-
Notifications
You must be signed in to change notification settings - Fork 257
/
Copy pathpruning_tutorial.html
1443 lines (1147 loc) ยท 98.9 KB
/
pruning_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="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ(Pruning) ํํ ๋ฆฌ์ผ — PyTorch Tutorials 1.10.2+cu102 documentation</title>
<link rel="stylesheet" href="../_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="../_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="../_static/copybutton.css" type="text/css" />
<link rel="stylesheet" href="../_static/gallery.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.10.0-beta/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.13.11/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="../_static/katex-math.css" type="text/css" />
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="(๋ฒ ํ) LSTM ๊ธฐ๋ฐ ๋จ์ด ๋จ์ ์ธ์ด ๋ชจ๋ธ์ ๋์ ์์ํ" href="../advanced/dynamic_quantization_tutorial.html" />
<link rel="prev" title="Parametrizations Tutorial" href="parametrizations.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">
1.10.2+cu102
</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"><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"><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"><span class="caption-text">Introduction to PyTorch on YouTube</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt.html">Introduction to PyTorch - YouTube Series</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/introyt1_tutorial.html">Introduction to PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/introyt/tensors_deeper_tutorial.html">Introduction to PyTorch Tensors</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"><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"><span class="caption-text">์ด๋ฏธ์ง/๋น๋์ค</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="torchvision_tutorial.html">TorchVision ๊ฐ์ฒด ๊ฒ์ถ ๋ฏธ์ธ์กฐ์ (Finetuning) ํํ ๋ฆฌ์ผ</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 ์ํ ๋น์ ํธ๋์คํฌ๋จธ(Vision Transformer) ๋ชจ๋ธ ์ต์ ํํ๊ธฐ</a></li>
</ul>
<p class="caption"><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="speech_command_classification_with_torchaudio_tutorial.html">Speech Command Classification with torchaudio</a></li>
<li class="toctree-l1"><a class="reference internal" href="text_to_speech_with_torchaudio.html">Text-to-speech with torchaudio</a></li>
<li class="toctree-l1"><a class="reference internal" href="forced_alignment_with_torchaudio_tutorial.html">Forced Alignment with Wav2Vec2</a></li>
</ul>
<p class="caption"><span class="caption-text">ํ
์คํธ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/transformer_tutorial.html">nn.Transformer ์ TorchText ๋ก ์ํ์ค-ํฌ-์ํ์ค(Sequence-to-Sequence) ๋ชจ๋ธ๋งํ๊ธฐ</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">nn.Transformer์ torchtext๋ก ์ธ์ด ๋ฒ์ญํ๊ธฐ</a></li>
</ul>
<p class="caption"><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="mario_rl_tutorial.html">Train a Mario-playing RL Agent</a></li>
</ul>
<p class="caption"><span class="caption-text">PyTorch ๋ชจ๋ธ์ ํ๋ก๋์
ํ๊ฒฝ์ ๋ฐฐํฌํ๊ธฐ</span></p>
<ul>
<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>
</ul>
<p class="caption"><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"><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="../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"><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>
</ul>
<p class="caption"><span class="caption-text">๋ชจ๋ธ ์ต์ ํ</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="../beginner/profiler.html">PyTorch ๋ชจ๋ ํ๋กํ์ผ๋ง ํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensorboard_profiler_tutorial.html">PyTorch Profiler With TensorBoard</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/hyperparameter_tuning_tutorial.html">Hyperparameter tuning with Ray Tune</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/vt_tutorial.html">๋ฐฐํฌ๋ฅผ ์ํ ๋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 current"><a class="current reference internal" href="#">๊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">(beta) Static Quantization with Eager Mode in PyTorch</a></li>
</ul>
<p class="caption"><span class="caption-text">๋ณ๋ ฌ ๋ฐ ๋ถ์ฐ ํ์ต</span></p>
<ul>
<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="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="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="pipeline_tutorial.html">ํ์ดํ๋ผ์ธ ๋ณ๋ ฌํ๋ก ํธ๋์คํฌ๋จธ ๋ชจ๋ธ ํ์ต์ํค๊ธฐ</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"><span class="caption-text">Mobile</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/deeplabv3_on_ios.html">iOS์์์ ์ด๋ฏธ์ง ๋ถํ DeepLapV3</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/deeplabv3_on_android.html">์๋๋ก์ด๋์์์ ์ด๋ฏธ์ง ๋ถํ DeepLapV3</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>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ(Pruning) ํํ ๋ฆฌ์ผ</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/intermediate/pruning_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/pruning_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="first admonition-title">Note</p>
<p class="last">Click <a class="reference internal" href="#sphx-glr-download-intermediate-pruning-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="pruning">
<span id="sphx-glr-intermediate-pruning-tutorial-py"></span><h1>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ(Pruning) ํํ ๋ฆฌ์ผ<a class="headerlink" href="#pruning" title="Permalink to this headline">ยถ</a></h1>
<p><strong>์ ์</strong>: <a class="reference external" href="https://github.com/mickypaganini">Michela Paganini</a>
<strong>๋ฒ์ญ</strong> : <a class="reference external" href="https://github.com/Justin-A">์์์ค</a></p>
<p>์ต์ฒจ๋จ ๋ฅ๋ฌ๋ ๋ชจ๋ธ๋ค์ ๊ต์ฅํ ๋ง์ ์์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค๋ก ๊ตฌ์ฑ๋๊ธฐ ๋๋ฌธ์, ์ฝ๊ฒ ๋ฐฐํฌ๋๊ธฐ ์ด๋ ต์ต๋๋ค.
์ด์ ๋ฐ๋๋ก, ์๋ฌผํ์ ์ ๊ฒฝ๋ง๋ค์ ํจ์จ์ ์ผ๋ก ํฌ์ํ๊ฒ ์ฐ๊ฒฐ๋ ๊ฒ์ผ๋ก ์๋ ค์ ธ ์์ต๋๋ค.
๋ชจ๋ธ์ ์ ํ๋๊ฐ ์์๋์ง ์๋ ๋ฒ์์์ ๋ฉ๋ชจ๋ฆฌ, ๋ฐฐํฐ๋ฆฌ, ํ๋์จ์ด ์๋น๋์ ์ค์ด๊ณ ,
๊ธฐ๊ธฐ์ ๊ฒฝ๋ํ๋ ๋ชจ๋ธ์ ๋ฐฐ์นํ๋ฉฐ, ๊ฐ์ธ์ด ์ด์ฉํ๊ณ ์๋ ๊ธฐ๊ธฐ์์ ํ๋ผ์ด๋ฒ์๊ฐ ๋ณด์ฅ๋๊ธฐ ์ํด์๋
๋ชจ๋ธ์ ํฌํจ๋ ํ๋ผ๋ฏธํฐ ์๋ฅผ ์ค์ฌ ์์ถํ๋ ์ต์ ์ ๊ธฐ๋ฒ์ ํ์
ํ๋ ๊ฒ์ด ์ค์ํฉ๋๋ค.
์ฐ๊ตฌ ์ธก๋ฉด์์๋, ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ๊ต์ฅํ ๋ง์ ์์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค๋ก ๊ตฌ์ฑ๋ ๋ชจ๋ธ๊ณผ
๊ต์ฅํ ์ ์ ์์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค๋ก ๊ตฌ์ฑ๋ ๋ชจ๋ธ ๊ฐ ํ์ต ์ญํ ์ฐจ์ด๋ฅผ ์กฐ์ฌํ๋๋ฐ ์ฃผ๋ก ์ด์ฉ๋๊ธฐ๋ ํ๋ฉฐ,
ํ์ ์ ๊ฒฝ๋ง ๋ชจ๋ธ๊ณผ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ ์ด๊ธฐํ๊ฐ ์ด์ด ์ข๊ฒ ์ ๋ ์ผ์ด์ค๋ฅผ ๋ฐํ์ผ๋ก
(โ<a class="reference external" href="https://arxiv.org/abs/1803.03635">lottery tickets</a>โ) ์ ๊ฒฝ๋ง ๊ตฌ์กฐ๋ฅผ ์ฐพ๋ ๊ธฐ์ ๋ค์ ๋ํด ๋ฐ๋ ์๊ฒฌ์ ์ ์ํ๊ธฐ๋ ํฉ๋๋ค.</p>
<p>์ด๋ฒ ํํ ๋ฆฌ์ผ์์๋, <code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune</span></code> ์ ์ด์ฉํ์ฌ ์ฌ๋ฌ๋ถ์ด ์ค๊ณํ ๋ฅ๋ฌ๋ ๋ชจ๋ธ์ ๋ํด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํด๋ณด๋ ๊ฒ์ ๋ฐฐ์๋ณด๊ณ ,
์ฌํ์ ์ผ๋ก ์ฌ๋ฌ๋ถ์ ๋ง์ถคํ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ๊ตฌํํ๋ ๋ฐฉ๋ฒ์ ๋ํด ๋ฐฐ์๋ณด๋๋ก ํ๊ฒ ์ต๋๋ค.</p>
<div class="section" id="id2">
<h2>์๊ตฌ์ฌํญ<a class="headerlink" href="#id2" title="Permalink to this headline">ยถ</a></h2>
<p><code class="docutils literal notranslate"><span class="pre">"torch>=1.4"</span></code></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">torch</span> <span class="kn">import</span> <span class="n">nn</span>
<span class="kn">import</span> <span class="nn">torch.nn.utils.prune</span> <span class="k">as</span> <span class="nn">prune</span>
<span class="kn">import</span> <span class="nn">torch.nn.functional</span> <span class="k">as</span> <span class="nn">F</span>
</pre></div>
</div>
</div>
<div class="section" id="id3">
<h2>๋ฅ๋ฌ๋ ๋ชจ๋ธ ์์ฑ<a class="headerlink" href="#id3" title="Permalink to this headline">ยถ</a></h2>
<p>์ด๋ฒ ํํ ๋ฆฌ์ผ์์๋, ์ ๋ฅด์ฟค ๊ต์๋์ ์ฐ๊ตฌ์ง๋ค์ด 1998๋
๋์ ๋ฐํํ <code class="docutils literal notranslate"><span class="pre">LeNet</span>
<span class="pre"><http://yann.lecun.com/exdb/publis/pdf/lecun-98.pdf></span></code> ์ ๋ชจ๋ธ ๊ตฌ์กฐ๋ฅผ ์ด์ฉํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">device</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">device</span><span class="p">(</span><span class="s2">"cuda"</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">is_available</span><span class="p">()</span> <span class="k">else</span> <span class="s2">"cpu"</span><span class="p">)</span>
<span class="k">class</span> <span class="nc">LeNet</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="nb">super</span><span class="p">(</span><span class="n">LeNet</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="c1"># 1๊ฐ ์ฑ๋ ์์ ์ด๋ฏธ์ง๋ฅผ ์
๋ ฅ๊ฐ์ผ๋ก ์ด์ฉํ์ฌ 6๊ฐ ์ฑ๋ ์์ ์ถ๋ ฅ๊ฐ์ ๊ณ์ฐํ๋ ๋ฐฉ์</span>
<span class="c1"># Convolution ์ฐ์ฐ์ ์งํํ๋ ์ปค๋(ํํฐ)์ ํฌ๊ธฐ๋ 3x3 ์ ์ด์ฉ</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv1</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">6</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv2</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="mi">6</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">3</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc1</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="mi">16</span> <span class="o">*</span> <span class="mi">5</span> <span class="o">*</span> <span class="mi">5</span><span class="p">,</span> <span class="mi">120</span><span class="p">)</span> <span class="c1"># Convolution ์ฐ์ฐ ๊ฒฐ๊ณผ 5x5 ํฌ๊ธฐ์ 16 ์ฑ๋ ์์ ์ด๋ฏธ์ง</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc2</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="mi">120</span><span class="p">,</span> <span class="mi">84</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fc3</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="mi">84</span><span class="p">,</span> <span class="mi">10</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">F</span><span class="o">.</span><span class="n">max_pool2d</span><span class="p">(</span><span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv1</span><span class="p">(</span><span class="n">x</span><span class="p">)),</span> <span class="p">(</span><span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">))</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">max_pool2d</span><span class="p">(</span><span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">conv2</span><span class="p">(</span><span class="n">x</span><span class="p">)),</span> <span class="mi">2</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="n">view</span><span class="p">(</span><span class="o">-</span><span class="mi">1</span><span class="p">,</span> <span class="nb">int</span><span class="p">(</span><span class="n">x</span><span class="o">.</span><span class="n">nelement</span><span class="p">()</span> <span class="o">/</span> <span class="n">x</span><span class="o">.</span><span class="n">shape</span><span class="p">[</span><span class="mi">0</span><span class="p">]))</span>
<span class="n">x</span> <span class="o">=</span> <span class="n">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">fc1</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">F</span><span class="o">.</span><span class="n">relu</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">fc2</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="bp">self</span><span class="o">.</span><span class="n">fc3</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">return</span> <span class="n">x</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">LeNet</span><span class="p">()</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">device</span><span class="o">=</span><span class="n">device</span><span class="p">)</span>
</pre></div>
</div>
</div>
<div class="section" id="id4">
<h2>๋ชจ๋ ์ ๊ฒ<a class="headerlink" href="#id4" title="Permalink to this headline">ยถ</a></h2>
<p>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋์ง ์์ LeNet ๋ชจ๋ธ์ <code class="docutils literal notranslate"><span class="pre">conv1</span></code> ์ธต์ ์ ๊ฒํด๋ด
์๋ค.
์ฌ๊ธฐ์๋ 2๊ฐ์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ธ <code class="docutils literal notranslate"><span class="pre">๊ฐ์ค์น</span></code> ๊ฐ๊ณผ <code class="docutils literal notranslate"><span class="pre">ํธํฅ</span></code> ๊ฐ์ด ํฌํจ๋ ๊ฒ์ด๋ฉฐ, ๋ฒํผ๋ ์กด์ฌํ์ง ์์ ๊ฒ์
๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">module</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">conv1</span>
<span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight', Parameter containing:
tensor([[[[-0.1599, 0.1967, 0.0294],
[-0.1911, 0.2040, 0.2047],
[-0.1330, -0.1417, -0.0624]]],
[[[-0.1863, 0.0249, 0.0698],
[-0.0324, -0.1226, 0.2558],
[ 0.1365, -0.1993, 0.3318]]],
[[[ 0.2656, 0.1135, -0.1520],
[ 0.1406, 0.1702, 0.0586],
[ 0.2396, -0.0773, 0.1192]]],
[[[-0.1009, -0.1442, -0.3307],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.1494, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0791, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.1659],
[ 0.0390, -0.3244, -0.0939]]]], device='cuda:0', requires_grad=True)), ('bias', Parameter containing:
tensor([-0.2589, 0.2286, 0.0883, -0.2710, -0.2214, -0.1360], device='cuda:0',
requires_grad=True))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[]
</pre></div>
</div>
</div>
<div class="section" id="id5">
<h2>๋ชจ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ ์ ์ฉ ์์ <a class="headerlink" href="#id5" title="Permalink to this headline">ยถ</a></h2>
<p>๋ชจ๋์ ๋ํด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๊ธฐ ์ํด (์ด๋ฒ ์์ ์์๋, LeNet ๋ชจ๋ธ์ <code class="docutils literal notranslate"><span class="pre">conv1</span></code> ์ธต)
์ฒซ ๋ฒ์งธ๋ก๋, <code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune</span></code> (๋๋ <code class="docutils literal notranslate"><span class="pre">BasePruningMethod</span></code> ์ ์๋ธ ํด๋์ค๋ก ์ง์
<a class="reference external" href="torch-nn-utils-prune">๊ตฌํ</a> )
๋ด ์กด์ฌํ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ํํฉ๋๋ค.
๊ทธ ํ, ํด๋น ๋ชจ๋ ๋ด์์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๊ณ ์ ํ๋ ๋ชจ๋๊ณผ ํ๋ผ๋ฏธํฐ๋ฅผ ์ง์ ํฉ๋๋ค.
๋ง์ง๋ง์ผ๋ก, ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ๋นํ ํค์๋ ์ธ์๊ฐ์ ์ด์ฉํ์ฌ ๊ฐ์ง์น๊ธฐ ๋งค๊ฐ๋ณ์๋ฅผ ์ง์ ํฉ๋๋ค.
์ด๋ฒ ์์ ์์๋, <code class="docutils literal notranslate"><span class="pre">conv1</span></code> ์ธต์ ๊ฐ์ค์น์ 30%๊ฐ๋ค์ ๋๋ค์ผ๋ก ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํด๋ณด๊ฒ ์ต๋๋ค.
๋ชจ๋์ ํจ์์ ๋ํ ์ฒซ ๋ฒ์งธ ์ธ์๊ฐ์ผ๋ก ์ ๋ฌ๋๋ฉฐ, <code class="docutils literal notranslate"><span class="pre">name</span></code> ์ ๋ฌธ์์ด ์๋ณ์๋ฅผ ์ด์ฉํ์ฌ ํด๋น ๋ชจ๋ ๋ด ๋งค๊ฐ๋ณ์๋ฅผ ๊ตฌ๋ถํฉ๋๋ค.
๊ทธ๋ฆฌ๊ณ , <code class="docutils literal notranslate"><span class="pre">amount</span></code> ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๊ธฐ ์ํ ๋์ ๊ฐ์ค์น๊ฐ๋ค์ ๋ฐฑ๋ถ์จ (0๊ณผ 1์ฌ์ด์ ์ค์๊ฐ),
ํน์ ๊ฐ์ค์น๊ฐ์ ์ฐ๊ฒฐ์ ๊ฐ์ (์์๊ฐ ์๋ ์ ์) ๋ฅผ ์ง์ ํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">prune</span><span class="o">.</span><span class="n">random_unstructured</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s2">"weight"</span><span class="p">,</span> <span class="n">amount</span><span class="o">=</span><span class="mf">0.3</span><span class="p">)</span>
</pre></div>
</div>
<p>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ๊ฐ์ค์น๊ฐ๋ค์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค๋ก๋ถํฐ ์ ๊ฑฐํ๊ณ <code class="docutils literal notranslate"><span class="pre">weight_orig</span></code> (์ฆ, ์ด๊ธฐ ๊ฐ์ค์น ์ด๋ฆ์ โ_origโ์ ๋ถ์ธ) ์ด๋ผ๋
์๋ก์ด ํ๋ผ๋ฏธํฐ๊ฐ์ผ๋ก ๋์ฒดํ๋ ๊ฒ์ผ๋ก ์คํ๋ฉ๋๋ค.
<code class="docutils literal notranslate"><span class="pre">weight_orig</span></code> ์ ํ
์๊ฐ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋์ง ์์ ์ํ๋ฅผ ์ ์ฅํฉ๋๋ค.
<code class="docutils literal notranslate"><span class="pre">bias</span></code> ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋์ง ์์๊ธฐ ๋๋ฌธ์ ๊ทธ๋๋ก ๋จ์ ์์ต๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('bias', Parameter containing:
tensor([-0.2589, 0.2286, 0.0883, -0.2710, -0.2214, -0.1360], device='cuda:0',
requires_grad=True)), ('weight_orig', Parameter containing:
tensor([[[[-0.1599, 0.1967, 0.0294],
[-0.1911, 0.2040, 0.2047],
[-0.1330, -0.1417, -0.0624]]],
[[[-0.1863, 0.0249, 0.0698],
[-0.0324, -0.1226, 0.2558],
[ 0.1365, -0.1993, 0.3318]]],
[[[ 0.2656, 0.1135, -0.1520],
[ 0.1406, 0.1702, 0.0586],
[ 0.2396, -0.0773, 0.1192]]],
[[[-0.1009, -0.1442, -0.3307],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.1494, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0791, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.1659],
[ 0.0390, -0.3244, -0.0939]]]], device='cuda:0', requires_grad=True))]
</pre></div>
</div>
<p>์์์ ์ ํํ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ํด ์์ฑ๋๋ ๊ฐ์ง์น๊ธฐ ๋ง์คํฌ๋ ์ด๊ธฐ ํ๋ผ๋ฏธํฐ <code class="docutils literal notranslate"><span class="pre">name</span></code> ์ <code class="docutils literal notranslate"><span class="pre">weight_mask</span></code>
(์ฆ, ์ด๊ธฐ ๊ฐ์ค์น ์ด๋ฆ์ โ_maskโ๋ฅผ ๋ถ์ธ) ์ด๋ฆ์ ๋ชจ๋ ๋ฒํผ๋ก ์ ์ฅ๋ฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight_mask', tensor([[[[0., 1., 0.],
[1., 0., 0.],
[1., 1., 1.]]],
[[[1., 0., 1.],
[1., 1., 0.],
[1., 0., 1.]]],
[[[1., 0., 0.],
[0., 1., 1.],
[1., 1., 1.]]],
[[[1., 0., 0.],
[1., 1., 1.],
[1., 1., 1.]]],
[[[1., 0., 1.],
[1., 1., 1.],
[0., 1., 1.]]],
[[[1., 1., 1.],
[1., 1., 0.],
[1., 1., 0.]]]], device='cuda:0'))]
</pre></div>
</div>
<p>์์ ์ด ๋์ง ์์ ์ํ์์ ์์ ํ๋ฅผ ์งํํ๊ธฐ ์ํด์๋ <code class="docutils literal notranslate"><span class="pre">๊ฐ์ค์น</span></code> ๊ฐ ์์ฑ์ด ์กด์ฌํด์ผ ํฉ๋๋ค.
<code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune</span></code> ๋ด ๊ตฌํ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ๊ฐ์ค์น๊ฐ๋ค์ ์ด์ฉํ์ฌ
(๊ธฐ์กด์ ๊ฐ์ค์น๊ฐ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋) ์์ ํ๋ฅผ ์งํํ๊ณ , <code class="docutils literal notranslate"><span class="pre">weight</span></code> ์์ฑ๊ฐ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ๊ฐ์ค์น๊ฐ๋ค์ ์ ์ฅํฉ๋๋ค.
์ด์ ๊ฐ์ค์น๊ฐ๋ค์ <code class="docutils literal notranslate"><span class="pre">module</span></code> ์ ๋งค๊ฐ๋ณ์๊ฐ ์๋๋ผ ํ๋์ ์์ฑ๊ฐ์ผ๋ก ์ทจ๊ธ๋๋ ์ ์ ์ฃผ์ํ์ธ์.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">weight</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>tensor([[[[-0.0000, 0.1967, 0.0000],
[-0.1911, 0.0000, 0.0000],
[-0.1330, -0.1417, -0.0624]]],
[[[-0.1863, 0.0000, 0.0698],
[-0.0324, -0.1226, 0.0000],
[ 0.1365, -0.0000, 0.3318]]],
[[[ 0.2656, 0.0000, -0.0000],
[ 0.0000, 0.1702, 0.0586],
[ 0.2396, -0.0773, 0.1192]]],
[[[-0.1009, -0.0000, -0.0000],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.0000, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0000, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.0000],
[ 0.0390, -0.3244, -0.0000]]]], device='cuda:0',
grad_fn=<MulBackward0>)
</pre></div>
</div>
<p>์ต์ข
์ ์ผ๋ก, ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ํ์ดํ ์น์ <code class="docutils literal notranslate"><span class="pre">forward_pre_hooks</span></code> ๋ฅผ ์ด์ฉํ์ฌ ๊ฐ ์์ ํ๊ฐ ์งํ๋๊ธฐ ์ ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ฉ๋๋ค.
๊ตฌ์ฒด์ ์ผ๋ก, ์ง๊ธ๊น์ง ์งํํ ๊ฒ ์ฒ๋ผ, ๋ชจ๋์ด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋์์ ๋,
๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ๊ฐ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ด <code class="docutils literal notranslate"><span class="pre">forward_pre_hook</span></code> ๋ฅผ ์ป๊ฒ๋ฉ๋๋ค.
์ด๋ฌํ ๊ฒฝ์ฐ, <code class="docutils literal notranslate"><span class="pre">weight</span></code> ์ด๋ฆ์ธ ๊ธฐ์กด ํ๋ผ๋ฏธํฐ๊ฐ์ ๋ํด์๋ง ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ์๊ธฐ ๋๋ฌธ์,
ํ
์ ์ค์ง 1๊ฐ๋ง ์กด์ฌํ ๊ฒ์
๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">_forward_pre_hooks</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>OrderedDict([(0, <torch.nn.utils.prune.RandomUnstructured object at 0x7fe6b0f1d730>)])
</pre></div>
</div>
<p>์๊ฒฐ์ฑ์ ์ํด, ํธํฅ๊ฐ์ ๋ํด์๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ ์ ์์ผ๋ฉฐ,
๋ชจ๋์ ํ๋ผ๋ฏธํฐ, ๋ฒํผ, ํ
, ์์ฑ๊ฐ๋ค์ด ์ด๋ป๊ฒ ๋ณ๊ฒฝ๋๋์ง ํ์ธํ ์ ์์ต๋๋ค.
๋ ๋ค๋ฅธ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํด๋ณด๊ธฐ ์ํด, <code class="docutils literal notranslate"><span class="pre">l1_unstructured</span></code> ๊ฐ์ง์น๊ธฐ ํจ์์์ ๊ตฌํ๋ ๋ด์ฉ๊ณผ ๊ฐ์ด,
L1 Norm ๊ฐ์ด ๊ฐ์ฅ ์์ ํธํฅ๊ฐ 3๊ฐ๋ฅผ ๊ฐ์ง์น๊ธฐ๋ฅผ ์๋ํด๋ด
์๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">prune</span><span class="o">.</span><span class="n">l1_unstructured</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s2">"bias"</span><span class="p">,</span> <span class="n">amount</span><span class="o">=</span><span class="mi">3</span><span class="p">)</span>
</pre></div>
</div>
<p>์ด์ ์์ ์ค์ตํ ๋ด์ฉ์ ํ ๋๋ก, ๋ช
๋ช
๋ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ด <code class="docutils literal notranslate"><span class="pre">weight_orig</span></code>, <code class="docutils literal notranslate"><span class="pre">bias_orig</span></code> 2๊ฐ๋ฅผ ๋ชจ๋ ํฌํจํ ๊ฒ์ด๋ผ ์์ํ ์ ์์ต๋๋ค.
๋ฒํผ๋ค์ <code class="docutils literal notranslate"><span class="pre">weight_mask</span></code>, <code class="docutils literal notranslate"><span class="pre">bias_mask</span></code> 2๊ฐ๋ฅผ ํฌํจํ ๊ฒ์
๋๋ค.
๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ 2๊ฐ์ ํ
์๊ฐ๋ค์ ๋ชจ๋์ ์์ฑ๊ฐ์ผ๋ก ์กด์ฌํ ๊ฒ์ด๋ฉฐ, ๋ชจ๋์ 2๊ฐ์ <code class="docutils literal notranslate"><span class="pre">forward_pre_hooks</span></code> ์ ๊ฐ๊ฒ ๋ ๊ฒ์
๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight_orig', Parameter containing:
tensor([[[[-0.1599, 0.1967, 0.0294],
[-0.1911, 0.2040, 0.2047],
[-0.1330, -0.1417, -0.0624]]],
[[[-0.1863, 0.0249, 0.0698],
[-0.0324, -0.1226, 0.2558],
[ 0.1365, -0.1993, 0.3318]]],
[[[ 0.2656, 0.1135, -0.1520],
[ 0.1406, 0.1702, 0.0586],
[ 0.2396, -0.0773, 0.1192]]],
[[[-0.1009, -0.1442, -0.3307],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.1494, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0791, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.1659],
[ 0.0390, -0.3244, -0.0939]]]], device='cuda:0', requires_grad=True)), ('bias_orig', Parameter containing:
tensor([-0.2589, 0.2286, 0.0883, -0.2710, -0.2214, -0.1360], device='cuda:0',
requires_grad=True))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight_mask', tensor([[[[0., 1., 0.],
[1., 0., 0.],
[1., 1., 1.]]],
[[[1., 0., 1.],
[1., 1., 0.],
[1., 0., 1.]]],
[[[1., 0., 0.],
[0., 1., 1.],
[1., 1., 1.]]],
[[[1., 0., 0.],
[1., 1., 1.],
[1., 1., 1.]]],
[[[1., 0., 1.],
[1., 1., 1.],
[0., 1., 1.]]],
[[[1., 1., 1.],
[1., 1., 0.],
[1., 1., 0.]]]], device='cuda:0')), ('bias_mask', tensor([1., 1., 0., 1., 0., 0.], device='cuda:0'))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">bias</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>tensor([-0.2589, 0.2286, 0.0000, -0.2710, -0.0000, -0.0000], device='cuda:0',
grad_fn=<MulBackward0>)
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">_forward_pre_hooks</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>OrderedDict([(0, <torch.nn.utils.prune.RandomUnstructured object at 0x7fe6b0f1d730>), (1, <torch.nn.utils.prune.L1Unstructured object at 0x7fe6b0fc61c0>)])
</pre></div>
</div>
</div>
<div class="section" id="id7">
<h2>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ ๋ฐ๋ณต ์ ์ฉ<a class="headerlink" href="#id7" title="Permalink to this headline">ยถ</a></h2>
<p>๋ชจ๋ ๋ด ๊ฐ์ ํ๋ผ๋ฏธํฐ๊ฐ์ ๋ํด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ฌ๋ฌ๋ฒ ์ ์ฉ๋ ์ ์์ผ๋ฉฐ, ๋ค์ํ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์กฐํฉ์ด ์ ์ฉ๋ ๊ฒ๊ณผ ๋์ผํ๊ฒ ์ ์ฉ๋ ์ ์์ต๋๋ค.
์๋ก์ด ๋ง์คํฌ์ ์ด์ ์ ๋ง์คํฌ์ ๊ฒฐํฉ์ <code class="docutils literal notranslate"><span class="pre">PruningContainer</span></code> ์ <code class="docutils literal notranslate"><span class="pre">compute_mask</span></code> ๋ฉ์๋๋ฅผ ํตํด ์ฒ๋ฆฌํ ์ ์์ต๋๋ค.</p>
<p>์๋ฅผ ๋ค์ด, ๋ง์ฝ <code class="docutils literal notranslate"><span class="pre">module.weight</span></code> ๊ฐ์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๊ณ ์ถ์ ๋, ํ
์์ 0๋ฒ์งธ ์ถ์ L2 norm๊ฐ์ ๊ธฐ์ค์ผ๋ก ๊ตฌ์กฐํ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํฉ๋๋ค.
(์ฌ๊ธฐ์ 0๋ฒ์งธ ์ถ์ด๋, ํฉ์ฑ๊ณฑ ์ฐ์ฐ์ ํตํด ๊ณ์ฐ๋ ์ถ๋ ฅ๊ฐ์ ๋ํด ๊ฐ ์ฑ๋๋ณ๋ก ์ ์ฉ๋๋ค๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค.)
์ด ๋ฐฉ์์ <code class="docutils literal notranslate"><span class="pre">ln_structured</span></code> ํจ์์ <code class="docutils literal notranslate"><span class="pre">n=2</span></code> ์ <code class="docutils literal notranslate"><span class="pre">dim=0</span></code> ์ ์ธ์๊ฐ์ ๋ฐํ์ผ๋ก ๊ตฌํ๋ ์ ์์ต๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">prune</span><span class="o">.</span><span class="n">ln_structured</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s2">"weight"</span><span class="p">,</span> <span class="n">amount</span><span class="o">=</span><span class="mf">0.5</span><span class="p">,</span> <span class="n">n</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span> <span class="n">dim</span><span class="o">=</span><span class="mi">0</span><span class="p">)</span>
</pre></div>
</div>
<p>์ฐ๋ฆฌ๊ฐ ํ์ธํ ์ ์๋ฏ์ด, ์ด์ ๋ง์คํฌ์ ์์ฉ์ ์ ์งํ๋ฉด์ ์ฑ๋์ 50% (6๊ฐ ์ค 3๊ฐ) ์ ํด๋น๋๋ ๋ชจ๋ ์ฐ๊ฒฐ์ 0์ผ๋ก ๋ณ๊ฒฝํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">weight</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>tensor([[[[-0.0000, 0.0000, 0.0000],
[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, -0.0000]]],
[[[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[ 0.0000, 0.0000, -0.0000],
[ 0.0000, 0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[-0.1009, -0.0000, -0.0000],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.0000, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0000, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.0000],
[ 0.0390, -0.3244, -0.0000]]]], device='cuda:0',
grad_fn=<MulBackward0>)
</pre></div>
</div>
<p>์ด์ ํด๋นํ๋ ํ
์ <code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune.PruningContainer</span></code> ํํ๋ก ์กด์ฌํ๋ฉฐ, ๊ฐ์ค์น์ ์ ์ฉ๋ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ด๋ ฅ์ ์ ์ฅํฉ๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">for</span> <span class="n">hook</span> <span class="ow">in</span> <span class="n">module</span><span class="o">.</span><span class="n">_forward_pre_hooks</span><span class="o">.</span><span class="n">values</span><span class="p">():</span>
<span class="k">if</span> <span class="n">hook</span><span class="o">.</span><span class="n">_tensor_name</span> <span class="o">==</span> <span class="s2">"weight"</span><span class="p">:</span> <span class="c1"># ๊ฐ์ค์น์ ํด๋นํ๋ ํ
์ ์ ํ</span>
<span class="k">break</span>
<span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">hook</span><span class="p">))</span> <span class="c1"># ์ปจํ
์ด๋ ๋ด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ด๋ ฅ</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[<torch.nn.utils.prune.RandomUnstructured object at 0x7fe6b0f1d730>, <torch.nn.utils.prune.LnStructured object at 0x7fe6b09969d0>]
</pre></div>
</div>
</div>
<div class="section" id="id8">
<h2>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ๋ชจ๋ธ์ ์ง๋ ฌํ<a class="headerlink" href="#id8" title="Permalink to this headline">ยถ</a></h2>
<p>๋ง์คํฌ ๋ฒํผ๋ค๊ณผ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ํ
์ ๊ณ์ฐ์ ์ฌ์ฉ๋ ๊ธฐ์กด์ ํ๋ผ๋ฏธํฐ๋ฅผ ํฌํจํ์ฌ ๊ด๋ จ๋ ๋ชจ๋ ํ
์๊ฐ๋ค์
ํ์ํ ๊ฒฝ์ฐ ๋ชจ๋ธ์ <code class="docutils literal notranslate"><span class="pre">state_dict</span></code> ์ ์ ์ฅ๋๊ธฐ ๋๋ฌธ์, ์ฝ๊ฒ ์ง๋ ฌํํ์ฌ ์ ์ฅํ ์ ์์ต๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">state_dict</span><span class="p">()</span><span class="o">.</span><span class="n">keys</span><span class="p">())</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>odict_keys(['conv1.weight_orig', 'conv1.bias_orig', 'conv1.weight_mask', 'conv1.bias_mask', 'conv2.weight', 'conv2.bias', 'fc1.weight', 'fc1.bias', 'fc2.weight', 'fc2.bias', 'fc3.weight', 'fc3.bias'])
</pre></div>
</div>
</div>
<div class="section" id="id9">
<h2>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ฌ-ํ๋ผ๋ฏธํฐํ ์ ๊ฑฐ<a class="headerlink" href="#id9" title="Permalink to this headline">ยถ</a></h2>
<p>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ๊ฒ์ ์๊ตฌ์ ์ผ๋ก ๋ง๋ค๊ธฐ ์ํด์, ์ฌ-ํ๋ผ๋ฏธํฐํ ๊ด์ ์
<code class="docutils literal notranslate"><span class="pre">weight_orig</span></code> ์ <code class="docutils literal notranslate"><span class="pre">weight_mask</span></code> ๊ฐ์ ์ ๊ฑฐํ๊ณ , <code class="docutils literal notranslate"><span class="pre">forward_pre_hook</span></code> ๊ฐ์ ์ ๊ฑฐํฉ๋๋ค.
์ ๊ฑฐํ๊ธฐ ์ํด <code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune</span></code> ๋ด <code class="docutils literal notranslate"><span class="pre">remove</span></code> ํจ์๋ฅผ ์ด์ฉํ ์ ์์ต๋๋ค.
๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋์ง ์์ ๊ฒ์ฒ๋ผ ์คํ๋๋ ๊ฒ์ด ์๋ ์ ์ ์ฃผ์ํ์ธ์.
์ด๋ ๋จ์ง ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ์ํ์์ ๊ฐ์ค์น ํ๋ผ๋ฏธํฐ๊ฐ์ ๋ชจ๋ธ ํ๋ผ๋ฏธํฐ๊ฐ์ผ๋ก ์ฌํ ๋นํ๋ ๊ฒ์ ํตํด ์๊ตฌ์ ์ผ๋ก ๋ง๋๋ ๊ฒ์ผ ๋ฟ์
๋๋ค.</p>
<p>์ฌ-ํ๋ผ๋ฏธํฐํ๋ฅผ ์ ๊ฑฐํ๊ธฐ ์ ์ํ</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight_orig', Parameter containing:
tensor([[[[-0.1599, 0.1967, 0.0294],
[-0.1911, 0.2040, 0.2047],
[-0.1330, -0.1417, -0.0624]]],
[[[-0.1863, 0.0249, 0.0698],
[-0.0324, -0.1226, 0.2558],
[ 0.1365, -0.1993, 0.3318]]],
[[[ 0.2656, 0.1135, -0.1520],
[ 0.1406, 0.1702, 0.0586],
[ 0.2396, -0.0773, 0.1192]]],
[[[-0.1009, -0.1442, -0.3307],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.1494, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0791, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.1659],
[ 0.0390, -0.3244, -0.0939]]]], device='cuda:0', requires_grad=True)), ('bias_orig', Parameter containing:
tensor([-0.2589, 0.2286, 0.0883, -0.2710, -0.2214, -0.1360], device='cuda:0',
requires_grad=True))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('weight_mask', tensor([[[[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]],
[[[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]],
[[[0., 0., 0.],
[0., 0., 0.],
[0., 0., 0.]]],
[[[1., 0., 0.],
[1., 1., 1.],
[1., 1., 1.]]],
[[[1., 0., 1.],
[1., 1., 1.],
[0., 1., 1.]]],
[[[1., 1., 1.],
[1., 1., 0.],
[1., 1., 0.]]]], device='cuda:0')), ('bias_mask', tensor([1., 1., 0., 1., 0., 0.], device='cuda:0'))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">weight</span><span class="p">)</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>tensor([[[[-0.0000, 0.0000, 0.0000],
[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, -0.0000]]],
[[[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[ 0.0000, 0.0000, -0.0000],
[ 0.0000, 0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[-0.1009, -0.0000, -0.0000],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.0000, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0000, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.0000],
[ 0.0390, -0.3244, -0.0000]]]], device='cuda:0',
grad_fn=<MulBackward0>)
</pre></div>
</div>
<p>์ฌ-ํ๋ผ๋ฏธํฐ๋ฅผ ์ ๊ฑฐํ ํ ์ํ</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">prune</span><span class="o">.</span><span class="n">remove</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_parameters</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('bias_orig', Parameter containing:
tensor([-0.2589, 0.2286, 0.0883, -0.2710, -0.2214, -0.1360], device='cuda:0',
requires_grad=True)), ('weight', Parameter containing:
tensor([[[[-0.0000, 0.0000, 0.0000],
[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, -0.0000]]],
[[[-0.0000, 0.0000, 0.0000],
[-0.0000, -0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[ 0.0000, 0.0000, -0.0000],
[ 0.0000, 0.0000, 0.0000],
[ 0.0000, -0.0000, 0.0000]]],
[[[-0.1009, -0.0000, -0.0000],
[ 0.0834, -0.3057, 0.2517],
[ 0.1163, -0.0654, -0.2401]]],
[[[ 0.1596, 0.0000, 0.0463],
[-0.1667, -0.2033, 0.2463],
[-0.0000, 0.1756, 0.0363]]],
[[[-0.0342, -0.2524, -0.0740],
[-0.2819, -0.2809, -0.0000],
[ 0.0390, -0.3244, -0.0000]]]], device='cuda:0', requires_grad=True))]
</pre></div>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="n">module</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">()))</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>[('bias_mask', tensor([1., 1., 0., 1., 0., 0.], device='cuda:0'))]
</pre></div>
</div>
</div>
<div class="section" id="id10">
<h2>๋ชจ๋ธ ๋ด ์ฌ๋ฌ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ ๋ํ์ฌ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ ์ ์ฉ<a class="headerlink" href="#id10" title="Permalink to this headline">ยถ</a></h2>
<p>๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๊ณ ์ถ์ ํ๋ผ๋ฏธํฐ๊ฐ๋ค์ ์ง์ ํจ์ผ๋ก์จ, ์ด๋ฒ ์์ ์์ ๋ณผ ์ ์๋ ๊ฒ ์ฒ๋ผ,
์ ๊ฒฝ๋ง ๋ชจ๋ธ ๋ด ์ฌ๋ฌ ํ
์๊ฐ๋ค์ ๋ํด์ ์ฝ๊ฒ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ ์ ์์ต๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">new_model</span> <span class="o">=</span> <span class="n">LeNet</span><span class="p">()</span>
<span class="k">for</span> <span class="n">name</span><span class="p">,</span> <span class="n">module</span> <span class="ow">in</span> <span class="n">new_model</span><span class="o">.</span><span class="n">named_modules</span><span class="p">():</span>
<span class="c1"># ๋ชจ๋ 2D-conv ์ธต์ 20% ์ฐ๊ฒฐ์ ๋ํด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉ</span>
<span class="k">if</span> <span class="nb">isinstance</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">Conv2d</span><span class="p">):</span>
<span class="n">prune</span><span class="o">.</span><span class="n">l1_unstructured</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">'weight'</span><span class="p">,</span> <span class="n">amount</span><span class="o">=</span><span class="mf">0.2</span><span class="p">)</span>
<span class="c1"># ๋ชจ๋ ์ ํ ์ธต์ 40% ์ฐ๊ฒฐ์ ๋ํด ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉ</span>
<span class="k">elif</span> <span class="nb">isinstance</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">Linear</span><span class="p">):</span>
<span class="n">prune</span><span class="o">.</span><span class="n">l1_unstructured</span><span class="p">(</span><span class="n">module</span><span class="p">,</span> <span class="n">name</span><span class="o">=</span><span class="s1">'weight'</span><span class="p">,</span> <span class="n">amount</span><span class="o">=</span><span class="mf">0.4</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="nb">dict</span><span class="p">(</span><span class="n">new_model</span><span class="o">.</span><span class="n">named_buffers</span><span class="p">())</span><span class="o">.</span><span class="n">keys</span><span class="p">())</span> <span class="c1"># ์กด์ฌํ๋ ๋ชจ๋ ๋ง์คํฌ๋ค์ ํ์ธ</span>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>dict_keys(['conv1.weight_mask', 'conv2.weight_mask', 'fc1.weight_mask', 'fc2.weight_mask', 'fc3.weight_mask'])
</pre></div>
</div>
</div>
<div class="section" id="id11">
<h2>์ ์ญ ๋ฒ์์ ๋ํ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ ์ ์ฉ<a class="headerlink" href="#id11" title="Permalink to this headline">ยถ</a></h2>
<p>์ง๊ธ๊น์ง, โ์ง์ญ ๋ณ์โ ์ ๋ํด์๋ง ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๋ ๋ฐฉ๋ฒ์ ์ดํด๋ณด์์ต๋๋ค.
(์ฆ, ๊ฐ์ค์น ๊ท๋ชจ, ํ์ฑํ ์ ๋, ๊ฒฝ์ฌ๊ฐ ๋ฑ์ ๊ฐ ํญ๋ชฉ์ ํต๊ณ๋์ ๋ฐํ์ผ๋ก ๋ชจ๋ธ ๋ด ํ
์๊ฐ ํ๋์ฉ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๋ ๋ฐฉ์)
๊ทธ๋ฌ๋, ๋ฒ์ฉ์ ์ด๊ณ ์๋ง ๋ ๊ฐ๋ ฅํ ๋ฐฉ๋ฒ์ ๊ฐ ์ธต์์ ๊ฐ์ฅ ๋ฎ์ 20%์ ์ฐ๊ฒฐ์ ์ ๊ฑฐํ๋๊ฒ ๋์ ์, ์ ์ฒด ๋ชจ๋ธ์ ๋ํด์ ๊ฐ์ฅ ๋ฎ์ 20% ์ฐ๊ฒฐ์ ํ๋ฒ์ ์ ๊ฑฐํ๋ ๊ฒ์
๋๋ค.
์ด๊ฒ์ ๊ฐ ์ธต์ ๋ํด์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๋ ์ฐ๊ฒฐ์ ๋ฐฑ๋ถ์จ๊ฐ์ ๋ค๋ฅด๊ฒ ๋ง๋ค ๊ฐ๋ฅ์ฑ์ด ์์ต๋๋ค.
<code class="docutils literal notranslate"><span class="pre">torch.nn.utils.prune</span></code> ๋ด <code class="docutils literal notranslate"><span class="pre">global_unstructured</span></code> ์ ์ด์ฉํ์ฌ ์ด๋ป๊ฒ ์ ์ญ ๋ฒ์์ ๋ํ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ ์ ์ฉํ๋์ง ์ดํด๋ด
์๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">model</span> <span class="o">=</span> <span class="n">LeNet</span><span class="p">()</span>
<span class="n">parameters_to_prune</span> <span class="o">=</span> <span class="p">(</span>
<span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">conv1</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">),</span>
<span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">conv2</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">),</span>
<span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">fc1</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">),</span>
<span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">fc2</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">),</span>
<span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">fc3</span><span class="p">,</span> <span class="s1">'weight'</span><span class="p">),</span>
<span class="p">)</span>
<span class="n">prune</span><span class="o">.</span><span class="n">global_unstructured</span><span class="p">(</span>
<span class="n">parameters_to_prune</span><span class="p">,</span>
<span class="n">pruning_method</span><span class="o">=</span><span class="n">prune</span><span class="o">.</span><span class="n">L1Unstructured</span><span class="p">,</span>
<span class="n">amount</span><span class="o">=</span><span class="mf">0.2</span><span class="p">,</span>
<span class="p">)</span>
</pre></div>
</div>
<p>์ด์ ๊ฐ ์ธต์ ์กด์ฌํ๋ ์ฐ๊ฒฐ๋ค์ ๊ฐ์ง์น๊ธฐ ๊ธฐ๋ฒ์ด ์ ์ฉ๋ ์ ๋๊ฐ 20%๊ฐ ์๋ ๊ฒ์ ํ์ธํ ์ ์์ต๋๋ค.
๊ทธ๋ฌ๋, ์ ์ฒด ๊ฐ์ง์น๊ธฐ ์ ์ฉ ๋ฒ์๋ ์ฝ 20%๊ฐ ๋ ๊ฒ์
๋๋ค.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span>
<span class="s2">"Sparsity in conv1.weight: </span><span class="si">{:.2f}</span><span class="s2">%"</span><span class="o">.</span><span class="n">format</span><span class="p">(</span>
<span class="mf">100.</span> <span class="o">*</span> <span class="nb">float</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">conv1</span><span class="o">.</span><span class="n">weight</span> <span class="o">==</span> <span class="mi">0</span><span class="p">))</span>