ ํจ์จ์ ์ธ
๋ฐฐ์น ํ๋ก์ธ์ฑ(batch processing)์ ํ์ฉํฉ๋๋ค.</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">from</span> <span class="nn">torchtext.datasets</span> <span class="kn">import</span> <span class="n">WikiText2</span>
<span class="kn">from</span> <span class="nn">torchtext.data.utils</span> <span class="kn">import</span> <span class="n">get_tokenizer</span>
<span class="kn">from</span> <span class="nn">torchtext.vocab</span> <span class="kn">import</span> <span class="n">build_vocab_from_iterator</span>
<span class="n">train_iter</span> <span class="o">=</span> <span class="n">WikiText2</span><span class="p">(</span><span class="n">split</span><span class="o">=</span><span class="s1">'train'</span><span class="p">)</span>
<span class="n">tokenizer</span> <span class="o">=</span> <span class="n">get_tokenizer</span><span class="p">(</span><span class="s1">'basic_english'</span><span class="p">)</span>
<span class="n">vocab</span> <span class="o">=</span> <span class="n">build_vocab_from_iterator</span><span class="p">(</span><span class="nb">map</span><span class="p">(</span><span class="n">tokenizer</span><span class="p">,</span> <span class="n">train_iter</span><span class="p">),</span> <span class="n">specials</span><span class="o">=</span><span class="p">[</span><span class="s2">"<unk>"</span><span class="p">])</span>
<span class="n">vocab</span><span class="o">.</span><span class="n">set_default_index</span><span class="p">(</span><span class="n">vocab</span><span class="p">[</span><span class="s2">"<unk>"</span><span class="p">])</span>
<span class="k">def</span> <span class="nf">data_process</span><span class="p">(</span><span class="n">raw_text_iter</span><span class="p">):</span>
<span class="n">data</span> <span class="o">=</span> <span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">tensor</span><span class="p">(</span><span class="n">vocab</span><span class="p">(</span><span class="n">tokenizer</span><span class="p">(</span><span class="n">item</span><span class="p">)),</span> <span class="n">dtype</span><span class="o">=</span><span class="n">torch</span><span class="o">.</span><span class="n">long</span><span class="p">)</span> <span class="k">for</span> <span class="n">item</span> <span class="ow">in</span> <span class="n">raw_text_iter</span><span class="p">]</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">cat</span><span class="p">(</span><span class="nb">tuple</span><span class="p">(</span><span class="nb">filter</span><span class="p">(</span><span class="k">lambda</span> <span class="n">t</span><span class="p">:</span> <span class="n">t</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span> <span class="o">></span> <span class="mi">0</span><span class="p">,</span> <span class="n">data</span><span class="p">)))</span>
<span class="n">train_iter</span><span class="p">,</span> <span class="n">val_iter</span><span class="p">,</span> <span class="n">test_iter</span> <span class="o">=</span> <span class="n">WikiText2</span><span class="p">()</span>
<span class="n">train_data</span> <span class="o">=</span> <span class="n">data_process</span><span class="p">(</span><span class="n">train_iter</span><span class="p">)</span>
<span class="n">val_data</span> <span class="o">=</span> <span class="n">data_process</span><span class="p">(</span><span class="n">val_iter</span><span class="p">)</span>
<span class="n">test_data</span> <span class="o">=</span> <span class="n">data_process</span><span class="p">(</span><span class="n">test_iter</span><span class="p">)</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"</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">batchify</span><span class="p">(</span><span class="n">data</span><span class="p">,</span> <span class="n">bsz</span><span class="p">):</span>
<span class="c1"># ๋ฐ์ดํฐ์
์ ``bsz`` ํํธ๋ค๋ก ๋๋๋๋ค.</span>
<span class="n">nbatch</span> <span class="o">=</span> <span class="n">data</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">//</span> <span class="n">bsz</span>
<span class="c1"># ๊น๋ํ๊ฒ ๋๋์ด ๋จ์ด์ง์ง ์๋ ์ถ๊ฐ์ ์ธ ๋ถ๋ถ(๋๋จธ์ง)์ ์๋ผ๋
๋๋ค.</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">data</span><span class="o">.</span><span class="n">narrow</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="n">nbatch</span> <span class="o">*</span> <span class="n">bsz</span><span class="p">)</span>
<span class="c1"># ๋ฐ์ดํฐ๋ฅผ ``bsz`` ๋ฐฐ์น๋ค๋ก ๋์ผํ๊ฒ ๋๋๋๋ค.</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">data</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="n">bsz</span><span class="p">,</span> <span class="o">-</span><span class="mi">1</span><span class="p">)</span><span class="o">.</span><span class="n">t</span><span class="p">()</span><span class="o">.</span><span class="n">contiguous</span><span class="p">()</span>
<span class="k">return</span> <span class="n">data</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">)</span>
<span class="n">batch_size</span> <span class="o">=</span> <span class="mi">20</span>
<span class="n">eval_batch_size</span> <span class="o">=</span> <span class="mi">10</span>
<span class="n">train_data</span> <span class="o">=</span> <span class="n">batchify</span><span class="p">(</span><span class="n">train_data</span><span class="p">,</span> <span class="n">batch_size</span><span class="p">)</span>
<span class="n">val_data</span> <span class="o">=</span> <span class="n">batchify</span><span class="p">(</span><span class="n">val_data</span><span class="p">,</span> <span class="n">eval_batch_size</span><span class="p">)</span>
<span class="n">test_data</span> <span class="o">=</span> <span class="n">batchify</span><span class="p">(</span><span class="n">test_data</span><span class="p">,</span> <span class="n">eval_batch_size</span><span class="p">)</span>
</pre></div>
</div>
<div class="sphx-glr-script-out highlight-pytb notranslate"><div class="highlight"><pre><span></span><span class="gt">Traceback (most recent call last):</span>
File <span class="nb">"/workspace/tutorials-kr/intermediate_source/pipeline_tutorial.py"</span>, line <span class="m">156</span>, in <span class="n"><module></span>
<span class="w"> </span><span class="n">vocab</span> <span class="o">=</span> <span class="n">build_vocab_from_iterator</span><span class="p">(</span><span class="nb">map</span><span class="p">(</span><span class="n">tokenizer</span><span class="p">,</span> <span class="n">train_iter</span><span class="p">),</span> <span class="n">specials</span><span class="o">=</span><span class="p">[</span><span class="s2">"<unk>"</span><span class="p">])</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchtext/vocab/vocab_factory.py"</span>, line <span class="m">98</span>, in <span class="n">build_vocab_from_iterator</span>
<span class="w"> </span><span class="k">for</span> <span class="n">tokens</span> <span class="ow">in</span> <span class="n">iterator</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/sharding.py"</span>, line <span class="m">76</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">item</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">):</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combinatorics.py"</span>, line <span class="m">127</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/plain_text_reader.py"</span>, line <span class="m">134</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">path</span><span class="p">,</span> <span class="n">file</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/fileopener.py"</span>, line <span class="m">66</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">get_file_binaries_from_pathnames</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">encoding</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/utils/common.py"</span>, line <span class="m">217</span>, in <span class="n">get_file_binaries_from_pathnames</span>
<span class="w"> </span><span class="k">for</span> <span class="n">pathname</span> <span class="ow">in</span> <span class="n">pathnames</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">55</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">dp</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">55</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">dp</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/cacheholder.py"</span>, line <span class="m">455</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">filename</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/saver.py"</span>, line <span class="m">53</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">filepath</span><span class="p">,</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/callable.py"</span>, line <span class="m">124</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/callable.py"</span>, line <span class="m">124</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/selecting.py"</span>, line <span class="m">72</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/ziparchiveloader.py"</span>, line <span class="m">53</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/fileopener.py"</span>, line <span class="m">66</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">get_file_binaries_from_pathnames</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">mode</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">encoding</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/utils/common.py"</span>, line <span class="m">217</span>, in <span class="n">get_file_binaries_from_pathnames</span>
<span class="w"> </span><span class="k">for</span> <span class="n">pathname</span> <span class="ow">in</span> <span class="n">pathnames</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/cacheholder.py"</span>, line <span class="m">230</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/cacheholder.py"</span>, line <span class="m">375</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">item</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">152</span>, in <span class="n">__next__</span>
<span class="w"> </span><span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_get_next</span><span class="p">()</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">140</span>, in <span class="n">_get_next</span>
<span class="w"> </span><span class="n">result</span> <span class="o">=</span> <span class="nb">next</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">iterator</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">468</span>, in <span class="n">get_next_element_by_instance</span>
<span class="w"> </span><span class="k">yield</span> <span class="bp">self</span><span class="o">.</span><span class="n">_find_next</span><span class="p">(</span><span class="n">instance_id</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">437</span>, in <span class="n">_find_next</span>
<span class="w"> </span><span class="n">value</span> <span class="o">=</span> <span class="nb">next</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_datapipe_iterator</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">55</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">dp</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/combining.py"</span>, line <span class="m">55</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield from</span> <span class="n">dp</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/cacheholder.py"</span>, line <span class="m">455</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">filename</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/saver.py"</span>, line <span class="m">53</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">filepath</span><span class="p">,</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/util/hashchecker.py"</span>, line <span class="m">67</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">file_name</span><span class="p">,</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">source_datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/callable.py"</span>, line <span class="m">124</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/iter/callable.py"</span>, line <span class="m">124</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">for</span> <span class="n">data</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">datapipe</span><span class="p">:</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torch/utils/data/datapipes/_hook_iterator.py"</span>, line <span class="m">181</span>, in <span class="n">wrap_generator</span>
<span class="w"> </span><span class="n">response</span> <span class="o">=</span> <span class="n">gen</span><span class="o">.</span><span class="n">send</span><span class="p">(</span><span class="kc">None</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/load/online.py"</span>, line <span class="m">93</span>, in <span class="n">__iter__</span>
<span class="w"> </span><span class="k">yield</span> <span class="n">_get_response_from_http</span><span class="p">(</span><span class="n">url</span><span class="p">,</span> <span class="n">timeout</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">timeout</span><span class="p">,</span> <span class="o">**</span><span class="bp">self</span><span class="o">.</span><span class="n">query_params</span><span class="p">)</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/torchdata/datapipes/iter/load/online.py"</span>, line <span class="m">40</span>, in <span class="n">_get_response_from_http</span>
<span class="w"> </span><span class="n">r</span><span class="o">.</span><span class="n">raise_for_status</span><span class="p">()</span>
File <span class="nb">"/usr/local/lib/python3.10/dist-packages/requests/models.py"</span>, line <span class="m">1024</span>, in <span class="n">raise_for_status</span>
<span class="w"> </span><span class="k">raise</span> <span class="n">HTTPError</span><span class="p">(</span><span class="n">http_error_msg</span><span class="p">,</span> <span class="n">response</span><span class="o">=</span><span class="bp">self</span><span class="p">)</span>
<span class="gr">requests.exceptions.HTTPError</span>: <span class="n">403 Client Error: Forbidden for url: https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-v1.zip</span>
<span class="x">This exception is thrown by __iter__ of HTTPReaderIterDataPipe(skip_on_error=False, source_datapipe=OnDiskCacheHolderIterDataPipe, timeout=None)</span>
</pre></div>
</div>
<div class="section" id="id5">
<h3>์
๋ ฅ๊ณผ ํ๊ฒ ์ํ์ค๋ฅผ ์์ฑํ๊ธฐ ์ํ ํจ์๋ค<a class="headerlink" href="#id5" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h3>
<p><code class="docutils literal notranslate"><span class="pre">get_batch()</span></code> ํจ์๋ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ์ ์ํ ์
๋ ฅ๊ณผ ํ๊ฒ ์ํ์ค๋ฅผ
์์ฑํฉ๋๋ค. ์ด ํจ์๋ ์์ค ๋ฐ์ดํฐ๋ฅผ <code class="docutils literal notranslate"><span class="pre">bptt</span></code> ๊ธธ์ด๋ฅผ ๊ฐ์ง ๋ฉ์ด๋ฆฌ๋ก ์ธ๋ถํํฉ๋๋ค.
์ธ์ด ๋ชจ๋ธ๋ง ๊ณผ์ ๋ฅผ ์ํด์, ๋ชจ๋ธ์ ๋ค์ ๋จ์ด์ธ <code class="docutils literal notranslate"><span class="pre">Target</span></code> ์ด ํ์ํฉ๋๋ค. ์๋ฅผ ๋ค์ด <code class="docutils literal notranslate"><span class="pre">bptt</span></code> ์ ๊ฐ์ด 2๋ผ๋ฉด,
<code class="docutils literal notranslate"><span class="pre">i</span></code> = 0 ์ผ ๋ ๋ค์์ 2 ๊ฐ ๋ณ์(Variable)๋ฅผ ์ป์ ์ ์์ต๋๋ค:</p>
<img alt="../_images/transformer_input_target.png" src="../_images/transformer_input_target.png" />
<p>๋ณ์ ๋ฉ์ด๋ฆฌ๋ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ์ย <code class="docutils literal notranslate"><span class="pre">S</span></code>ย ์ฐจ์๊ณผ ์ผ์นํ๋ 0 ์ฐจ์์ ํด๋นํฉ๋๋ค.
๋ฐฐ์น ์ฐจ์ย <code class="docutils literal notranslate"><span class="pre">N</span></code>ย ์ 1 ์ฐจ์์ ํด๋นํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">bptt</span> <span class="o">=</span> <span class="mi">25</span>
<span class="k">def</span> <span class="nf">get_batch</span><span class="p">(</span><span class="n">source</span><span class="p">,</span> <span class="n">i</span><span class="p">):</span>
<span class="n">seq_len</span> <span class="o">=</span> <span class="nb">min</span><span class="p">(</span><span class="n">bptt</span><span class="p">,</span> <span class="nb">len</span><span class="p">(</span><span class="n">source</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span> <span class="o">-</span> <span class="n">i</span><span class="p">)</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">source</span><span class="p">[</span><span class="n">i</span><span class="p">:</span><span class="n">i</span><span class="o">+</span><span class="n">seq_len</span><span class="p">]</span>
<span class="n">target</span> <span class="o">=</span> <span class="n">source</span><span class="p">[</span><span class="n">i</span><span class="o">+</span><span class="mi">1</span><span class="p">:</span><span class="n">i</span><span class="o">+</span><span class="mi">1</span><span class="o">+</span><span class="n">seq_len</span><span class="p">]</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">)</span>
<span class="c1"># ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ฅผ ์ํด ๋จผ์ ๋ฐฐ์น ์ฐจ์์ด ํ์ํฉ๋๋ค.</span>
<span class="k">return</span> <span class="n">data</span><span class="o">.</span><span class="n">t</span><span class="p">(),</span> <span class="n">target</span>
</pre></div>
</div>
</div>
</div>
<div class="section" id="id6">
<h2>๋ชจ๋ธ ๊ท๋ชจ์ ํ์ดํ ์ด๊ธฐํ<a class="headerlink" href="#id6" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ฅผ ํ์ฉํ ๋ํ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ ์ฆ๋ช
ํ๊ธฐ ์ํด์
ํธ๋์คํฌ๋จธ ๊ณ์ธต ๊ท๋ชจ๋ฅผ ์ ์ ํ ํ์ฅ์ํต๋๋ค. 4096์ฐจ์์ ์๋ฒ ๋ฉ ๋ฒกํฐ, 4096์ ์๋ ์ฌ์ด์ฆ,
16๊ฐ์ ์ดํ
์
ํค๋(attention head)์ ์ด 12 ๊ฐ์ ํธ๋์คํฌ๋จธ ๊ณ์ธต
(<code class="docutils literal notranslate"><span class="pre">nn.TransformerEncoderLayer</span></code>)๋ฅผ ์ฌ์ฉํฉ๋๋ค. ์ด๋ ์ต๋
<strong>1.4์ต</strong> ๊ฐ์ ํ๋ผ๋ฏธํฐ๋ฅผ ๊ฐ๋ ๋ชจ๋ธ์ ์์ฑํฉ๋๋ค.</p>
<p>Pipe๋ <a class="reference external" href="https://pytorch.org/docs/stable/rpc.html#rref">RRef</a> ๋ฅผ ํตํด
<a class="reference external" href="https://pytorch.org/docs/stable/rpc.html">RPC ํ๋ ์์ํฌ</a> ์ ์์กดํ๋๋ฐ
์ด๋ ํฅํ ํธ์คํธ ํ์ดํ๋ผ์ธ์ ๊ต์ฐจ ํ์ฅํ ์ ์๋๋ก ํ๊ธฐ ๋๋ฌธ์
RPC ํ๋ ์์ํฌ๋ฅผ ์ด๊ธฐํํด์ผ ํฉ๋๋ค. ์ด๋ RPC ํ๋ ์์ํฌ๋ ์ค์ง ํ๋์ ํ๋์ worker๋ก ์ด๊ธฐํ๋ฅผ ํด์ผ ํ๋๋ฐ,
์ฌ๋ฌ GPU๋ฅผ ๋ค๋ฃจ๊ธฐ ์ํด ํ๋ก์ธ์ค ํ๋๋ง ์ฌ์ฉํ๊ณ ์๊ธฐ ๋๋ฌธ์
๋๋ค.</p>
<p>๊ทธ๋ฐ ๋ค์ ํ์ดํ๋ผ์ธ์ ํ GPU์ 8๊ฐ์ ํธ๋์คํฌ๋จธ์
๋ค๋ฅธ GPU์ 8๊ฐ์ ํธ๋์คํฌ๋จธ ๋ ์ด์ด๋ก ์ด๊ธฐํ๋ฉ๋๋ค.</p>
<div class="admonition note">
<p class="admonition-title">์ฐธ๊ณ </p>
<p>ํจ์จ์ฑ์ ์ํด <code class="docutils literal notranslate"><span class="pre">Pipe</span></code> ์ ์ ๋ฌ๋ <code class="docutils literal notranslate"><span class="pre">nn.Sequential</span></code> ์ด
์ค์ง ๋ ๊ฐ์ ์์(2๊ฐ์ GPU)๋ก๋ง ๊ตฌ์ฑ๋๋๋ก ํฉ๋๋ค. ์ด๋ ๊ฒ ํ๋ฉด
Pipe๊ฐ ๋ ๊ฐ์ ํํฐ์
์์๋ง ์๋ํ๊ณ
ํํฐ์
๊ฐ ์ค๋ฒํค๋๋ฅผ ํผํ ์ ์์ต๋๋ค.</p>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">ntokens</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">vocab</span><span class="p">)</span> <span class="c1"># ๋จ์ด ์ฌ์ (์ดํ์ง)์ ํฌ๊ธฐ</span>
<span class="n">emsize</span> <span class="o">=</span> <span class="mi">4096</span> <span class="c1"># ์๋ฒ ๋ฉ ์ฐจ์</span>
<span class="n">nhid</span> <span class="o">=</span> <span class="mi">4096</span> <span class="c1"># ``nn.TransformerEncoder`` ์์ ์์ ํ(feedforward) ์ ๊ฒฝ๋ง ๋ชจ๋ธ์ ์ฐจ์</span>
<span class="n">nlayers</span> <span class="o">=</span> <span class="mi">12</span> <span class="c1"># ``nn.TransformerEncoder`` ๋ด๋ถ์ ``nn.TransformerEncoderLayer`` ๊ฐ์</span>
<span class="n">nhead</span> <span class="o">=</span> <span class="mi">16</span> <span class="c1"># Multihead Attention ๋ชจ๋ธ์ ํค๋ ๊ฐ์</span>
<span class="n">dropout</span> <span class="o">=</span> <span class="mf">0.2</span> <span class="c1"># dropout ๊ฐ</span>
<span class="kn">from</span> <span class="nn">torch.distributed</span> <span class="kn">import</span> <span class="n">rpc</span>
<span class="n">tmpfile</span> <span class="o">=</span> <span class="n">tempfile</span><span class="o">.</span><span class="n">NamedTemporaryFile</span><span class="p">()</span>
<span class="n">rpc</span><span class="o">.</span><span class="n">init_rpc</span><span class="p">(</span>
<span class="n">name</span><span class="o">=</span><span class="s2">"worker"</span><span class="p">,</span>
<span class="n">rank</span><span class="o">=</span><span class="mi">0</span><span class="p">,</span>
<span class="n">world_size</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="n">rpc_backend_options</span><span class="o">=</span><span class="n">rpc</span><span class="o">.</span><span class="n">TensorPipeRpcBackendOptions</span><span class="p">(</span>
<span class="n">init_method</span><span class="o">=</span><span class="s2">"file://</span><span class="si">{}</span><span class="s2">"</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">tmpfile</span><span class="o">.</span><span class="n">name</span><span class="p">),</span>
<span class="c1"># _transports์ _channels๋ฅผ ์ง์ ํ๋ ๊ฒ์ด ํด๊ฒฐ ๋ฐฉ๋ฒ์ด๋ฉฐ</span>
<span class="c1"># PyTorch ๋ฒ์ >= 1.8.1 ์์๋ _transports์ _channels๋ฅผ</span>
<span class="c1"># ์ง์ ํ์ง ์์๋ ๋ฉ๋๋ค.</span>
<span class="n">_transports</span><span class="o">=</span><span class="p">[</span><span class="s2">"ibv"</span><span class="p">,</span> <span class="s2">"uv"</span><span class="p">],</span>
<span class="n">_channels</span><span class="o">=</span><span class="p">[</span><span class="s2">"cuda_ipc"</span><span class="p">,</span> <span class="s2">"cuda_basic"</span><span class="p">],</span>
<span class="p">)</span>
<span class="p">)</span>
<span class="n">num_gpus</span> <span class="o">=</span> <span class="mi">2</span>
<span class="n">partition_len</span> <span class="o">=</span> <span class="p">((</span><span class="n">nlayers</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="n">num_gpus</span><span class="p">)</span> <span class="o">+</span> <span class="mi">1</span>
<span class="c1"># ์ฒ์์ ์ธ์ฝ๋๋ฅผ ์ถ๊ฐํฉ๋๋ค.</span>
<span class="n">tmp_list</span> <span class="o">=</span> <span class="p">[</span><span class="n">Encoder</span><span class="p">(</span><span class="n">ntokens</span><span class="p">,</span> <span class="n">emsize</span><span class="p">,</span> <span class="n">dropout</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="mi">0</span><span class="p">)]</span>
<span class="n">module_list</span> <span class="o">=</span> <span class="p">[]</span>
<span class="c1"># ํ์ํ ๋ชจ๋ ํธ๋์คํฌ๋จธ ๋ธ๋ก๋ค์ ์ถ๊ฐํฉ๋๋ค.</span>
<span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="n">nlayers</span><span class="p">):</span>
<span class="n">transformer_block</span> <span class="o">=</span> <span class="n">TransformerEncoderLayer</span><span class="p">(</span><span class="n">emsize</span><span class="p">,</span> <span class="n">nhead</span><span class="p">,</span> <span class="n">nhid</span><span class="p">,</span> <span class="n">dropout</span><span class="p">)</span>
<span class="k">if</span> <span class="n">i</span> <span class="o">!=</span> <span class="mi">0</span> <span class="ow">and</span> <span class="n">i</span> <span class="o">%</span> <span class="p">(</span><span class="n">partition_len</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>
<span class="n">module_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">tmp_list</span><span class="p">))</span>
<span class="n">tmp_list</span> <span class="o">=</span> <span class="p">[]</span>
<span class="n">device</span> <span class="o">=</span> <span class="n">i</span> <span class="o">//</span> <span class="p">(</span><span class="n">partition_len</span><span class="p">)</span>
<span class="n">tmp_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">transformer_block</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="p">))</span>
<span class="c1"># ๋ง์ง๋ง์ ๋์ฝ๋๋ฅผ ์ถ๊ฐํฉ๋๋ค.</span>
<span class="n">tmp_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">Decoder</span><span class="p">(</span><span class="n">ntokens</span><span class="p">,</span> <span class="n">emsize</span><span class="p">)</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="n">num_gpus</span> <span class="o">-</span> <span class="mi">1</span><span class="p">))</span>
<span class="n">module_list</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">tmp_list</span><span class="p">))</span>
<span class="kn">from</span> <span class="nn">torch.distributed.pipeline.sync</span> <span class="kn">import</span> <span class="n">Pipe</span>
<span class="c1"># ํ์ดํ๋ผ์ธ์ ๋น๋ํฉ๋๋ค.</span>
<span class="n">chunks</span> <span class="o">=</span> <span class="mi">8</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">Pipe</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Sequential</span><span class="p">(</span><span class="o">*</span><span class="n">module_list</span><span class="p">),</span> <span class="n">chunks</span> <span class="o">=</span> <span class="n">chunks</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">get_total_params</span><span class="p">(</span><span class="n">module</span><span class="p">:</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">):</span>
<span class="n">total_params</span> <span class="o">=</span> <span class="mi">0</span>
<span class="k">for</span> <span class="n">param</span> <span class="ow">in</span> <span class="n">module</span><span class="o">.</span><span class="n">parameters</span><span class="p">():</span>
<span class="n">total_params</span> <span class="o">+=</span> <span class="n">param</span><span class="o">.</span><span class="n">numel</span><span class="p">()</span>
<span class="k">return</span> <span class="n">total_params</span>
<span class="nb">print</span> <span class="p">(</span><span class="s1">'Total parameters in model: </span><span class="si">{:,}</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">get_total_params</span><span class="p">(</span><span class="n">model</span><span class="p">)))</span>
</pre></div>
</div>
</div>
<div class="section" id="id7">
<h2>๋ชจ๋ธ ์คํํ๊ธฐ<a class="headerlink" href="#id7" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>์์ค(loss)์ ์ถ์ ํ๊ธฐ ์ํด <a class="reference external" href="https://pytorch.org/docs/master/nn.html?highlight=crossentropyloss#torch.nn.CrossEntropyLoss">CrossEntropyLoss</a> ๊ฐ
์ ์ฉ๋๋ฉฐ, ์ตํฐ๋ง์ด์ (optimizer)๋ก์ <a class="reference external" href="https://pytorch.org/docs/master/optim.html?highlight=sgd#torch.optim.SGD">SGD</a>
๋ ํ๋ฅ ์ ๊ฒฝ์ฌํ๊ฐ๋ฒ(stochastic gradient descent method)์ ๊ตฌํํฉ๋๋ค. ์ด๊ธฐ
ํ์ต๋ฅ (learning rate)์ 5.0๋ก ์ค์ ๋ฉ๋๋ค. <a class="reference external" href="https://pytorch.org/docs/master/optim.html?highlight=steplr#torch.optim.lr_scheduler.StepLR">StepLR</a> ๋
์ํญ(epoch)์ ๋ฐ๋ผ์ ํ์ต๋ฅ ์ ์กฐ์ ํ๋ ๋ฐ ์ฌ์ฉ๋ฉ๋๋ค. ํ์ตํ๋ ๋์,
๊ธฐ์ธ๊ธฐ ํญ๋ฐ(gradient exploding)์ ๋ฐฉ์งํ๊ธฐ ์ํด ๋ชจ๋ ๊ธฐ์ธ๊ธฐ๋ฅผ ํจ๊ป ์กฐ์ (scale)ํ๋ ํจ์
<a class="reference external" href="https://pytorch.org/docs/master/nn.html?highlight=nn%20utils%20clip_grad_norm#torch.nn.utils.clip_grad_norm_">nn.utils.clip_grad_norm_</a>
์ ์ด์ฉํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">criterion</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="n">lr</span> <span class="o">=</span> <span class="mf">5.0</span> <span class="c1"># ํ์ต๋ฅ </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="n">lr</span><span class="p">)</span>
<span class="n">scheduler</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">lr_scheduler</span><span class="o">.</span><span class="n">StepLR</span><span class="p">(</span><span class="n">optimizer</span><span class="p">,</span> <span class="mf">1.0</span><span class="p">,</span> <span class="n">gamma</span><span class="o">=</span><span class="mf">0.95</span><span class="p">)</span>
<span class="kn">import</span> <span class="nn">time</span>
<span class="k">def</span> <span class="nf">train</span><span class="p">():</span>
<span class="n">model</span><span class="o">.</span><span class="n">train</span><span class="p">()</span> <span class="c1"># ํ๋ จ ๋ชจ๋๋ก ์ ํ</span>
<span class="n">total_loss</span> <span class="o">=</span> <span class="mf">0.</span>
<span class="n">start_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span>
<span class="n">ntokens</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">vocab</span><span class="p">)</span>
<span class="c1"># ์คํฌ๋ฆฝํธ ์คํ ์๊ฐ์ ์งง๊ฒ ์ ์งํ๊ธฐ ์ํด์ 50 ๋ฐฐ์น๋ง ํ์ตํฉ๋๋ค.</span>
<span class="n">nbatches</span> <span class="o">=</span> <span class="nb">min</span><span class="p">(</span><span class="mi">50</span> <span class="o">*</span> <span class="n">bptt</span><span class="p">,</span> <span class="n">train_data</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span>
<span class="k">for</span> <span class="n">batch</span><span class="p">,</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="nb">range</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">nbatches</span><span class="p">,</span> <span class="n">bptt</span><span class="p">)):</span>
<span class="n">data</span><span class="p">,</span> <span class="n">targets</span> <span class="o">=</span> <span class="n">get_batch</span><span class="p">(</span><span class="n">train_data</span><span class="p">,</span> <span class="n">i</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="c1"># Pipe๋ ๋จ์ผ ํธ์คํธ ๋ด์ ์๊ณ </span>
<span class="c1"># forward ๋ฉ์๋๋ก ๋ฐํ๋ ``RRef`` ํ๋ก์ธ์ค๋ ์ด ๋
ธ๋์ ๊ตญํ๋์ด ์๊ธฐ ๋๋ฌธ์</span>
<span class="c1"># ``RRef.local_value()`` ๋ฅผ ํตํด ๊ฐ๋จํ ์ฐพ์ ์ ์์ต๋๋ค.</span>
<span class="n">output</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span><span class="o">.</span><span class="n">local_value</span><span class="p">()</span>
<span class="c1"># ํ๊ฒ์ ํ์ดํ๋ผ์ธ ์ถ๋ ฅ์ด ์๋</span>
<span class="c1"># ์ฅ์น๋ก ์ฎ๊ฒจ์ผํฉ๋๋ค.</span>
<span class="n">loss</span> <span class="o">=</span> <span class="n">criterion</span><span class="p">(</span><span class="n">output</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">ntokens</span><span class="p">),</span> <span class="n">targets</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="mi">1</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">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">clip_grad_norm_</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="mf">0.5</span><span class="p">)</span>
<span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
<span class="n">total_loss</span> <span class="o">+=</span> <span class="n">loss</span><span class="o">.</span><span class="n">item</span><span class="p">()</span>
<span class="n">log_interval</span> <span class="o">=</span> <span class="mi">10</span>
<span class="k">if</span> <span class="n">batch</span> <span class="o">%</span> <span class="n">log_interval</span> <span class="o">==</span> <span class="mi">0</span> <span class="ow">and</span> <span class="n">batch</span> <span class="o">></span> <span class="mi">0</span><span class="p">:</span>
<span class="n">cur_loss</span> <span class="o">=</span> <span class="n">total_loss</span> <span class="o">/</span> <span class="n">log_interval</span>
<span class="n">elapsed</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span> <span class="o">-</span> <span class="n">start_time</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'| epoch </span><span class="si">{:3d}</span><span class="s1"> | </span><span class="si">{:5d}</span><span class="s1">/</span><span class="si">{:5d}</span><span class="s1"> batches | '</span>
<span class="s1">'lr </span><span class="si">{:02.2f}</span><span class="s1"> | ms/batch </span><span class="si">{:5.2f}</span><span class="s1"> | '</span>
<span class="s1">'loss </span><span class="si">{:5.2f}</span><span class="s1"> | ppl </span><span class="si">{:8.2f}</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
<span class="n">epoch</span><span class="p">,</span> <span class="n">batch</span><span class="p">,</span> <span class="n">nbatches</span> <span class="o">//</span> <span class="n">bptt</span><span class="p">,</span> <span class="n">scheduler</span><span class="o">.</span><span class="n">get_lr</span><span class="p">()[</span><span class="mi">0</span><span class="p">],</span>
<span class="n">elapsed</span> <span class="o">*</span> <span class="mi">1000</span> <span class="o">/</span> <span class="n">log_interval</span><span class="p">,</span>
<span class="n">cur_loss</span><span class="p">,</span> <span class="n">math</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="n">cur_loss</span><span class="p">)))</span>
<span class="n">total_loss</span> <span class="o">=</span> <span class="mi">0</span>
<span class="n">start_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">evaluate</span><span class="p">(</span><span class="n">eval_model</span><span class="p">,</span> <span class="n">data_source</span><span class="p">):</span>
<span class="n">eval_model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span> <span class="c1"># ํ๊ฐ ๋ชจ๋๋ก ์ ํ</span>
<span class="n">total_loss</span> <span class="o">=</span> <span class="mf">0.</span>
<span class="n">ntokens</span> <span class="o">=</span> <span class="nb">len</span><span class="p">(</span><span class="n">vocab</span><span class="p">)</span>
<span class="c1"># ์คํฌ๋ฆฝํธ ์คํ ์๊ฐ์ ์งง๊ฒ ์ ์งํ๊ธฐ ์ํด 50 ๋ฐฐ์น๋ง ํ๊ฐํฉ๋๋ค.</span>
<span class="n">nbatches</span> <span class="o">=</span> <span class="nb">min</span><span class="p">(</span><span class="mi">50</span> <span class="o">*</span> <span class="n">bptt</span><span class="p">,</span> <span class="n">data_source</span><span class="o">.</span><span class="n">size</span><span class="p">(</span><span class="mi">0</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span>
<span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">no_grad</span><span class="p">():</span>
<span class="k">for</span> <span class="n">i</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">nbatches</span><span class="p">,</span> <span class="n">bptt</span><span class="p">):</span>
<span class="n">data</span><span class="p">,</span> <span class="n">targets</span> <span class="o">=</span> <span class="n">get_batch</span><span class="p">(</span><span class="n">data_source</span><span class="p">,</span> <span class="n">i</span><span class="p">)</span>
<span class="n">output</span> <span class="o">=</span> <span class="n">eval_model</span><span class="p">(</span><span class="n">data</span><span class="p">)</span><span class="o">.</span><span class="n">local_value</span><span class="p">()</span>
<span class="n">output_flat</span> <span class="o">=</span> <span class="n">output</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="n">ntokens</span><span class="p">)</span>
<span class="c1"># ํ๊ฒ์ ํ์ดํ๋ผ์ธ ์ถ๋ ฅ์ด ์๋</span>
<span class="c1"># ์ฅ์น๋ก ์ฎ๊ฒจ์ผํฉ๋๋ค.</span>
<span class="n">total_loss</span> <span class="o">+=</span> <span class="nb">len</span><span class="p">(</span><span class="n">data</span><span class="p">)</span> <span class="o">*</span> <span class="n">criterion</span><span class="p">(</span><span class="n">output_flat</span><span class="p">,</span> <span class="n">targets</span><span class="o">.</span><span class="n">cuda</span><span class="p">(</span><span class="mi">1</span><span class="p">))</span><span class="o">.</span><span class="n">item</span><span class="p">()</span>
<span class="k">return</span> <span class="n">total_loss</span> <span class="o">/</span> <span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">data_source</span><span class="p">)</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span>
</pre></div>
</div>
<p>์ํญ์ ๋ฐ๋ณตํฉ๋๋ค. ๋ง์ฝ ๊ฒ์ฆ ์ค์ฐจ(validation loss)๊ฐ ์ง๊ธ๊น์ง ๊ด์ฐฐํ ๊ฒ ์ค ์ต์ ์ด๋ผ๋ฉด
๋ชจ๋ธ์ ์ ์ฅํฉ๋๋ค. ๊ฐ ์ํญ ์ดํ์ ํ์ต๋ฅ ์ ์กฐ์ ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">best_val_loss</span> <span class="o">=</span> <span class="nb">float</span><span class="p">(</span><span class="s2">"inf"</span><span class="p">)</span>
<span class="n">epochs</span> <span class="o">=</span> <span class="mi">3</span> <span class="c1"># ์ํญ ์</span>
<span class="n">best_model</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">for</span> <span class="n">epoch</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="n">epochs</span> <span class="o">+</span> <span class="mi">1</span><span class="p">):</span>
<span class="n">epoch_start_time</span> <span class="o">=</span> <span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span>
<span class="n">train</span><span class="p">()</span>
<span class="n">val_loss</span> <span class="o">=</span> <span class="n">evaluate</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">val_data</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'-'</span> <span class="o">*</span> <span class="mi">89</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'| end of epoch </span><span class="si">{:3d}</span><span class="s1"> | time: </span><span class="si">{:5.2f}</span><span class="s1">s | valid loss </span><span class="si">{:5.2f}</span><span class="s1"> | '</span>
<span class="s1">'valid ppl </span><span class="si">{:8.2f}</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">epoch</span><span class="p">,</span> <span class="p">(</span><span class="n">time</span><span class="o">.</span><span class="n">time</span><span class="p">()</span> <span class="o">-</span> <span class="n">epoch_start_time</span><span class="p">),</span>
<span class="n">val_loss</span><span class="p">,</span> <span class="n">math</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="n">val_loss</span><span class="p">)))</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'-'</span> <span class="o">*</span> <span class="mi">89</span><span class="p">)</span>
<span class="k">if</span> <span class="n">val_loss</span> <span class="o"><</span> <span class="n">best_val_loss</span><span class="p">:</span>
<span class="n">best_val_loss</span> <span class="o">=</span> <span class="n">val_loss</span>
<span class="n">best_model</span> <span class="o">=</span> <span class="n">model</span>
<span class="n">scheduler</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
</pre></div>
</div>
</div>
<div class="section" id="id8">
<h2>ํ๊ฐ ๋ฐ์ดํฐ์
์ผ๋ก ๋ชจ๋ธ ํ๊ฐํ๊ธฐ<a class="headerlink" href="#id8" title="์ด ์ ๋ชฉ์ ๋ํ ํผ๋จธ๋งํฌ">ยถ</a></h2>
<p>ํ๊ฐ ๋ฐ์ดํฐ์
์์์ ๊ฒฐ๊ณผ๋ฅผ ํ์ธํ๊ธฐ ์ํด ์ต๊ณ ์ ๋ชจ๋ธ์ ์ ์ฉํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">test_loss</span> <span class="o">=</span> <span class="n">evaluate</span><span class="p">(</span><span class="n">best_model</span><span class="p">,</span> <span class="n">test_data</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'='</span> <span class="o">*</span> <span class="mi">89</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'| End of training | test loss </span><span class="si">{:5.2f}</span><span class="s1"> | test ppl </span><span class="si">{:8.2f}</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
<span class="n">test_loss</span><span class="p">,</span> <span class="n">math</span><span class="o">.</span><span class="n">exp</span><span class="p">(</span><span class="n">test_loss</span><span class="p">)))</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'='</span> <span class="o">*</span> <span class="mi">89</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 0 minutes 0.564 seconds)</p>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-intermediate-pipeline-tutorial-py">
<div class="sphx-glr-download sphx-glr-download-python docutils container">
<p><a class="reference download internal" download="" href="../_downloads/b4afbcfb1c1ac5f5cd7da108c2236f09/pipeline_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">pipeline_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/4cefa4723023eb5d85ed047dadc7f491/pipeline_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">pipeline_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>
<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="#">ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ</a><ul>
<li><a class="reference internal" href="#id3">๋ชจ๋ธ ์ ์ํ๊ธฐ</a></li>
<li><a class="reference internal" href="#id4">๋ฐ์ดํฐ ๋ก๋ํ๊ณ ๋ฐฐ์น ๋ง๋ค๊ธฐ</a><ul>
<li><a class="reference internal" href="#id5">์
๋ ฅ๊ณผ ํ๊ฒ ์ํ์ค๋ฅผ ์์ฑํ๊ธฐ ์ํ ํจ์๋ค</a></li>
</ul>
</li>