-
Notifications
You must be signed in to change notification settings - Fork 257
/
Copy pathstatic_quantization_tutorial.html
1432 lines (1177 loc) ยท 125 KB
/
static_quantization_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>(experimental) Static Quantization with Eager Mode in PyTorch — PyTorch Tutorials 1.4.0 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/gallery.css" type="text/css" />
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="(experimental) Quantized Transfer Learning for Computer Vision Tutorial" href="../intermediate/quantized_transfer_learning_tutorial.html" />
<link rel="prev" title="(experimental) Dynamic Quantization on an LSTM Word Language Model" href="dynamic_quantization_tutorial.html" />
<script src="../_static/js/modernizr.min.js"></script>
</head>
<a class="github-ribbon" href="https://github.com/9bow/PyTorch-tutorials-kr">
<img class="ribbon-img" width="149" height="149" src="https://github.blog/wp-content/uploads/2008/12/forkme_left_red_aa0000.png?resize=149%2C149" class="attachment-full size-full" alt="Fork me on GitHub" data-recalc-dims="1">
</a>
<div class="container-fluid header-holder tutorials-header" id="header-holder">
<div class="container">
<div class="header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<a href="https://pytorch.org/features">Features</a>
</li>
<li>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</li>
<li>
<a href="https://pytorch.org/blog/">Blog</a>
</li>
<li class="active">
<a href="https://pytorch.org/tutorials">Tutorials</a>
</li>
<li>
<a href="https://pytorch.org/docs/stable/index.html">Docs</a>
</li>
<li>
<a href="https://pytorch.org/resources">Resources</a>
</li>
<li>
<a href="https://github.com/pytorch/pytorch">Github</a>
</li>
</ul>
</div>
<a class="main-menu-open-button" href="#" data-behavior="open-mobile-menu"></a>
</div>
</div>
</div>
<body class="pytorch-body">
<div class="table-of-contents-link-wrapper">
<span>Table of Contents</span>
<a href="#" class="toggle-table-of-contents" data-behavior="toggle-table-of-contents"></a>
</div>
<nav data-toggle="wy-nav-shift" class="pytorch-left-menu" id="pytorch-left-menu">
<div class="pytorch-side-scroll">
<div class="pytorch-menu pytorch-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
<div class="pytorch-left-menu-search">
<div class="version">
1.4.0
</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">์์ํ๊ธฐ (Getting Started)</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/deep_learning_60min_blitz.html">ํ์ดํ ์น(PyTorch)๋ก ๋ฅ๋ฌ๋ํ๊ธฐ: 60๋ถ๋ง์ ๋์ฅ๋ด๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/data_loading_tutorial.html">์ฌ์ฉ์ ์ ์ Dataset, Dataloader, Transforms ์์ฑํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/tensorboard_tutorial.html">Visualizing Models, Data, and Training with TensorBoard</a></li>
</ul>
<p class="caption"><span class="caption-text">์ด๋ฏธ์ง (Image)</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/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="../intermediate/spatial_transformer_tutorial.html">๊ณต๊ฐ ๋ณํ๊ธฐ ๋คํธ์ํฌ(Spatial Transformer Networks) ํํ ๋ฆฌ์ผ</a></li>
<li class="toctree-l1"><a class="reference internal" href="neural_style_tutorial.html">PyTorch๋ฅผ ์ด์ฉํ ์ ๊ฒฝ๋ง-๋ณํ(Neural-Transfer)</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/fgsm_tutorial.html">์ ๋์ ์์ ์์ฑ(Adversarial Example Generation)</a></li>
</ul>
<p class="caption"><span class="caption-text">์ค๋์ค (Audio)</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../beginner/audio_preprocessing_tutorial.html">torchaudio Tutorial</a></li>
</ul>
<p class="caption"><span class="caption-text">ํ
์คํธ (Text)</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/char_rnn_classification_tutorial.html">๊ธฐ์ด๋ถํฐ ์์ํ๋ NLP: ๋ฌธ์-๋จ์ RNN์ผ๋ก ์ด๋ฆ ๋ถ๋ฅํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/char_rnn_generation_tutorial.html">๊ธฐ์ด๋ถํฐ ์์ํ๋ NLP: ๋ฌธ์-๋จ์ RNN์ผ๋ก ์ด๋ฆ ์์ฑํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/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">Text Classification with TorchText</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/torchtext_translation_tutorial.html">TorchText๋ก ์ธ์ด ๋ฒ์ญํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/transformer_tutorial.html">Sequence-to-Sequence Modeling with nn.Transformer and TorchText</a></li>
</ul>
<p class="caption"><span class="caption-text">Named Tensor (experimental)</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/named_tensor_tutorial.html">(experimental) Introduction to Named Tensors in PyTorch</a></li>
</ul>
<p class="caption"><span class="caption-text">๊ฐํ ํ์ต</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/reinforcement_q_learning.html">๊ฐํ ํ์ต (DQN) ํํ ๋ฆฌ์ผ</a></li>
</ul>
<p class="caption"><span class="caption-text">PyTorch ๋ชจ๋ธ์ ์ด์ํ๊ฒฝ์ ๋ฐฐํฌํ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/flask_rest_api_tutorial.html">Flask๋ฅผ ์ด์ฉํ์ฌ 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="cpp_export.html">C++์์ TorchScript ๋ชจ๋ธ ๋ก๋ฉํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="super_resolution_with_onnxruntime.html">(์ ํ) PyTorch ๋ชจ๋ธ์ ONNX์ผ๋ก ๋ณํํ๊ณ ONNX ๋ฐํ์์์ ์คํํ๊ธฐ</a></li>
</ul>
<p class="caption"><span class="caption-text">๋ณ๋ ฌ & ๋ถ์ฐ ํ์ต</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/model_parallel_tutorial.html">Single-Machine Model Parallel Best Practices</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/ddp_tutorial.html">Getting Started with Distributed Data Parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/dist_tuto.html">PyTorch๋ก ๋ถ์ฐ ์ดํ๋ฆฌ์ผ์ด์
๊ฐ๋ฐํ๊ธฐ</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/rpc_tutorial.html">Getting Started with Distributed RPC Framework</a></li>
<li class="toctree-l1"><a class="reference internal" href="../beginner/aws_distributed_training_tutorial.html">(advanced) PyTorch 1.0 Distributed Trainer with Amazon AWS</a></li>
</ul>
<p class="caption"><span class="caption-text">PyTorch ํ์ฅํ๊ธฐ</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="torch_script_custom_ops.html">Extending TorchScript with Custom C++ Operators</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch_script_custom_classes.html">Extending TorchScript with Custom C++ Classes</a></li>
<li class="toctree-l1"><a class="reference internal" href="numpy_extensions_tutorial.html">Creating Extensions Using numpy and scipy</a></li>
<li class="toctree-l1"><a class="reference internal" href="cpp_extension.html">Custom C++ and CUDA Extensions</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="dynamic_quantization_tutorial.html">(experimental) Dynamic Quantization on an LSTM Word Language Model</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">(experimental) Static Quantization with Eager Mode in PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/quantized_transfer_learning_tutorial.html">(experimental) Quantized Transfer Learning for Computer Vision Tutorial</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/dynamic_quantization_bert_tutorial.html">(experimental) Dynamic Quantization on BERT</a></li>
<li class="toctree-l1"><a class="reference internal" href="../intermediate/pruning_tutorial.html">Pruning Tutorial</a></li>
</ul>
<p class="caption"><span class="caption-text">๋ค๋ฅธ ์ธ์ด์์์ PyTorch</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="cpp_frontend.html">Using the PyTorch C++ Frontend</a></li>
</ul>
<p class="caption"><span class="caption-text">PyTorch Fundamentals In-Depth</span></p>
<ul>
<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">What is <cite>torch.nn</cite> <em>really</em>?</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>(experimental) Static Quantization with Eager Mode in PyTorch</li>
<li class="pytorch-breadcrumbs-aside">
<a href="../_sources/advanced/static_quantization_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">advanced/static_quantization_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">Note</p>
<p>Click <a class="reference internal" href="#sphx-glr-download-advanced-static-quantization-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="experimental-static-quantization-with-eager-mode-in-pytorch">
<span id="sphx-glr-advanced-static-quantization-tutorial-py"></span><h1>(experimental) Static Quantization with Eager Mode in PyTorch<a class="headerlink" href="#experimental-static-quantization-with-eager-mode-in-pytorch" title="Permalink to this headline">ยถ</a></h1>
<p><strong>Author</strong>: <a class="reference external" href="https://github.com/raghuramank100">Raghuraman Krishnamoorthi</a></p>
<p><strong>Edited by</strong>: <a class="reference external" href="https://github.com/SethHWeidman/">Seth Weidman</a></p>
<p>This tutorial shows how to do post-training static quantization, as well as illustrating
two more advanced techniques - per-channel quantization and quantization-aware training -
to further improve the modelโs accuracy. Note that quantization is currently only supported
for CPUs, so we will not be utilizing GPUs / CUDA in this tutorial.</p>
<p>By the end of this tutorial, you will see how quantization in PyTorch can result in
significant decreases in model size while increasing speed. Furthermore, youโll see how
to easily apply some advanced quantization techniques shown
<a class="reference external" href="https://arxiv.org/abs/1806.08342">here</a> so that your quantized models take much less
of an accuracy hit than they would otherwise.</p>
<p>Warning: we use a lot of boilerplate code from other PyTorch repos to, for example,
define the <code class="docutils literal notranslate"><span class="pre">MobileNetV2</span></code> model archtecture, define data loaders, and so on. We of course
encourage you to read it; but if you want to get to the quantization features, feel free
to skip to the โ4. Post-training static quantizationโ section.</p>
<p>Weโll start by doing the necessary imports:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</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">torchvision</span>
<span class="kn">from</span> <span class="nn">torch.utils.data</span> <span class="k">import</span> <span class="n">DataLoader</span>
<span class="kn">from</span> <span class="nn">torchvision</span> <span class="k">import</span> <span class="n">datasets</span>
<span class="kn">import</span> <span class="nn">torchvision.transforms</span> <span class="k">as</span> <span class="nn">transforms</span>
<span class="kn">import</span> <span class="nn">os</span>
<span class="kn">import</span> <span class="nn">time</span>
<span class="kn">import</span> <span class="nn">sys</span>
<span class="kn">import</span> <span class="nn">torch.quantization</span>
<span class="c1"># # Setup warnings</span>
<span class="kn">import</span> <span class="nn">warnings</span>
<span class="n">warnings</span><span class="o">.</span><span class="n">filterwarnings</span><span class="p">(</span>
<span class="n">action</span><span class="o">=</span><span class="s1">'ignore'</span><span class="p">,</span>
<span class="n">category</span><span class="o">=</span><span class="ne">DeprecationWarning</span><span class="p">,</span>
<span class="n">module</span><span class="o">=</span><span class="sa">r</span><span class="s1">'.*'</span>
<span class="p">)</span>
<span class="n">warnings</span><span class="o">.</span><span class="n">filterwarnings</span><span class="p">(</span>
<span class="n">action</span><span class="o">=</span><span class="s1">'default'</span><span class="p">,</span>
<span class="n">module</span><span class="o">=</span><span class="sa">r</span><span class="s1">'torch.quantization'</span>
<span class="p">)</span>
<span class="c1"># Specify random seed for repeatable results</span>
<span class="n">torch</span><span class="o">.</span><span class="n">manual_seed</span><span class="p">(</span><span class="mi">191009</span><span class="p">)</span>
</pre></div>
</div>
<div class="section" id="model-architecture">
<h2>1. Model architecture<a class="headerlink" href="#model-architecture" title="Permalink to this headline">ยถ</a></h2>
<p>We first define the MobileNetV2 model architecture, with several notable modifications
to enable quantization:</p>
<ul class="simple">
<li><p>Replacing addition with <code class="docutils literal notranslate"><span class="pre">nn.quantized.FloatFunctional</span></code></p></li>
<li><p>Insert <code class="docutils literal notranslate"><span class="pre">QuantStub</span></code> and <code class="docutils literal notranslate"><span class="pre">DeQuantStub</span></code> at the beginning and end of the network.</p></li>
<li><p>Replace ReLU6 with ReLU</p></li>
</ul>
<p>Note: this code is taken from
<a class="reference external" href="https://github.com/pytorch/vision/blob/master/torchvision/models/mobilenet.py">here</a>.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">torch.quantization</span> <span class="k">import</span> <span class="n">QuantStub</span><span class="p">,</span> <span class="n">DeQuantStub</span>
<span class="k">def</span> <span class="nf">_make_divisible</span><span class="p">(</span><span class="n">v</span><span class="p">,</span> <span class="n">divisor</span><span class="p">,</span> <span class="n">min_value</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
<span class="sd">"""</span>
<span class="sd"> This function is taken from the original tf repo.</span>
<span class="sd"> It ensures that all layers have a channel number that is divisible by 8</span>
<span class="sd"> It can be seen here:</span>
<span class="sd"> https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py</span>
<span class="sd"> :param v:</span>
<span class="sd"> :param divisor:</span>
<span class="sd"> :param min_value:</span>
<span class="sd"> :return:</span>
<span class="sd"> """</span>
<span class="k">if</span> <span class="n">min_value</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">min_value</span> <span class="o">=</span> <span class="n">divisor</span>
<span class="n">new_v</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span><span class="n">min_value</span><span class="p">,</span> <span class="nb">int</span><span class="p">(</span><span class="n">v</span> <span class="o">+</span> <span class="n">divisor</span> <span class="o">/</span> <span class="mi">2</span><span class="p">)</span> <span class="o">//</span> <span class="n">divisor</span> <span class="o">*</span> <span class="n">divisor</span><span class="p">)</span>
<span class="c1"># Make sure that round down does not go down by more than 10%.</span>
<span class="k">if</span> <span class="n">new_v</span> <span class="o"><</span> <span class="mf">0.9</span> <span class="o">*</span> <span class="n">v</span><span class="p">:</span>
<span class="n">new_v</span> <span class="o">+=</span> <span class="n">divisor</span>
<span class="k">return</span> <span class="n">new_v</span>
<span class="k">class</span> <span class="nc">ConvBNReLU</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="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">in_planes</span><span class="p">,</span> <span class="n">out_planes</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">groups</span><span class="o">=</span><span class="mi">1</span><span class="p">):</span>
<span class="n">padding</span> <span class="o">=</span> <span class="p">(</span><span class="n">kernel_size</span> <span class="o">-</span> <span class="mi">1</span><span class="p">)</span> <span class="o">//</span> <span class="mi">2</span>
<span class="nb">super</span><span class="p">(</span><span class="n">ConvBNReLU</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="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">in_planes</span><span class="p">,</span> <span class="n">out_planes</span><span class="p">,</span> <span class="n">kernel_size</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">padding</span><span class="p">,</span> <span class="n">groups</span><span class="o">=</span><span class="n">groups</span><span class="p">,</span> <span class="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
<span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">out_planes</span><span class="p">,</span> <span class="n">momentum</span><span class="o">=</span><span class="mf">0.1</span><span class="p">),</span>
<span class="c1"># Replace with ReLU</span>
<span class="n">nn</span><span class="o">.</span><span class="n">ReLU</span><span class="p">(</span><span class="n">inplace</span><span class="o">=</span><span class="kc">False</span><span class="p">)</span>
<span class="p">)</span>
<span class="k">class</span> <span class="nc">InvertedResidual</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="nf">__init__</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="n">oup</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">expand_ratio</span><span class="p">):</span>
<span class="nb">super</span><span class="p">(</span><span class="n">InvertedResidual</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">stride</span> <span class="o">=</span> <span class="n">stride</span>
<span class="k">assert</span> <span class="n">stride</span> <span class="ow">in</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="n">hidden_dim</span> <span class="o">=</span> <span class="nb">int</span><span class="p">(</span><span class="nb">round</span><span class="p">(</span><span class="n">inp</span> <span class="o">*</span> <span class="n">expand_ratio</span><span class="p">))</span>
<span class="bp">self</span><span class="o">.</span><span class="n">use_res_connect</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">stride</span> <span class="o">==</span> <span class="mi">1</span> <span class="ow">and</span> <span class="n">inp</span> <span class="o">==</span> <span class="n">oup</span>
<span class="n">layers</span> <span class="o">=</span> <span class="p">[]</span>
<span class="k">if</span> <span class="n">expand_ratio</span> <span class="o">!=</span> <span class="mi">1</span><span class="p">:</span>
<span class="c1"># pw</span>
<span class="n">layers</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">ConvBNReLU</span><span class="p">(</span><span class="n">inp</span><span class="p">,</span> <span class="n">hidden_dim</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">1</span><span class="p">))</span>
<span class="n">layers</span><span class="o">.</span><span class="n">extend</span><span class="p">([</span>
<span class="c1"># dw</span>
<span class="n">ConvBNReLU</span><span class="p">(</span><span class="n">hidden_dim</span><span class="p">,</span> <span class="n">hidden_dim</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="n">stride</span><span class="p">,</span> <span class="n">groups</span><span class="o">=</span><span class="n">hidden_dim</span><span class="p">),</span>
<span class="c1"># pw-linear</span>
<span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">(</span><span class="n">hidden_dim</span><span class="p">,</span> <span class="n">oup</span><span class="p">,</span> <span class="mi">1</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="n">bias</span><span class="o">=</span><span class="kc">False</span><span class="p">),</span>
<span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">(</span><span class="n">oup</span><span class="p">,</span> <span class="n">momentum</span><span class="o">=</span><span class="mf">0.1</span><span class="p">),</span>
<span class="p">])</span>
<span class="bp">self</span><span class="o">.</span><span class="n">conv</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">layers</span><span class="p">)</span>
<span class="c1"># Replace torch.add with floatfunctional</span>
<span class="bp">self</span><span class="o">.</span><span class="n">skip_add</span> <span class="o">=</span> <span class="n">nn</span><span class="o">.</span><span class="n">quantized</span><span class="o">.</span><span class="n">FloatFunctional</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="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">use_res_connect</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">skip_add</span><span class="o">.</span><span class="n">add</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">))</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">conv</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">class</span> <span class="nc">MobileNetV2</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="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">num_classes</span><span class="o">=</span><span class="mi">1000</span><span class="p">,</span> <span class="n">width_mult</span><span class="o">=</span><span class="mf">1.0</span><span class="p">,</span> <span class="n">inverted_residual_setting</span><span class="o">=</span><span class="kc">None</span><span class="p">,</span> <span class="n">round_nearest</span><span class="o">=</span><span class="mi">8</span><span class="p">):</span>
<span class="sd">"""</span>
<span class="sd"> MobileNet V2 main class</span>
<span class="sd"> Args:</span>
<span class="sd"> num_classes (int): Number of classes</span>
<span class="sd"> width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount</span>
<span class="sd"> inverted_residual_setting: Network structure</span>
<span class="sd"> round_nearest (int): Round the number of channels in each layer to be a multiple of this number</span>
<span class="sd"> Set to 1 to turn off rounding</span>
<span class="sd"> """</span>
<span class="nb">super</span><span class="p">(</span><span class="n">MobileNetV2</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="n">block</span> <span class="o">=</span> <span class="n">InvertedResidual</span>
<span class="n">input_channel</span> <span class="o">=</span> <span class="mi">32</span>
<span class="n">last_channel</span> <span class="o">=</span> <span class="mi">1280</span>
<span class="k">if</span> <span class="n">inverted_residual_setting</span> <span class="ow">is</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">inverted_residual_setting</span> <span class="o">=</span> <span class="p">[</span>
<span class="c1"># t, c, n, s</span>
<span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">16</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">],</span>
<span class="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">24</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="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">32</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">64</span><span class="p">,</span> <span class="mi">4</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">96</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">1</span><span class="p">],</span>
<span class="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">160</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">2</span><span class="p">],</span>
<span class="p">[</span><span class="mi">6</span><span class="p">,</span> <span class="mi">320</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">1</span><span class="p">],</span>
<span class="p">]</span>
<span class="c1"># only check the first element, assuming user knows t,c,n,s are required</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">inverted_residual_setting</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span> <span class="ow">or</span> <span class="nb">len</span><span class="p">(</span><span class="n">inverted_residual_setting</span><span class="p">[</span><span class="mi">0</span><span class="p">])</span> <span class="o">!=</span> <span class="mi">4</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"inverted_residual_setting should be non-empty "</span>
<span class="s2">"or a 4-element list, got </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">inverted_residual_setting</span><span class="p">))</span>
<span class="c1"># building first layer</span>
<span class="n">input_channel</span> <span class="o">=</span> <span class="n">_make_divisible</span><span class="p">(</span><span class="n">input_channel</span> <span class="o">*</span> <span class="n">width_mult</span><span class="p">,</span> <span class="n">round_nearest</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">last_channel</span> <span class="o">=</span> <span class="n">_make_divisible</span><span class="p">(</span><span class="n">last_channel</span> <span class="o">*</span> <span class="nb">max</span><span class="p">(</span><span class="mf">1.0</span><span class="p">,</span> <span class="n">width_mult</span><span class="p">),</span> <span class="n">round_nearest</span><span class="p">)</span>
<span class="n">features</span> <span class="o">=</span> <span class="p">[</span><span class="n">ConvBNReLU</span><span class="p">(</span><span class="mi">3</span><span class="p">,</span> <span class="n">input_channel</span><span class="p">,</span> <span class="n">stride</span><span class="o">=</span><span class="mi">2</span><span class="p">)]</span>
<span class="c1"># building inverted residual blocks</span>
<span class="k">for</span> <span class="n">t</span><span class="p">,</span> <span class="n">c</span><span class="p">,</span> <span class="n">n</span><span class="p">,</span> <span class="n">s</span> <span class="ow">in</span> <span class="n">inverted_residual_setting</span><span class="p">:</span>
<span class="n">output_channel</span> <span class="o">=</span> <span class="n">_make_divisible</span><span class="p">(</span><span class="n">c</span> <span class="o">*</span> <span class="n">width_mult</span><span class="p">,</span> <span class="n">round_nearest</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="n">n</span><span class="p">):</span>
<span class="n">stride</span> <span class="o">=</span> <span class="n">s</span> <span class="k">if</span> <span class="n">i</span> <span class="o">==</span> <span class="mi">0</span> <span class="k">else</span> <span class="mi">1</span>
<span class="n">features</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">block</span><span class="p">(</span><span class="n">input_channel</span><span class="p">,</span> <span class="n">output_channel</span><span class="p">,</span> <span class="n">stride</span><span class="p">,</span> <span class="n">expand_ratio</span><span class="o">=</span><span class="n">t</span><span class="p">))</span>
<span class="n">input_channel</span> <span class="o">=</span> <span class="n">output_channel</span>
<span class="c1"># building last several layers</span>
<span class="n">features</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">ConvBNReLU</span><span class="p">(</span><span class="n">input_channel</span><span class="p">,</span> <span class="bp">self</span><span class="o">.</span><span class="n">last_channel</span><span class="p">,</span> <span class="n">kernel_size</span><span class="o">=</span><span class="mi">1</span><span class="p">))</span>
<span class="c1"># make it nn.Sequential</span>
<span class="bp">self</span><span class="o">.</span><span class="n">features</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">features</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">quant</span> <span class="o">=</span> <span class="n">QuantStub</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">dequant</span> <span class="o">=</span> <span class="n">DeQuantStub</span><span class="p">()</span>
<span class="c1"># building classifier</span>
<span class="bp">self</span><span class="o">.</span><span class="n">classifier</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="n">nn</span><span class="o">.</span><span class="n">Dropout</span><span class="p">(</span><span class="mf">0.2</span><span class="p">),</span>
<span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">last_channel</span><span class="p">,</span> <span class="n">num_classes</span><span class="p">),</span>
<span class="p">)</span>
<span class="c1"># weight initialization</span>
<span class="k">for</span> <span class="n">m</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">modules</span><span class="p">():</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Conv2d</span><span class="p">):</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">kaiming_normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">,</span> <span class="n">mode</span><span class="o">=</span><span class="s1">'fan_out'</span><span class="p">)</span>
<span class="k">if</span> <span class="n">m</span><span class="o">.</span><span class="n">bias</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">zeros_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="p">)</span>
<span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">BatchNorm2d</span><span class="p">):</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">ones_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">)</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">zeros_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</span><span class="p">)</span>
<span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="n">nn</span><span class="o">.</span><span class="n">Linear</span><span class="p">):</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">normal_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">weight</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="mf">0.01</span><span class="p">)</span>
<span class="n">nn</span><span class="o">.</span><span class="n">init</span><span class="o">.</span><span class="n">zeros_</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">bias</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="bp">self</span><span class="o">.</span><span class="n">quant</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">features</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="n">mean</span><span class="p">([</span><span class="mi">2</span><span class="p">,</span> <span class="mi">3</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">classifier</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">dequant</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="c1"># Fuse Conv+BN and Conv+BN+Relu modules prior to quantization</span>
<span class="c1"># This operation does not change the numerics</span>
<span class="k">def</span> <span class="nf">fuse_model</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">for</span> <span class="n">m</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">modules</span><span class="p">():</span>
<span class="k">if</span> <span class="nb">type</span><span class="p">(</span><span class="n">m</span><span class="p">)</span> <span class="o">==</span> <span class="n">ConvBNReLU</span><span class="p">:</span>
<span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">fuse_modules</span><span class="p">(</span><span class="n">m</span><span class="p">,</span> <span class="p">[</span><span class="s1">'0'</span><span class="p">,</span> <span class="s1">'1'</span><span class="p">,</span> <span class="s1">'2'</span><span class="p">],</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="k">if</span> <span class="nb">type</span><span class="p">(</span><span class="n">m</span><span class="p">)</span> <span class="o">==</span> <span class="n">InvertedResidual</span><span class="p">:</span>
<span class="k">for</span> <span class="n">idx</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">conv</span><span class="p">)):</span>
<span class="k">if</span> <span class="nb">type</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">conv</span><span class="p">[</span><span class="n">idx</span><span class="p">])</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">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">fuse_modules</span><span class="p">(</span><span class="n">m</span><span class="o">.</span><span class="n">conv</span><span class="p">,</span> <span class="p">[</span><span class="nb">str</span><span class="p">(</span><span class="n">idx</span><span class="p">),</span> <span class="nb">str</span><span class="p">(</span><span class="n">idx</span> <span class="o">+</span> <span class="mi">1</span><span class="p">)],</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
</pre></div>
</div>
</div>
<div class="section" id="helper-functions">
<h2>2. Helper functions<a class="headerlink" href="#helper-functions" title="Permalink to this headline">ยถ</a></h2>
<p>We next define several helper functions to help with model evaluation. These mostly come from
<a class="reference external" href="https://github.com/pytorch/examples/blob/master/imagenet/main.py">here</a>.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">class</span> <span class="nc">AverageMeter</span><span class="p">(</span><span class="nb">object</span><span class="p">):</span>
<span class="sd">"""Computes and stores the average and current value"""</span>
<span class="k">def</span> <span class="nf">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">name</span><span class="p">,</span> <span class="n">fmt</span><span class="o">=</span><span class="s1">':f'</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">name</span> <span class="o">=</span> <span class="n">name</span>
<span class="bp">self</span><span class="o">.</span><span class="n">fmt</span> <span class="o">=</span> <span class="n">fmt</span>
<span class="bp">self</span><span class="o">.</span><span class="n">reset</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">reset</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">val</span> <span class="o">=</span> <span class="mi">0</span>
<span class="bp">self</span><span class="o">.</span><span class="n">avg</span> <span class="o">=</span> <span class="mi">0</span>
<span class="bp">self</span><span class="o">.</span><span class="n">sum</span> <span class="o">=</span> <span class="mi">0</span>
<span class="bp">self</span><span class="o">.</span><span class="n">count</span> <span class="o">=</span> <span class="mi">0</span>
<span class="k">def</span> <span class="nf">update</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">val</span><span class="p">,</span> <span class="n">n</span><span class="o">=</span><span class="mi">1</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">val</span> <span class="o">=</span> <span class="n">val</span>
<span class="bp">self</span><span class="o">.</span><span class="n">sum</span> <span class="o">+=</span> <span class="n">val</span> <span class="o">*</span> <span class="n">n</span>
<span class="bp">self</span><span class="o">.</span><span class="n">count</span> <span class="o">+=</span> <span class="n">n</span>
<span class="bp">self</span><span class="o">.</span><span class="n">avg</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">sum</span> <span class="o">/</span> <span class="bp">self</span><span class="o">.</span><span class="n">count</span>
<span class="k">def</span> <span class="nf">__str__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="n">fmtstr</span> <span class="o">=</span> <span class="s1">'</span><span class="si">{name}</span><span class="s1"> {val'</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">fmt</span> <span class="o">+</span> <span class="s1">'} ({avg'</span> <span class="o">+</span> <span class="bp">self</span><span class="o">.</span><span class="n">fmt</span> <span class="o">+</span> <span class="s1">'})'</span>
<span class="k">return</span> <span class="n">fmtstr</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="o">**</span><span class="bp">self</span><span class="o">.</span><span class="vm">__dict__</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">accuracy</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">,</span> <span class="n">topk</span><span class="o">=</span><span class="p">(</span><span class="mi">1</span><span class="p">,)):</span>
<span class="sd">"""Computes the accuracy over the k top predictions for the specified values of k"""</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="n">maxk</span> <span class="o">=</span> <span class="nb">max</span><span class="p">(</span><span class="n">topk</span><span class="p">)</span>
<span class="n">batch_size</span> <span class="o">=</span> <span class="n">target</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="n">_</span><span class="p">,</span> <span class="n">pred</span> <span class="o">=</span> <span class="n">output</span><span class="o">.</span><span class="n">topk</span><span class="p">(</span><span class="n">maxk</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="kc">True</span><span class="p">,</span> <span class="kc">True</span><span class="p">)</span>
<span class="n">pred</span> <span class="o">=</span> <span class="n">pred</span><span class="o">.</span><span class="n">t</span><span class="p">()</span>
<span class="n">correct</span> <span class="o">=</span> <span class="n">pred</span><span class="o">.</span><span class="n">eq</span><span class="p">(</span><span class="n">target</span><span class="o">.</span><span class="n">view</span><span class="p">(</span><span class="mi">1</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">expand_as</span><span class="p">(</span><span class="n">pred</span><span class="p">))</span>
<span class="n">res</span> <span class="o">=</span> <span class="p">[]</span>
<span class="k">for</span> <span class="n">k</span> <span class="ow">in</span> <span class="n">topk</span><span class="p">:</span>
<span class="n">correct_k</span> <span class="o">=</span> <span class="n">correct</span><span class="p">[:</span><span class="n">k</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="o">.</span><span class="n">float</span><span class="p">()</span><span class="o">.</span><span class="n">sum</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="n">keepdim</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">res</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">correct_k</span><span class="o">.</span><span class="n">mul_</span><span class="p">(</span><span class="mf">100.0</span> <span class="o">/</span> <span class="n">batch_size</span><span class="p">))</span>
<span class="k">return</span> <span class="n">res</span>
<span class="k">def</span> <span class="nf">evaluate</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader</span><span class="p">,</span> <span class="n">neval_batches</span><span class="p">):</span>
<span class="n">model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
<span class="n">top1</span> <span class="o">=</span> <span class="n">AverageMeter</span><span class="p">(</span><span class="s1">'Acc@1'</span><span class="p">,</span> <span class="s1">':6.2f'</span><span class="p">)</span>
<span class="n">top5</span> <span class="o">=</span> <span class="n">AverageMeter</span><span class="p">(</span><span class="s1">'Acc@5'</span><span class="p">,</span> <span class="s1">':6.2f'</span><span class="p">)</span>
<span class="n">cnt</span> <span class="o">=</span> <span class="mi">0</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">image</span><span class="p">,</span> <span class="n">target</span> <span class="ow">in</span> <span class="n">data_loader</span><span class="p">:</span>
<span class="n">output</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="n">image</span><span class="p">)</span>
<span class="n">loss</span> <span class="o">=</span> <span class="n">criterion</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
<span class="n">cnt</span> <span class="o">+=</span> <span class="mi">1</span>
<span class="n">acc1</span><span class="p">,</span> <span class="n">acc5</span> <span class="o">=</span> <span class="n">accuracy</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">,</span> <span class="n">topk</span><span class="o">=</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">5</span><span class="p">))</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'.'</span><span class="p">,</span> <span class="n">end</span> <span class="o">=</span> <span class="s1">''</span><span class="p">)</span>
<span class="n">top1</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">acc1</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">image</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="n">top5</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">acc5</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">image</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="k">if</span> <span class="n">cnt</span> <span class="o">>=</span> <span class="n">neval_batches</span><span class="p">:</span>
<span class="k">return</span> <span class="n">top1</span><span class="p">,</span> <span class="n">top5</span>
<span class="k">return</span> <span class="n">top1</span><span class="p">,</span> <span class="n">top5</span>
<span class="k">def</span> <span class="nf">load_model</span><span class="p">(</span><span class="n">model_file</span><span class="p">):</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">MobileNetV2</span><span class="p">()</span>
<span class="n">state_dict</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">load</span><span class="p">(</span><span class="n">model_file</span><span class="p">)</span>
<span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">state_dict</span><span class="p">)</span>
<span class="n">model</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s1">'cpu'</span><span class="p">)</span>
<span class="k">return</span> <span class="n">model</span>
<span class="k">def</span> <span class="nf">print_size_of_model</span><span class="p">(</span><span class="n">model</span><span class="p">):</span>
<span class="n">torch</span><span class="o">.</span><span class="n">save</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="s2">"temp.p"</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Size (MB):'</span><span class="p">,</span> <span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">getsize</span><span class="p">(</span><span class="s2">"temp.p"</span><span class="p">)</span><span class="o">/</span><span class="mf">1e6</span><span class="p">)</span>
<span class="n">os</span><span class="o">.</span><span class="n">remove</span><span class="p">(</span><span class="s1">'temp.p'</span><span class="p">)</span>
</pre></div>
</div>
</div>
<div class="section" id="define-dataset-and-data-loaders">
<h2>3. Define dataset and data loaders<a class="headerlink" href="#define-dataset-and-data-loaders" title="Permalink to this headline">ยถ</a></h2>
<p>As our last major setup step, we define our dataloaders for our training and testing set.</p>
<div class="section" id="imagenet-data">
<h3>ImageNet Data<a class="headerlink" href="#imagenet-data" title="Permalink to this headline">ยถ</a></h3>
<p>The specific dataset weโve created for this tutorial contains just 1000 images from the ImageNet data, one from
each class (this dataset, at just over 250 MB, is small enough that it can be downloaded
relatively easily). The URL for this custom dataset is:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">https</span><span class="p">:</span><span class="o">//</span><span class="n">s3</span><span class="o">.</span><span class="n">amazonaws</span><span class="o">.</span><span class="n">com</span><span class="o">/</span><span class="n">pytorch</span><span class="o">-</span><span class="n">tutorial</span><span class="o">-</span><span class="n">assets</span><span class="o">/</span><span class="n">imagenet_1k</span><span class="o">.</span><span class="n">zip</span>
</pre></div>
</div>
<p>To download this data locally using Python, you could use:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">requests</span>
<span class="n">url</span> <span class="o">=</span> <span class="s1">'https://s3.amazonaws.com/pytorch-tutorial-assets/imagenet_1k.zip`</span>
<span class="n">filename</span> <span class="o">=</span> <span class="s1">'~/Downloads/imagenet_1k_data.zip'</span>
<span class="n">r</span> <span class="o">=</span> <span class="n">requests</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="n">url</span><span class="p">)</span>
<span class="k">with</span> <span class="nb">open</span><span class="p">(</span><span class="n">filename</span><span class="p">,</span> <span class="s1">'wb'</span><span class="p">)</span> <span class="k">as</span> <span class="n">f</span><span class="p">:</span>
<span class="n">f</span><span class="o">.</span><span class="n">write</span><span class="p">(</span><span class="n">r</span><span class="o">.</span><span class="n">content</span><span class="p">)</span>
</pre></div>
</div>
<p>For this tutorial to run, we download this data and move it to the right place using
<a class="reference external" href="https://github.com/pytorch/tutorials/blob/master/Makefile#L97-L98">these lines</a>
from the <a class="reference external" href="https://github.com/pytorch/tutorials/blob/master/Makefile">Makefile</a>.</p>
<p>To run the code in this tutorial using the entire ImageNet dataset, on the other hand, you could download
the data using <code class="docutils literal notranslate"><span class="pre">torchvision</span></code> following
<a class="reference external" href="https://pytorch.org/docs/stable/torchvision/datasets.html#imagenet">here</a>. For example,
to download the training set and apply some standard transformations to it, you could use:</p>
<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span> <span class="nn">torchvision</span>
<span class="kn">import</span> <span class="nn">torchvision.transforms</span> <span class="kn">as</span> <span class="nn">transforms</span>
<span class="n">imagenet_dataset</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">datasets</span><span class="o">.</span><span class="n">ImageNet</span><span class="p">(</span>
<span class="s1">'~/.data/imagenet'</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">download</span><span class="o">=</span><span class="bp">True</span><span class="p">,</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">RandomResizedCrop</span><span class="p">(</span><span class="mi">224</span><span class="p">),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">RandomHorizontalFlip</span><span class="p">(),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">(</span><span class="n">mean</span><span class="o">=</span><span class="p">[</span><span class="mf">0.485</span><span class="p">,</span> <span class="mf">0.456</span><span class="p">,</span> <span class="mf">0.406</span><span class="p">],</span>
<span class="n">std</span><span class="o">=</span><span class="p">[</span><span class="mf">0.229</span><span class="p">,</span> <span class="mf">0.224</span><span class="p">,</span> <span class="mf">0.225</span><span class="p">]),</span>
<span class="p">])</span>
</pre></div>
</div>
<p>With the data downloaded, we show functions below that define dataloaders weโll use to read
in this data. These functions mostly come from
<a class="reference external" href="https://github.com/pytorch/vision/blob/master/references/detection/train.py">here</a>.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">prepare_data_loaders</span><span class="p">(</span><span class="n">data_path</span><span class="p">):</span>
<span class="n">traindir</span> <span class="o">=</span> <span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">data_path</span><span class="p">,</span> <span class="s1">'train'</span><span class="p">)</span>
<span class="n">valdir</span> <span class="o">=</span> <span class="n">os</span><span class="o">.</span><span class="n">path</span><span class="o">.</span><span class="n">join</span><span class="p">(</span><span class="n">data_path</span><span class="p">,</span> <span class="s1">'val'</span><span class="p">)</span>
<span class="n">normalize</span> <span class="o">=</span> <span class="n">transforms</span><span class="o">.</span><span class="n">Normalize</span><span class="p">(</span><span class="n">mean</span><span class="o">=</span><span class="p">[</span><span class="mf">0.485</span><span class="p">,</span> <span class="mf">0.456</span><span class="p">,</span> <span class="mf">0.406</span><span class="p">],</span>
<span class="n">std</span><span class="o">=</span><span class="p">[</span><span class="mf">0.229</span><span class="p">,</span> <span class="mf">0.224</span><span class="p">,</span> <span class="mf">0.225</span><span class="p">])</span>
<span class="n">dataset</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">datasets</span><span class="o">.</span><span class="n">ImageFolder</span><span class="p">(</span>
<span class="n">traindir</span><span class="p">,</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">RandomResizedCrop</span><span class="p">(</span><span class="mi">224</span><span class="p">),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">RandomHorizontalFlip</span><span class="p">(),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
<span class="n">normalize</span><span class="p">,</span>
<span class="p">]))</span>
<span class="n">dataset_test</span> <span class="o">=</span> <span class="n">torchvision</span><span class="o">.</span><span class="n">datasets</span><span class="o">.</span><span class="n">ImageFolder</span><span class="p">(</span>
<span class="n">valdir</span><span class="p">,</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">Compose</span><span class="p">([</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">Resize</span><span class="p">(</span><span class="mi">256</span><span class="p">),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">CenterCrop</span><span class="p">(</span><span class="mi">224</span><span class="p">),</span>
<span class="n">transforms</span><span class="o">.</span><span class="n">ToTensor</span><span class="p">(),</span>
<span class="n">normalize</span><span class="p">,</span>
<span class="p">]))</span>
<span class="n">train_sampler</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">RandomSampler</span><span class="p">(</span><span class="n">dataset</span><span class="p">)</span>
<span class="n">test_sampler</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">SequentialSampler</span><span class="p">(</span><span class="n">dataset_test</span><span class="p">)</span>
<span class="n">data_loader</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span><span class="p">(</span>
<span class="n">dataset</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="n">train_batch_size</span><span class="p">,</span>
<span class="n">sampler</span><span class="o">=</span><span class="n">train_sampler</span><span class="p">)</span>
<span class="n">data_loader_test</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">utils</span><span class="o">.</span><span class="n">data</span><span class="o">.</span><span class="n">DataLoader</span><span class="p">(</span>
<span class="n">dataset_test</span><span class="p">,</span> <span class="n">batch_size</span><span class="o">=</span><span class="n">eval_batch_size</span><span class="p">,</span>
<span class="n">sampler</span><span class="o">=</span><span class="n">test_sampler</span><span class="p">)</span>
<span class="k">return</span> <span class="n">data_loader</span><span class="p">,</span> <span class="n">data_loader_test</span>
</pre></div>
</div>
<p>Next, weโll load in the pre-trained MobileNetV2 model. We provide the URL to download the data from in <code class="docutils literal notranslate"><span class="pre">torchvision</span></code>
<a class="reference external" href="https://github.com/pytorch/vision/blob/master/torchvision/models/mobilenet.py#L9">here</a>.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">data_path</span> <span class="o">=</span> <span class="s1">'data/imagenet_1k'</span>
<span class="n">saved_model_dir</span> <span class="o">=</span> <span class="s1">'data/'</span>
<span class="n">float_model_file</span> <span class="o">=</span> <span class="s1">'mobilenet_pretrained_float.pth'</span>
<span class="n">scripted_float_model_file</span> <span class="o">=</span> <span class="s1">'mobilenet_quantization_scripted.pth'</span>
<span class="n">scripted_quantized_model_file</span> <span class="o">=</span> <span class="s1">'mobilenet_quantization_scripted_quantized.pth'</span>
<span class="n">train_batch_size</span> <span class="o">=</span> <span class="mi">30</span>
<span class="n">eval_batch_size</span> <span class="o">=</span> <span class="mi">30</span>
<span class="n">data_loader</span><span class="p">,</span> <span class="n">data_loader_test</span> <span class="o">=</span> <span class="n">prepare_data_loaders</span><span class="p">(</span><span class="n">data_path</span><span class="p">)</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">float_model</span> <span class="o">=</span> <span class="n">load_model</span><span class="p">(</span><span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">float_model_file</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s1">'cpu'</span><span class="p">)</span>
</pre></div>
</div>
<p>Next, weโll โfuse modulesโ; this can both make the model faster by saving on memory access
while also improving numerical accuracy. While this can be used with any model, this is
especially common with quantized models.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1"> Inverted Residual Block: Before fusion </span><span class="se">\n\n</span><span class="s1">'</span><span class="p">,</span> <span class="n">float_model</span><span class="o">.</span><span class="n">features</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">conv</span><span class="p">)</span>
<span class="n">float_model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
<span class="c1"># Fuses modules</span>
<span class="n">float_model</span><span class="o">.</span><span class="n">fuse_model</span><span class="p">()</span>
<span class="c1"># Note fusion of Conv+BN+Relu and Conv+Relu</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1"> Inverted Residual Block: After fusion</span><span class="se">\n\n</span><span class="s1">'</span><span class="p">,</span><span class="n">float_model</span><span class="o">.</span><span class="n">features</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">conv</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>Inverted Residual Block: Before fusion
Sequential(
(0): ConvBNReLU(
(0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): Conv2d(32, 16, kernel_size=(1, 1), stride=(1, 1), bias=False)
(2): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
Inverted Residual Block: After fusion
Sequential(
(0): ConvBNReLU(
(0): ConvReLU2d(
(0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32)
(1): ReLU()
)
(1): Identity()
(2): Identity()
)
(1): Conv2d(32, 16, kernel_size=(1, 1), stride=(1, 1))
(2): Identity()
)
</pre></div>
</div>
<p>Finally to get a โbaselineโ accuracy, letโs see the accuracy of our un-quantized model
with fused modules</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">num_eval_batches</span> <span class="o">=</span> <span class="mi">10</span>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Size of baseline model"</span><span class="p">)</span>
<span class="n">print_size_of_model</span><span class="p">(</span><span class="n">float_model</span><span class="p">)</span>
<span class="n">top1</span><span class="p">,</span> <span class="n">top5</span> <span class="o">=</span> <span class="n">evaluate</span><span class="p">(</span><span class="n">float_model</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader_test</span><span class="p">,</span> <span class="n">neval_batches</span><span class="o">=</span><span class="n">num_eval_batches</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Evaluation accuracy on </span><span class="si">%d</span><span class="s1"> images, </span><span class="si">%2.2f</span><span class="s1">'</span><span class="o">%</span><span class="p">(</span><span class="n">num_eval_batches</span> <span class="o">*</span> <span class="n">eval_batch_size</span><span class="p">,</span> <span class="n">top1</span><span class="o">.</span><span class="n">avg</span><span class="p">))</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">float_model</span><span class="p">),</span> <span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">scripted_float_model_file</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>Size of baseline model
Size (MB): 13.980272
..........Evaluation accuracy on 300 images, 78.00
</pre></div>
</div>
<p>We see 78% accuracy on 300 images, a solid baseline for ImageNet,
especially considering our model is just 14.0 MB.</p>
<p>This will be our baseline to compare to. Next, letโs try different quantization methods</p>
</div>
</div>
<div class="section" id="post-training-static-quantization">
<h2>4. Post-training static quantization<a class="headerlink" href="#post-training-static-quantization" title="Permalink to this headline">ยถ</a></h2>
<p>Post-training static quantization involves not just converting the weights from float to int,
as in dynamic quantization, but also performing the additional step of first feeding batches
of data through the network and computing the resulting distributions of the different activations
(specifically, this is done by inserting <cite>observer</cite> modules at different points that record this
data). These distributions are then used to determine how the specifically the different activations
should be quantized at inference time (a simple technique would be to simply divide the entire range
of activations into 256 levels, but we support more sophisticated methods as well). Importantly,
this additional step allows us to pass quantized values between operations instead of converting these
values to floats - and then back to ints - between every operation, resulting in a significant speed-up.</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">num_calibration_batches</span> <span class="o">=</span> <span class="mi">10</span>
<span class="n">myModel</span> <span class="o">=</span> <span class="n">load_model</span><span class="p">(</span><span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">float_model_file</span><span class="p">)</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="s1">'cpu'</span><span class="p">)</span>
<span class="n">myModel</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
<span class="c1"># Fuse Conv, bn and relu</span>
<span class="n">myModel</span><span class="o">.</span><span class="n">fuse_model</span><span class="p">()</span>
<span class="c1"># Specify quantization configuration</span>
<span class="c1"># Start with simple min/max range estimation and per-tensor quantization of weights</span>
<span class="n">myModel</span><span class="o">.</span><span class="n">qconfig</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">default_qconfig</span>
<span class="nb">print</span><span class="p">(</span><span class="n">myModel</span><span class="o">.</span><span class="n">qconfig</span><span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">prepare</span><span class="p">(</span><span class="n">myModel</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="c1"># Calibrate first</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Post Training Quantization Prepare: Inserting Observers'</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1"> Inverted Residual Block:After observer insertion </span><span class="se">\n\n</span><span class="s1">'</span><span class="p">,</span> <span class="n">myModel</span><span class="o">.</span><span class="n">features</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">conv</span><span class="p">)</span>
<span class="c1"># Calibrate with the training set</span>
<span class="n">evaluate</span><span class="p">(</span><span class="n">myModel</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader</span><span class="p">,</span> <span class="n">neval_batches</span><span class="o">=</span><span class="n">num_calibration_batches</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Post Training Quantization: Calibration done'</span><span class="p">)</span>
<span class="c1"># Convert to quantized model</span>
<span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">convert</span><span class="p">(</span><span class="n">myModel</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Post Training Quantization: Convert done'</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'</span><span class="se">\n</span><span class="s1"> Inverted Residual Block: After fusion and quantization, note fused modules: </span><span class="se">\n\n</span><span class="s1">'</span><span class="p">,</span><span class="n">myModel</span><span class="o">.</span><span class="n">features</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">conv</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Size of model after quantization"</span><span class="p">)</span>
<span class="n">print_size_of_model</span><span class="p">(</span><span class="n">myModel</span><span class="p">)</span>
<span class="n">top1</span><span class="p">,</span> <span class="n">top5</span> <span class="o">=</span> <span class="n">evaluate</span><span class="p">(</span><span class="n">myModel</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader_test</span><span class="p">,</span> <span class="n">neval_batches</span><span class="o">=</span><span class="n">num_eval_batches</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Evaluation accuracy on </span><span class="si">%d</span><span class="s1"> images, </span><span class="si">%2.2f</span><span class="s1">'</span><span class="o">%</span><span class="p">(</span><span class="n">num_eval_batches</span> <span class="o">*</span> <span class="n">eval_batch_size</span><span class="p">,</span> <span class="n">top1</span><span class="o">.</span><span class="n">avg</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>QConfig(activation=functools.partial(<class 'torch.quantization.observer.MinMaxObserver'>, reduce_range=True), weight=functools.partial(<class 'torch.quantization.observer.MinMaxObserver'>, dtype=torch.qint8, qscheme=torch.per_tensor_symmetric))
Post Training Quantization Prepare: Inserting Observers
Inverted Residual Block:After observer insertion
Sequential(
(0): ConvBNReLU(
(0): ConvReLU2d(
(0): Conv2d(
32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32
(observer): MinMaxObserver(min_val=None, max_val=None)
)
(1): ReLU(
(observer): MinMaxObserver(min_val=None, max_val=None)
)
)
(1): Identity()
(2): Identity()
)
(1): Conv2d(
32, 16, kernel_size=(1, 1), stride=(1, 1)
(observer): MinMaxObserver(min_val=None, max_val=None)
)
(2): Identity()
)
..........Post Training Quantization: Calibration done
Post Training Quantization: Convert done
Inverted Residual Block: After fusion and quantization, note fused modules:
Sequential(
(0): ConvBNReLU(
(0): QuantizedConvReLU2d(32, 32, kernel_size=(3, 3), stride=(1, 1), scale=0.15047618746757507, zero_point=0, padding=(1, 1), groups=32)
(1): Identity()
(2): Identity()
)
(1): QuantizedConv2d(32, 16, kernel_size=(1, 1), stride=(1, 1), scale=0.1900387406349182, zero_point=75)
(2): Identity()
)
Size of model after quantization
Size (MB): 3.586995
..........Evaluation accuracy on 300 images, 62.67
</pre></div>
</div>
<p>For this quantized model, we see a significantly lower accuracy of just ~62% on these same 300
images. Nevertheless, we did reduce the size of our model down to just under 3.6 MB, almost a 4x
decrease.</p>
<p>In addition, we can significantly improve on the accuracy simply by using a different
quantization configuration. We repeat the same exercise with the recommended configuration for
quantizing for x86 architectures. This configuration does the following:</p>
<ul class="simple">
<li><p>Quantizes weights on a per-channel basis</p></li>
<li><p>Uses a histogram observer that collects a histogram of activations and then picks
quantization parameters in an optimal manner.</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">per_channel_quantized_model</span> <span class="o">=</span> <span class="n">load_model</span><span class="p">(</span><span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">float_model_file</span><span class="p">)</span>
<span class="n">per_channel_quantized_model</span><span class="o">.</span><span class="n">eval</span><span class="p">()</span>
<span class="n">per_channel_quantized_model</span><span class="o">.</span><span class="n">fuse_model</span><span class="p">()</span>
<span class="n">per_channel_quantized_model</span><span class="o">.</span><span class="n">qconfig</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">get_default_qconfig</span><span class="p">(</span><span class="s1">'fbgemm'</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="o">.</span><span class="n">qconfig</span><span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">prepare</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">evaluate</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="p">,</span><span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader</span><span class="p">,</span> <span class="n">num_calibration_batches</span><span class="p">)</span>
<span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">convert</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">top1</span><span class="p">,</span> <span class="n">top5</span> <span class="o">=</span> <span class="n">evaluate</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">data_loader_test</span><span class="p">,</span> <span class="n">neval_batches</span><span class="o">=</span><span class="n">num_eval_batches</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Evaluation accuracy on </span><span class="si">%d</span><span class="s1"> images, </span><span class="si">%2.2f</span><span class="s1">'</span><span class="o">%</span><span class="p">(</span><span class="n">num_eval_batches</span> <span class="o">*</span> <span class="n">eval_batch_size</span><span class="p">,</span> <span class="n">top1</span><span class="o">.</span><span class="n">avg</span><span class="p">))</span>
<span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">save</span><span class="p">(</span><span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">script</span><span class="p">(</span><span class="n">per_channel_quantized_model</span><span class="p">),</span> <span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">scripted_quantized_model_file</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>QConfig(activation=functools.partial(<class 'torch.quantization.observer.HistogramObserver'>, reduce_range=True), weight=functools.partial(<class 'torch.quantization.observer.PerChannelMinMaxObserver'>, dtype=torch.qint8, qscheme=torch.per_channel_symmetric))
....................Evaluation accuracy on 300 images, 75.00
</pre></div>
</div>
<p>Changing just this quantization configuration method resulted in an increase
of the accuracy to over 76%! Still, this is 1-2% worse than the baseline of 78% achieved above.
So lets try quantization aware training.</p>
</div>
<div class="section" id="quantization-aware-training">
<h2>5. Quantization-aware training<a class="headerlink" href="#quantization-aware-training" title="Permalink to this headline">ยถ</a></h2>
<p>Quantization-aware training (QAT) is the quantization method that typically results in the highest accuracy.
With QAT, all weights and activations are โfake quantizedโ during both the forward and backward passes of
training: that is, float values are rounded to mimic int8 values, but all computations are still done with
floating point numbers. Thus, all the weight adjustments during training are made while โawareโ of the fact
that the model will ultimately be quantized; after quantizing, therefore, this method will usually yield
higher accuracy than either dynamic quantization or post-training static quantization.</p>
<p>The overall workflow for actually performing QAT is very similar to before:</p>
<ul class="simple">
<li><p>We can use the same model as before: there is no additional preparation needed for quantization-aware
training.</p></li>
<li><p>We need to use a <code class="docutils literal notranslate"><span class="pre">qconfig</span></code> specifying what kind of fake-quantization is to be inserted after weights
and activations, instead of specifying observers</p></li>
</ul>
<p>We first define a training function:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="k">def</span> <span class="nf">train_one_epoch</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">criterion</span><span class="p">,</span> <span class="n">optimizer</span><span class="p">,</span> <span class="n">data_loader</span><span class="p">,</span> <span class="n">device</span><span class="p">,</span> <span class="n">ntrain_batches</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="n">top1</span> <span class="o">=</span> <span class="n">AverageMeter</span><span class="p">(</span><span class="s1">'Acc@1'</span><span class="p">,</span> <span class="s1">':6.2f'</span><span class="p">)</span>
<span class="n">top5</span> <span class="o">=</span> <span class="n">AverageMeter</span><span class="p">(</span><span class="s1">'Acc@5'</span><span class="p">,</span> <span class="s1">':6.2f'</span><span class="p">)</span>
<span class="n">avgloss</span> <span class="o">=</span> <span class="n">AverageMeter</span><span class="p">(</span><span class="s1">'Loss'</span><span class="p">,</span> <span class="s1">'1.5f'</span><span class="p">)</span>
<span class="n">cnt</span> <span class="o">=</span> <span class="mi">0</span>
<span class="k">for</span> <span class="n">image</span><span class="p">,</span> <span class="n">target</span> <span class="ow">in</span> <span class="n">data_loader</span><span class="p">:</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="nb">print</span><span class="p">(</span><span class="s1">'.'</span><span class="p">,</span> <span class="n">end</span> <span class="o">=</span> <span class="s1">''</span><span class="p">)</span>
<span class="n">cnt</span> <span class="o">+=</span> <span class="mi">1</span>
<span class="n">image</span><span class="p">,</span> <span class="n">target</span> <span class="o">=</span> <span class="n">image</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">target</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">output</span> <span class="o">=</span> <span class="n">model</span><span class="p">(</span><span class="n">image</span><span class="p">)</span>
<span class="n">loss</span> <span class="o">=</span> <span class="n">criterion</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">)</span>
<span class="n">optimizer</span><span class="o">.</span><span class="n">zero_grad</span><span class="p">()</span>
<span class="n">loss</span><span class="o">.</span><span class="n">backward</span><span class="p">()</span>
<span class="n">optimizer</span><span class="o">.</span><span class="n">step</span><span class="p">()</span>
<span class="n">acc1</span><span class="p">,</span> <span class="n">acc5</span> <span class="o">=</span> <span class="n">accuracy</span><span class="p">(</span><span class="n">output</span><span class="p">,</span> <span class="n">target</span><span class="p">,</span> <span class="n">topk</span><span class="o">=</span><span class="p">(</span><span class="mi">1</span><span class="p">,</span> <span class="mi">5</span><span class="p">))</span>
<span class="n">top1</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">acc1</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">image</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="n">top5</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">acc5</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="n">image</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="n">avgloss</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">loss</span><span class="p">,</span> <span class="n">image</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="k">if</span> <span class="n">cnt</span> <span class="o">>=</span> <span class="n">ntrain_batches</span><span class="p">:</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Loss'</span><span class="p">,</span> <span class="n">avgloss</span><span class="o">.</span><span class="n">avg</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Training: * Acc@1 </span><span class="si">{top1.avg:.3f}</span><span class="s1"> Acc@5 </span><span class="si">{top5.avg:.3f}</span><span class="s1">'</span>
<span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">top1</span><span class="o">=</span><span class="n">top1</span><span class="p">,</span> <span class="n">top5</span><span class="o">=</span><span class="n">top5</span><span class="p">))</span>
<span class="k">return</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Full imagenet train set: * Acc@1 </span><span class="si">{top1.global_avg:.3f}</span><span class="s1"> Acc@5 </span><span class="si">{top5.global_avg:.3f}</span><span class="s1">'</span>
<span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">top1</span><span class="o">=</span><span class="n">top1</span><span class="p">,</span> <span class="n">top5</span><span class="o">=</span><span class="n">top5</span><span class="p">))</span>
<span class="k">return</span>
</pre></div>
</div>
<p>We fuse modules as before</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">qat_model</span> <span class="o">=</span> <span class="n">load_model</span><span class="p">(</span><span class="n">saved_model_dir</span> <span class="o">+</span> <span class="n">float_model_file</span><span class="p">)</span>
<span class="n">qat_model</span><span class="o">.</span><span class="n">fuse_model</span><span class="p">()</span>
<span class="n">optimizer</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">optim</span><span class="o">.</span><span class="n">SGD</span><span class="p">(</span><span class="n">qat_model</span><span class="o">.</span><span class="n">parameters</span><span class="p">(),</span> <span class="n">lr</span> <span class="o">=</span> <span class="mf">0.0001</span><span class="p">)</span>
<span class="n">qat_model</span><span class="o">.</span><span class="n">qconfig</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">get_default_qat_qconfig</span><span class="p">(</span><span class="s1">'fbgemm'</span><span class="p">)</span>
</pre></div>
</div>
<p>Finally, <code class="docutils literal notranslate"><span class="pre">prepare_qat</span></code> performs the โfake quantizationโ, preparing the model for quantization-aware
training</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">torch</span><span class="o">.</span><span class="n">quantization</span><span class="o">.</span><span class="n">prepare_qat</span><span class="p">(</span><span class="n">qat_model</span><span class="p">,</span> <span class="n">inplace</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="s1">'Inverted Residual Block: After preparation for QAT, note fake-quantization modules </span><span class="se">\n</span><span class="s1">'</span><span class="p">,</span><span class="n">qat_model</span><span class="o">.</span><span class="n">features</span><span class="p">[</span><span class="mi">1</span><span class="p">]</span><span class="o">.</span><span class="n">conv</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>Inverted Residual Block: After preparation for QAT, note fake-quantization modules
Sequential(
(0): ConvBNReLU(
(0): ConvBnReLU2d(
32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=32, bias=False
(observer): FakeQuantize(
fake_quant_enabled=True, observer_enabled=True, scale=None, zero_point=None
(observer): MovingAverageMinMaxObserver(min_val=None, max_val=None)
)
(weight_fake_quant): FakeQuantize(
fake_quant_enabled=True, observer_enabled=True, scale=None, zero_point=None
(observer): MovingAveragePerChannelMinMaxObserver(min_val=None, max_val=None)
)
)
(1): Identity()
(2): Identity()
)
(1): ConvBn2d(
32, 16, kernel_size=(1, 1), stride=(1, 1), bias=False
(observer): FakeQuantize(
fake_quant_enabled=True, observer_enabled=True, scale=None, zero_point=None
(observer): MovingAverageMinMaxObserver(min_val=None, max_val=None)
)
(weight_fake_quant): FakeQuantize(
fake_quant_enabled=True, observer_enabled=True, scale=None, zero_point=None
(observer): MovingAveragePerChannelMinMaxObserver(min_val=None, max_val=None)
)
)
(2): Identity()
)
</pre></div>
</div>
<p>Training a quantized model with high accuracy requires accurate modeling of numerics at
inference. For quantization aware training, therefore, we modify the training loop by:</p>
<ul class="simple">
<li><p>Switch batch norm to use running mean and variance towards the end of training to better
match inference numerics.</p></li>
<li><p>We also freeze the quantizer parameters (scale and zero-point) and fine tune the weights.</p></li>
</ul>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">num_train_batches</span> <span class="o">=</span> <span class="mi">20</span>