-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathquantization-support.html
1713 lines (1515 loc) · 191 KB
/
quantization-support.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<!--[if IE 8]><html class="no-js lt-ie9" lang="en" > <![endif]-->
<!--[if gt IE 8]><!--> <html class="no-js" lang="en" > <!--<![endif]-->
<head>
<meta name="robots" content="noindex">
<meta charset="utf-8">
<meta name="generator" content="Docutils 0.18.1: http://docutils.sourceforge.net/" />
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Quantization API Reference — PyTorch 2.0 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/quantization-support.html"/>
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<!-- <link rel="stylesheet" href="_static/pygments.css" type="text/css" /> -->
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="_static/css/theme.css" type="text/css" />
<link rel="stylesheet" href="_static/copybutton.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/katex.min.css" type="text/css" />
<link rel="stylesheet" href="_static/katex-math.css" type="text/css" />
<link rel="stylesheet" href="_static/sphinx-dropdown.css" type="text/css" />
<link rel="stylesheet" href="_static/panels-bootstrap.min.css" type="text/css" />
<link rel="stylesheet" href="_static/css/jit.css" type="text/css" />
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="quantize" href="generated/torch.ao.quantization.quantize.html" />
<link rel="prev" title="Quantization" href="quantization.html" />
<!-- Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-117752657-2"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-117752657-2');
</script>
<!-- End Google Analytics -->
<script src="_static/js/modernizr.min.js"></script>
<!-- Preload the theme fonts -->
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-book.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-Medium.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/FreightSans/freight-sans-medium-italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="_static/fonts/IBMPlexMono/IBMPlexMono-SemiBold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<!-- Preload the katex fonts -->
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Math-Italic.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Main-Bold.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size1-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size4-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size2-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Size3-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="preload" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/fonts/KaTeX_Caligraphic-Regular.woff2" as="font" type="font/woff2" crossorigin="anonymous">
<link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.15.2/css/all.css" integrity="sha384-vSIIfh2YWi9wW0r9iZe7RJPrKwp6bG+s9QZMoITbCckVJqGCCRhc+ccxNcdpHuYu" crossorigin="anonymous">
</head>
<div class="container-fluid header-holder tutorials-header" id="header-holder">
<div class="container">
<div class="header-container">
<a class="header-logo" href="https://pytorch.org/" aria-label="PyTorch"></a>
<div class="main-menu">
<ul>
<li>
<a href="https://pytorch.org/get-started">Get Started</a>
</li>
<li>
<a href="https://pytorch.org/ecosystem">Ecosystem</a>
</li>
<li>
<a href="https://pytorch.org/mobile">Mobile</a>
</li>
<li>
<a href="https://pytorch.org/blog/">Blog</a>
</li>
<li>
<a href="https://pytorch.org/tutorials">Tutorials</a>
</li>
<li class="active docs-active">
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-orange-arrow">
Docs
</a>
<div class="resources-dropdown-menu">
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/docs/stable/index.html">
<span class="dropdown-title">PyTorch</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/audio/stable/index.html">
<span class="dropdown-title">torchaudio</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/text/stable/index.html">
<span class="dropdown-title">torchtext</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/vision/stable/index.html">
<span class="dropdown-title">torchvision</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torcharrow">
<span class="dropdown-title">torcharrow</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/data">
<span class="dropdown-title">TorchData</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torchrec">
<span class="dropdown-title">TorchRec</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/serve/">
<span class="dropdown-title">TorchServe</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/torchx/">
<span class="dropdown-title">TorchX</span>
<p></p>
</a>
<a class="doc-dropdown-option nav-dropdown-item" href="https://pytorch.org/xla">
<span class="dropdown-title">PyTorch on XLA Devices</span>
<p></p>
</a>
</div>
</li>
<li>
<div id="resourcesDropdownButton" data-toggle="resources-dropdown" class="resources-dropdown">
<a class="resource-option with-down-arrow">
Resources
</a>
<div class="resources-dropdown-menu">
<a class="nav-dropdown-item" href="https://pytorch.org/features">
<span class="dropdown-title">About</span>
<p>Learn about PyTorch’s features and capabilities</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/foundation">
<span class="dropdown-title">PyTorch Foundation</span>
<p>Learn about the PyTorch foundation</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/#community-module">
<span class="dropdown-title">Community</span>
<p>Join the PyTorch developer community to contribute, learn, and get your questions answered.</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/community-stories">
<span class="dropdown-title">Community Stories</span>
<p>Learn how our community solves real, everyday machine learning problems with PyTorch.</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/resources">
<span class="dropdown-title">Developer Resources</span>
<p>Find resources and get questions answered</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/events">
<span class="dropdown-title">Events</span>
<p>Find events, webinars, and podcasts</p>
</a>
<a class="nav-dropdown-item" href="https://discuss.pytorch.org/" target="_blank">
<span class="dropdown-title">Forums</span>
<p>A place to discuss PyTorch code, issues, install, research</p>
</a>
<a class="nav-dropdown-item" href="https://pytorch.org/hub">
<span class="dropdown-title">Models (Beta)</span>
<p>Discover, publish, and reuse pre-trained models</p>
</a>
</div>
</div>
</li>
<li>
<a href="https://github.com/pytorch/pytorch">GitHub</a>
</li>
</ul>
</div>
<a class="main-menu-open-button" href="#" data-behavior="open-mobile-menu"></a>
</div>
</div>
</div>
<body class="pytorch-body">
<div class="table-of-contents-link-wrapper">
<span>Table of Contents</span>
<a href="#" class="toggle-table-of-contents" data-behavior="toggle-table-of-contents"></a>
</div>
<nav data-toggle="wy-nav-shift" class="pytorch-left-menu" id="pytorch-left-menu">
<div class="pytorch-side-scroll">
<div class="pytorch-menu pytorch-menu-vertical" data-spy="affix" role="navigation" aria-label="main navigation">
<div class="pytorch-left-menu-search">
<div class="version">
<a href='https://pytorch.org/docs/versions.html'>2.0 ▼</a>
</div>
<div role="search">
<form id="rtd-search-form" class="wy-form" action="search.html" method="get">
<input type="text" name="q" placeholder="Search Docs" />
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
</div>
<p class="caption" role="heading"><span class="caption-text">Community</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="community/build_ci_governance.html">PyTorch Governance | Build + CI</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/contribution_guide.html">PyTorch Contribution Guide</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/design.html">PyTorch Design Philosophy</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/governance.html">PyTorch Governance | Mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="community/persons_of_interest.html">PyTorch Governance | Maintainers</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Developer Notes</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="notes/amp_examples.html">CUDA Automatic Mixed Precision examples</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/autograd.html">Autograd mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/broadcasting.html">Broadcasting semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cpu_threading_torchscript_inference.html">CPU threading and TorchScript inference</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/cuda.html">CUDA semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/ddp.html">Distributed Data Parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/extending.html">Extending PyTorch</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/extending.func.html">Extending torch.func with autograd.Function</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/faq.html">Frequently Asked Questions</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/gradcheck.html">Gradcheck mechanics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/hip.html">HIP (ROCm) semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/large_scale_deployments.html">Features for large-scale deployments</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/modules.html">Modules</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/mps.html">MPS backend</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/multiprocessing.html">Multiprocessing best practices</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/numerical_accuracy.html">Numerical accuracy</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/randomness.html">Reproducibility</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/serialization.html">Serialization semantics</a></li>
<li class="toctree-l1"><a class="reference internal" href="notes/windows.html">Windows FAQ</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">torch.compile</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="dynamo/index.html">TorchDynamo Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/installation.html">Installing TorchDynamo</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/get-started.html">Getting Started</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/guards-overview.html">Guards Overview</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/custom-backends.html">Custom Backends</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/deep-dive.html">TorchDynamo Deeper Dive</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/troubleshooting.html">TorchDynamo Troubleshooting</a></li>
<li class="toctree-l1"><a class="reference internal" href="dynamo/faq.html">Frequently Asked Questions</a></li>
<li class="toctree-l1"><a class="reference internal" href="ir.html">IRs</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Language Bindings</span></p>
<ul>
<li class="toctree-l1"><a class="reference internal" href="cpp_index.html">C++</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/javadoc/">Javadoc</a></li>
<li class="toctree-l1"><a class="reference internal" href="deploy.html">torch::deploy</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Python API</span></p>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="torch.html">torch</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.html">torch.nn</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.functional.html">torch.nn.functional</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensors.html">torch.Tensor</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_attributes.html">Tensor Attributes</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensor_view.html">Tensor Views</a></li>
<li class="toctree-l1"><a class="reference internal" href="amp.html">torch.amp</a></li>
<li class="toctree-l1"><a class="reference internal" href="autograd.html">torch.autograd</a></li>
<li class="toctree-l1"><a class="reference internal" href="library.html">torch.library</a></li>
<li class="toctree-l1"><a class="reference internal" href="cuda.html">torch.cuda</a></li>
<li class="toctree-l1"><a class="reference internal" href="mps.html">torch.mps</a></li>
<li class="toctree-l1"><a class="reference internal" href="backends.html">torch.backends</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.html">torch.distributed</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.algorithms.join.html">torch.distributed.algorithms.join</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.elastic.html">torch.distributed.elastic</a></li>
<li class="toctree-l1"><a class="reference internal" href="fsdp.html">torch.distributed.fsdp</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.optim.html">torch.distributed.optim</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.tensor.parallel.html">torch.distributed.tensor.parallel</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributed.checkpoint.html">torch.distributed.checkpoint</a></li>
<li class="toctree-l1"><a class="reference internal" href="distributions.html">torch.distributions</a></li>
<li class="toctree-l1"><a class="reference internal" href="_dynamo.html">torch._dynamo</a></li>
<li class="toctree-l1"><a class="reference internal" href="fft.html">torch.fft</a></li>
<li class="toctree-l1"><a class="reference internal" href="func.html">torch.func</a></li>
<li class="toctree-l1"><a class="reference internal" href="futures.html">torch.futures</a></li>
<li class="toctree-l1"><a class="reference internal" href="fx.html">torch.fx</a></li>
<li class="toctree-l1"><a class="reference internal" href="hub.html">torch.hub</a></li>
<li class="toctree-l1"><a class="reference internal" href="jit.html">torch.jit</a></li>
<li class="toctree-l1"><a class="reference internal" href="linalg.html">torch.linalg</a></li>
<li class="toctree-l1"><a class="reference internal" href="monitor.html">torch.monitor</a></li>
<li class="toctree-l1"><a class="reference internal" href="signal.html">torch.signal</a></li>
<li class="toctree-l1"><a class="reference internal" href="special.html">torch.special</a></li>
<li class="toctree-l1"><a class="reference internal" href="torch.overrides.html">torch.overrides</a></li>
<li class="toctree-l1"><a class="reference internal" href="package.html">torch.package</a></li>
<li class="toctree-l1"><a class="reference internal" href="profiler.html">torch.profiler</a></li>
<li class="toctree-l1"><a class="reference internal" href="nn.init.html">torch.nn.init</a></li>
<li class="toctree-l1"><a class="reference internal" href="onnx.html">torch.onnx</a></li>
<li class="toctree-l1"><a class="reference internal" href="onnx_diagnostics.html">torch.onnx diagnostics</a></li>
<li class="toctree-l1"><a class="reference internal" href="optim.html">torch.optim</a></li>
<li class="toctree-l1"><a class="reference internal" href="complex_numbers.html">Complex Numbers</a></li>
<li class="toctree-l1"><a class="reference internal" href="ddp_comm_hooks.html">DDP Communication Hooks</a></li>
<li class="toctree-l1"><a class="reference internal" href="pipeline.html">Pipeline Parallelism</a></li>
<li class="toctree-l1 current"><a class="reference internal" href="quantization.html">Quantization</a></li>
<li class="toctree-l1"><a class="reference internal" href="rpc.html">Distributed RPC Framework</a></li>
<li class="toctree-l1"><a class="reference internal" href="random.html">torch.random</a></li>
<li class="toctree-l1"><a class="reference internal" href="masked.html">torch.masked</a></li>
<li class="toctree-l1"><a class="reference internal" href="nested.html">torch.nested</a></li>
<li class="toctree-l1"><a class="reference internal" href="sparse.html">torch.sparse</a></li>
<li class="toctree-l1"><a class="reference internal" href="storage.html">torch.Storage</a></li>
<li class="toctree-l1"><a class="reference internal" href="testing.html">torch.testing</a></li>
<li class="toctree-l1"><a class="reference internal" href="benchmark_utils.html">torch.utils.benchmark</a></li>
<li class="toctree-l1"><a class="reference internal" href="bottleneck.html">torch.utils.bottleneck</a></li>
<li class="toctree-l1"><a class="reference internal" href="checkpoint.html">torch.utils.checkpoint</a></li>
<li class="toctree-l1"><a class="reference internal" href="cpp_extension.html">torch.utils.cpp_extension</a></li>
<li class="toctree-l1"><a class="reference internal" href="data.html">torch.utils.data</a></li>
<li class="toctree-l1"><a class="reference internal" href="jit_utils.html">torch.utils.jit</a></li>
<li class="toctree-l1"><a class="reference internal" href="dlpack.html">torch.utils.dlpack</a></li>
<li class="toctree-l1"><a class="reference internal" href="mobile_optimizer.html">torch.utils.mobile_optimizer</a></li>
<li class="toctree-l1"><a class="reference internal" href="model_zoo.html">torch.utils.model_zoo</a></li>
<li class="toctree-l1"><a class="reference internal" href="tensorboard.html">torch.utils.tensorboard</a></li>
<li class="toctree-l1"><a class="reference internal" href="type_info.html">Type Info</a></li>
<li class="toctree-l1"><a class="reference internal" href="named_tensor.html">Named Tensors</a></li>
<li class="toctree-l1"><a class="reference internal" href="name_inference.html">Named Tensors operator coverage</a></li>
<li class="toctree-l1"><a class="reference internal" href="config_mod.html">torch.__config__</a></li>
</ul>
<p class="caption" role="heading"><span class="caption-text">Libraries</span></p>
<ul>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/audio/stable">torchaudio</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/data">TorchData</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/torchrec">TorchRec</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/serve">TorchServe</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/text/stable">torchtext</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/vision/stable">torchvision</a></li>
<li class="toctree-l1"><a class="reference external" href="https://pytorch.org/xla/">PyTorch on XLA Devices</a></li>
</ul>
</div>
</div>
</nav>
<div class="pytorch-container">
<div class="pytorch-page-level-bar" id="pytorch-page-level-bar">
<div class="pytorch-breadcrumbs-wrapper">
<div role="navigation" aria-label="breadcrumbs navigation">
<ul class="pytorch-breadcrumbs">
<li>
<a href="index.html">
Docs
</a> >
</li>
<li><a href="quantization.html">Quantization</a> ></li>
<li>Quantization API Reference</li>
<li class="pytorch-breadcrumbs-aside">
<a href="_sources/quantization-support.rst.txt" rel="nofollow"><img src="_static/images/view-page-source-icon.svg"></a>
</li>
</ul>
</div>
</div>
<div class="pytorch-shortcuts-wrapper" id="pytorch-shortcuts-wrapper">
Shortcuts
</div>
</div>
<section data-toggle="wy-nav-shift" id="pytorch-content-wrap" class="pytorch-content-wrap">
<div class="pytorch-content-left">
<div class="rst-content">
<div role="main" class="main-content" itemscope="itemscope" itemtype="http://schema.org/Article">
<article itemprop="articleBody" id="pytorch-article" class="pytorch-article">
<section id="quantization-api-reference">
<h1>Quantization API Reference<a class="headerlink" href="#quantization-api-reference" title="Permalink to this heading">¶</a></h1>
<section id="torch-ao-quantization">
<h2>torch.ao.quantization<a class="headerlink" href="#torch-ao-quantization" title="Permalink to this heading">¶</a></h2>
<p>This module contains Eager mode quantization APIs.</p>
<section id="top-level-apis">
<h3>Top level APIs<a class="headerlink" href="#top-level-apis" title="Permalink to this heading">¶</a></h3>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.quantize"/><a class="reference internal" href="generated/torch.ao.quantization.quantize.html#torch.ao.quantization.quantize" title="torch.ao.quantization.quantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">quantize</span></code></a></p></td>
<td><p>Quantize the input float model with post training static quantization.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.quantize_dynamic"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_dynamic.html#torch.ao.quantization.quantize_dynamic" title="torch.ao.quantization.quantize_dynamic"><code class="xref py py-obj docutils literal notranslate"><span class="pre">quantize_dynamic</span></code></a></p></td>
<td><p>Converts a float model to dynamic (i.e.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.quantize_qat"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_qat.html#torch.ao.quantization.quantize_qat" title="torch.ao.quantization.quantize_qat"><code class="xref py py-obj docutils literal notranslate"><span class="pre">quantize_qat</span></code></a></p></td>
<td><p>Do quantization aware training and output a quantized model</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.prepare"/><a class="reference internal" href="generated/torch.ao.quantization.prepare.html#torch.ao.quantization.prepare" title="torch.ao.quantization.prepare"><code class="xref py py-obj docutils literal notranslate"><span class="pre">prepare</span></code></a></p></td>
<td><p>Prepares a copy of the model for quantization calibration or quantization-aware training.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.prepare_qat"/><a class="reference internal" href="generated/torch.ao.quantization.prepare_qat.html#torch.ao.quantization.prepare_qat" title="torch.ao.quantization.prepare_qat"><code class="xref py py-obj docutils literal notranslate"><span class="pre">prepare_qat</span></code></a></p></td>
<td><p>Prepares a copy of the model for quantization calibration or quantization-aware training and converts it to quantized version.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.convert"/><a class="reference internal" href="generated/torch.ao.quantization.convert.html#torch.ao.quantization.convert" title="torch.ao.quantization.convert"><code class="xref py py-obj docutils literal notranslate"><span class="pre">convert</span></code></a></p></td>
<td><p>Converts submodules in input module to a different module according to <cite>mapping</cite> by calling <cite>from_float</cite> method on the target module class.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="preparing-model-for-quantization">
<h3>Preparing model for quantization<a class="headerlink" href="#preparing-model-for-quantization" title="Permalink to this heading">¶</a></h3>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fuse_modules"/><a class="reference internal" href="generated/torch.ao.quantization.fuse_modules.html#torch.ao.quantization.fuse_modules" title="torch.ao.quantization.fuse_modules"><code class="xref py py-obj docutils literal notranslate"><span class="pre">fuse_modules</span></code></a></p></td>
<td><p>Fuses a list of modules into a single module</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.QuantStub"/><a class="reference internal" href="generated/torch.ao.quantization.QuantStub.html#torch.ao.quantization.QuantStub" title="torch.ao.quantization.QuantStub"><code class="xref py py-obj docutils literal notranslate"><span class="pre">QuantStub</span></code></a></p></td>
<td><p>Quantize stub module, before calibration, this is same as an observer, it will be swapped as <cite>nnq.Quantize</cite> in <cite>convert</cite>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.DeQuantStub"/><a class="reference internal" href="generated/torch.ao.quantization.DeQuantStub.html#torch.ao.quantization.DeQuantStub" title="torch.ao.quantization.DeQuantStub"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DeQuantStub</span></code></a></p></td>
<td><p>Dequantize stub module, before calibration, this is same as identity, this will be swapped as <cite>nnq.DeQuantize</cite> in <cite>convert</cite>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.QuantWrapper"/><a class="reference internal" href="generated/torch.ao.quantization.QuantWrapper.html#torch.ao.quantization.QuantWrapper" title="torch.ao.quantization.QuantWrapper"><code class="xref py py-obj docutils literal notranslate"><span class="pre">QuantWrapper</span></code></a></p></td>
<td><p>A wrapper class that wraps the input module, adds QuantStub and DeQuantStub and surround the call to module with call to quant and dequant modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.add_quant_dequant"/><a class="reference internal" href="generated/torch.ao.quantization.add_quant_dequant.html#torch.ao.quantization.add_quant_dequant" title="torch.ao.quantization.add_quant_dequant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">add_quant_dequant</span></code></a></p></td>
<td><p>Wrap the leaf child module in QuantWrapper if it has a valid qconfig Note that this function will modify the children of module inplace and it can return a new module which wraps the input module as well.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="utility-functions">
<h3>Utility functions<a class="headerlink" href="#utility-functions" title="Permalink to this heading">¶</a></h3>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.swap_module"/><a class="reference internal" href="generated/torch.ao.quantization.swap_module.html#torch.ao.quantization.swap_module" title="torch.ao.quantization.swap_module"><code class="xref py py-obj docutils literal notranslate"><span class="pre">swap_module</span></code></a></p></td>
<td><p>Swaps the module if it has a quantized counterpart and it has an <cite>observer</cite> attached.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.propagate_qconfig_"/><a class="reference internal" href="generated/torch.ao.quantization.propagate_qconfig_.html#torch.ao.quantization.propagate_qconfig_" title="torch.ao.quantization.propagate_qconfig_"><code class="xref py py-obj docutils literal notranslate"><span class="pre">propagate_qconfig_</span></code></a></p></td>
<td><p>Propagate qconfig through the module hierarchy and assign <cite>qconfig</cite> attribute on each leaf module</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.default_eval_fn"/><a class="reference internal" href="generated/torch.ao.quantization.default_eval_fn.html#torch.ao.quantization.default_eval_fn" title="torch.ao.quantization.default_eval_fn"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_eval_fn</span></code></a></p></td>
<td><p>Default evaluation function takes a torch.utils.data.Dataset or a list of input Tensors and run the model on the dataset</p></td>
</tr>
</tbody>
</table>
</section>
</section>
<section id="torch-ao-quantization-quantize-fx">
<h2>torch.ao.quantization.quantize_fx<a class="headerlink" href="#torch-ao-quantization-quantize-fx" title="Permalink to this heading">¶</a></h2>
<p>This module contains FX graph mode quantization APIs (prototype).</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.quantize_fx.prepare_fx"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.prepare_fx.html#torch.ao.quantization.quantize_fx.prepare_fx" title="torch.ao.quantization.quantize_fx.prepare_fx"><code class="xref py py-obj docutils literal notranslate"><span class="pre">prepare_fx</span></code></a></p></td>
<td><p>Prepare a model for post training static quantization</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.quantize_fx.prepare_qat_fx"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.prepare_qat_fx.html#torch.ao.quantization.quantize_fx.prepare_qat_fx" title="torch.ao.quantization.quantize_fx.prepare_qat_fx"><code class="xref py py-obj docutils literal notranslate"><span class="pre">prepare_qat_fx</span></code></a></p></td>
<td><p>Prepare a model for quantization aware training</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.quantize_fx.convert_fx"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.convert_fx.html#torch.ao.quantization.quantize_fx.convert_fx" title="torch.ao.quantization.quantize_fx.convert_fx"><code class="xref py py-obj docutils literal notranslate"><span class="pre">convert_fx</span></code></a></p></td>
<td><p>Convert a calibrated or trained model to a quantized model</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.quantize_fx.fuse_fx"/><a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.fuse_fx.html#torch.ao.quantization.quantize_fx.fuse_fx" title="torch.ao.quantization.quantize_fx.fuse_fx"><code class="xref py py-obj docutils literal notranslate"><span class="pre">fuse_fx</span></code></a></p></td>
<td><p>Fuse modules like conv+bn, conv+bn+relu etc, model must be in eval mode.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-qconfig-mapping">
<h2>torch.ao.quantization.qconfig_mapping<a class="headerlink" href="#torch-ao-quantization-qconfig-mapping" title="Permalink to this heading">¶</a></h2>
<p>This module contains QConfigMapping for configuring FX graph mode quantization.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig_mapping.QConfigMapping"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig_mapping.QConfigMapping.html#torch.ao.quantization.qconfig_mapping.QConfigMapping" title="torch.ao.quantization.qconfig_mapping.QConfigMapping"><code class="xref py py-obj docutils literal notranslate"><span class="pre">QConfigMapping</span></code></a></p></td>
<td><p>Mapping from model ops to <code class="xref py py-class docutils literal notranslate"><span class="pre">torch.ao.quantization.QConfig</span></code> s.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig_mapping.get_default_qconfig_mapping"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig_mapping.get_default_qconfig_mapping.html#torch.ao.quantization.qconfig_mapping.get_default_qconfig_mapping" title="torch.ao.quantization.qconfig_mapping.get_default_qconfig_mapping"><code class="xref py py-obj docutils literal notranslate"><span class="pre">get_default_qconfig_mapping</span></code></a></p></td>
<td><p>Return the default QConfigMapping for post training quantization.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig_mapping.get_default_qat_qconfig_mapping"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig_mapping.get_default_qat_qconfig_mapping.html#torch.ao.quantization.qconfig_mapping.get_default_qat_qconfig_mapping" title="torch.ao.quantization.qconfig_mapping.get_default_qat_qconfig_mapping"><code class="xref py py-obj docutils literal notranslate"><span class="pre">get_default_qat_qconfig_mapping</span></code></a></p></td>
<td><p>Return the default QConfigMapping for quantization aware training.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-backend-config">
<h2>torch.ao.quantization.backend_config<a class="headerlink" href="#torch-ao-quantization-backend-config" title="Permalink to this heading">¶</a></h2>
<p>This module contains BackendConfig, a config object that defines how quantization is supported
in a backend. Currently only used by FX Graph Mode Quantization, but we may extend Eager Mode
Quantization to work with this as well.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.backend_config.BackendConfig"/><a class="reference internal" href="generated/torch.ao.quantization.backend_config.BackendConfig.html#torch.ao.quantization.backend_config.BackendConfig" title="torch.ao.quantization.backend_config.BackendConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BackendConfig</span></code></a></p></td>
<td><p>Config that defines the set of patterns that can be quantized on a given backend, and how reference quantized models can be produced from these patterns.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.backend_config.BackendPatternConfig"/><a class="reference internal" href="generated/torch.ao.quantization.backend_config.BackendPatternConfig.html#torch.ao.quantization.backend_config.BackendPatternConfig" title="torch.ao.quantization.backend_config.BackendPatternConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BackendPatternConfig</span></code></a></p></td>
<td><p>Config object that specifies quantization behavior for a given operator pattern.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.backend_config.DTypeConfig"/><a class="reference internal" href="generated/torch.ao.quantization.backend_config.DTypeConfig.html#torch.ao.quantization.backend_config.DTypeConfig" title="torch.ao.quantization.backend_config.DTypeConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DTypeConfig</span></code></a></p></td>
<td><p>Config object that specifies the supported data types passed as arguments to quantize ops in the reference model spec, for input and output activations, weights, and biases.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.backend_config.DTypeWithConstraints"/><a class="reference internal" href="generated/torch.ao.quantization.backend_config.DTypeWithConstraints.html#torch.ao.quantization.backend_config.DTypeWithConstraints" title="torch.ao.quantization.backend_config.DTypeWithConstraints"><code class="xref py py-obj docutils literal notranslate"><span class="pre">DTypeWithConstraints</span></code></a></p></td>
<td><p>Config for specifying additional constraints for a given dtype, such as quantization value ranges, scale value ranges, and fixed quantization params, to be used in <a class="reference internal" href="generated/torch.ao.quantization.backend_config.DTypeConfig.html#torch.ao.quantization.backend_config.DTypeConfig" title="torch.ao.quantization.backend_config.DTypeConfig"><code class="xref py py-class docutils literal notranslate"><span class="pre">DTypeConfig</span></code></a>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.backend_config.ObservationType"/><a class="reference internal" href="generated/torch.ao.quantization.backend_config.ObservationType.html#torch.ao.quantization.backend_config.ObservationType" title="torch.ao.quantization.backend_config.ObservationType"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ObservationType</span></code></a></p></td>
<td><p>An enum that represents different ways of how an operator/operator pattern should be observed</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-fx-custom-config">
<h2>torch.ao.quantization.fx.custom_config<a class="headerlink" href="#torch-ao-quantization-fx-custom-config" title="Permalink to this heading">¶</a></h2>
<p>This module contains a few CustomConfig classes that’s used in both eager mode and FX graph mode quantization</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fx.custom_config.FuseCustomConfig"/><a class="reference internal" href="generated/torch.ao.quantization.fx.custom_config.FuseCustomConfig.html#torch.ao.quantization.fx.custom_config.FuseCustomConfig" title="torch.ao.quantization.fx.custom_config.FuseCustomConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FuseCustomConfig</span></code></a></p></td>
<td><p>Custom configuration for <a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.fuse_fx.html#torch.ao.quantization.quantize_fx.fuse_fx" title="torch.ao.quantization.quantize_fx.fuse_fx"><code class="xref py py-func docutils literal notranslate"><span class="pre">fuse_fx()</span></code></a>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fx.custom_config.PrepareCustomConfig"/><a class="reference internal" href="generated/torch.ao.quantization.fx.custom_config.PrepareCustomConfig.html#torch.ao.quantization.fx.custom_config.PrepareCustomConfig" title="torch.ao.quantization.fx.custom_config.PrepareCustomConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">PrepareCustomConfig</span></code></a></p></td>
<td><p>Custom configuration for <a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.prepare_fx.html#torch.ao.quantization.quantize_fx.prepare_fx" title="torch.ao.quantization.quantize_fx.prepare_fx"><code class="xref py py-func docutils literal notranslate"><span class="pre">prepare_fx()</span></code></a> and <a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.prepare_qat_fx.html#torch.ao.quantization.quantize_fx.prepare_qat_fx" title="torch.ao.quantization.quantize_fx.prepare_qat_fx"><code class="xref py py-func docutils literal notranslate"><span class="pre">prepare_qat_fx()</span></code></a>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fx.custom_config.ConvertCustomConfig"/><a class="reference internal" href="generated/torch.ao.quantization.fx.custom_config.ConvertCustomConfig.html#torch.ao.quantization.fx.custom_config.ConvertCustomConfig" title="torch.ao.quantization.fx.custom_config.ConvertCustomConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvertCustomConfig</span></code></a></p></td>
<td><p>Custom configuration for <a class="reference internal" href="generated/torch.ao.quantization.quantize_fx.convert_fx.html#torch.ao.quantization.quantize_fx.convert_fx" title="torch.ao.quantization.quantize_fx.convert_fx"><code class="xref py py-func docutils literal notranslate"><span class="pre">convert_fx()</span></code></a>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fx.custom_config.StandaloneModuleConfigEntry"/><a class="reference internal" href="generated/torch.ao.quantization.fx.custom_config.StandaloneModuleConfigEntry.html#torch.ao.quantization.fx.custom_config.StandaloneModuleConfigEntry" title="torch.ao.quantization.fx.custom_config.StandaloneModuleConfigEntry"><code class="xref py py-obj docutils literal notranslate"><span class="pre">StandaloneModuleConfigEntry</span></code></a></p></td>
<td><p></p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-quantization-related-functions">
<h2>torch (quantization related functions)<a class="headerlink" href="#torch-quantization-related-functions" title="Permalink to this heading">¶</a></h2>
<p>This describes the quantization related functions of the <cite>torch</cite> namespace.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.quantize_per_tensor"/><a class="reference internal" href="generated/torch.quantize_per_tensor.html#torch.quantize_per_tensor" title="torch.quantize_per_tensor"><code class="xref py py-obj docutils literal notranslate"><span class="pre">quantize_per_tensor</span></code></a></p></td>
<td><p>Converts a float tensor to a quantized tensor with given scale and zero point.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.quantize_per_channel"/><a class="reference internal" href="generated/torch.quantize_per_channel.html#torch.quantize_per_channel" title="torch.quantize_per_channel"><code class="xref py py-obj docutils literal notranslate"><span class="pre">quantize_per_channel</span></code></a></p></td>
<td><p>Converts a float tensor to a per-channel quantized tensor with given scales and zero points.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.dequantize"/><a class="reference internal" href="generated/torch.dequantize.html#torch.dequantize" title="torch.dequantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">dequantize</span></code></a></p></td>
<td><p>Returns an fp32 Tensor by dequantizing a quantized Tensor</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-tensor-quantization-related-methods">
<h2>torch.Tensor (quantization related methods)<a class="headerlink" href="#torch-tensor-quantization-related-methods" title="Permalink to this heading">¶</a></h2>
<p>Quantized Tensors support a limited subset of data manipulation methods of the
regular full-precision tensor.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.Tensor.view"/><a class="reference internal" href="generated/torch.Tensor.view.html#torch.Tensor.view" title="torch.Tensor.view"><code class="xref py py-obj docutils literal notranslate"><span class="pre">view</span></code></a></p></td>
<td><p>Returns a new tensor with the same data as the <code class="xref py py-attr docutils literal notranslate"><span class="pre">self</span></code> tensor but of a different <code class="xref py py-attr docutils literal notranslate"><span class="pre">shape</span></code>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.as_strided"/><a class="reference internal" href="generated/torch.Tensor.as_strided.html#torch.Tensor.as_strided" title="torch.Tensor.as_strided"><code class="xref py py-obj docutils literal notranslate"><span class="pre">as_strided</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.as_strided.html#torch.as_strided" title="torch.as_strided"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.as_strided()</span></code></a></p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.expand"/><a class="reference internal" href="generated/torch.Tensor.expand.html#torch.Tensor.expand" title="torch.Tensor.expand"><code class="xref py py-obj docutils literal notranslate"><span class="pre">expand</span></code></a></p></td>
<td><p>Returns a new view of the <code class="xref py py-attr docutils literal notranslate"><span class="pre">self</span></code> tensor with singleton dimensions expanded to a larger size.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.flatten"/><a class="reference internal" href="generated/torch.Tensor.flatten.html#torch.Tensor.flatten" title="torch.Tensor.flatten"><code class="xref py py-obj docutils literal notranslate"><span class="pre">flatten</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.flatten.html#torch.flatten" title="torch.flatten"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.flatten()</span></code></a></p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.select"/><a class="reference internal" href="generated/torch.Tensor.select.html#torch.Tensor.select" title="torch.Tensor.select"><code class="xref py py-obj docutils literal notranslate"><span class="pre">select</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.select.html#torch.select" title="torch.select"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.select()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.ne"/><a class="reference internal" href="generated/torch.Tensor.ne.html#torch.Tensor.ne" title="torch.Tensor.ne"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ne</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.ne.html#torch.ne" title="torch.ne"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.ne()</span></code></a>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.eq"/><a class="reference internal" href="generated/torch.Tensor.eq.html#torch.Tensor.eq" title="torch.Tensor.eq"><code class="xref py py-obj docutils literal notranslate"><span class="pre">eq</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.eq.html#torch.eq" title="torch.eq"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.eq()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.ge"/><a class="reference internal" href="generated/torch.Tensor.ge.html#torch.Tensor.ge" title="torch.Tensor.ge"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ge</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.ge.html#torch.ge" title="torch.ge"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.ge()</span></code></a>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.le"/><a class="reference internal" href="generated/torch.Tensor.le.html#torch.Tensor.le" title="torch.Tensor.le"><code class="xref py py-obj docutils literal notranslate"><span class="pre">le</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.le.html#torch.le" title="torch.le"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.le()</span></code></a>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.gt"/><a class="reference internal" href="generated/torch.Tensor.gt.html#torch.Tensor.gt" title="torch.Tensor.gt"><code class="xref py py-obj docutils literal notranslate"><span class="pre">gt</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.gt.html#torch.gt" title="torch.gt"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.gt()</span></code></a>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.lt"/><a class="reference internal" href="generated/torch.Tensor.lt.html#torch.Tensor.lt" title="torch.Tensor.lt"><code class="xref py py-obj docutils literal notranslate"><span class="pre">lt</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.lt.html#torch.lt" title="torch.lt"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.lt()</span></code></a>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.copy_"/><a class="reference internal" href="generated/torch.Tensor.copy_.html#torch.Tensor.copy_" title="torch.Tensor.copy_"><code class="xref py py-obj docutils literal notranslate"><span class="pre">copy_</span></code></a></p></td>
<td><p>Copies the elements from <code class="xref py py-attr docutils literal notranslate"><span class="pre">src</span></code> into <code class="xref py py-attr docutils literal notranslate"><span class="pre">self</span></code> tensor and returns <code class="xref py py-attr docutils literal notranslate"><span class="pre">self</span></code>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.clone"/><a class="reference internal" href="generated/torch.Tensor.clone.html#torch.Tensor.clone" title="torch.Tensor.clone"><code class="xref py py-obj docutils literal notranslate"><span class="pre">clone</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.clone.html#torch.clone" title="torch.clone"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.clone()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.dequantize"/><a class="reference internal" href="generated/torch.Tensor.dequantize.html#torch.Tensor.dequantize" title="torch.Tensor.dequantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">dequantize</span></code></a></p></td>
<td><p>Given a quantized Tensor, dequantize it and return the dequantized float Tensor.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.equal"/><a class="reference internal" href="generated/torch.Tensor.equal.html#torch.Tensor.equal" title="torch.Tensor.equal"><code class="xref py py-obj docutils literal notranslate"><span class="pre">equal</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.equal.html#torch.equal" title="torch.equal"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.equal()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.int_repr"/><a class="reference internal" href="generated/torch.Tensor.int_repr.html#torch.Tensor.int_repr" title="torch.Tensor.int_repr"><code class="xref py py-obj docutils literal notranslate"><span class="pre">int_repr</span></code></a></p></td>
<td><p>Given a quantized Tensor, <code class="docutils literal notranslate"><span class="pre">self.int_repr()</span></code> returns a CPU Tensor with uint8_t as data type that stores the underlying uint8_t values of the given Tensor.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.max"/><a class="reference internal" href="generated/torch.Tensor.max.html#torch.Tensor.max" title="torch.Tensor.max"><code class="xref py py-obj docutils literal notranslate"><span class="pre">max</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.max.html#torch.max" title="torch.max"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.max()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.mean"/><a class="reference internal" href="generated/torch.Tensor.mean.html#torch.Tensor.mean" title="torch.Tensor.mean"><code class="xref py py-obj docutils literal notranslate"><span class="pre">mean</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.mean.html#torch.mean" title="torch.mean"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.mean()</span></code></a></p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.min"/><a class="reference internal" href="generated/torch.Tensor.min.html#torch.Tensor.min" title="torch.Tensor.min"><code class="xref py py-obj docutils literal notranslate"><span class="pre">min</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.min.html#torch.min" title="torch.min"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.min()</span></code></a></p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.q_scale"/><a class="reference internal" href="generated/torch.Tensor.q_scale.html#torch.Tensor.q_scale" title="torch.Tensor.q_scale"><code class="xref py py-obj docutils literal notranslate"><span class="pre">q_scale</span></code></a></p></td>
<td><p>Given a Tensor quantized by linear(affine) quantization, returns the scale of the underlying quantizer().</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.q_zero_point"/><a class="reference internal" href="generated/torch.Tensor.q_zero_point.html#torch.Tensor.q_zero_point" title="torch.Tensor.q_zero_point"><code class="xref py py-obj docutils literal notranslate"><span class="pre">q_zero_point</span></code></a></p></td>
<td><p>Given a Tensor quantized by linear(affine) quantization, returns the zero_point of the underlying quantizer().</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.q_per_channel_scales"/><a class="reference internal" href="generated/torch.Tensor.q_per_channel_scales.html#torch.Tensor.q_per_channel_scales" title="torch.Tensor.q_per_channel_scales"><code class="xref py py-obj docutils literal notranslate"><span class="pre">q_per_channel_scales</span></code></a></p></td>
<td><p>Given a Tensor quantized by linear (affine) per-channel quantization, returns a Tensor of scales of the underlying quantizer.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.q_per_channel_zero_points"/><a class="reference internal" href="generated/torch.Tensor.q_per_channel_zero_points.html#torch.Tensor.q_per_channel_zero_points" title="torch.Tensor.q_per_channel_zero_points"><code class="xref py py-obj docutils literal notranslate"><span class="pre">q_per_channel_zero_points</span></code></a></p></td>
<td><p>Given a Tensor quantized by linear (affine) per-channel quantization, returns a tensor of zero_points of the underlying quantizer.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.q_per_channel_axis"/><a class="reference internal" href="generated/torch.Tensor.q_per_channel_axis.html#torch.Tensor.q_per_channel_axis" title="torch.Tensor.q_per_channel_axis"><code class="xref py py-obj docutils literal notranslate"><span class="pre">q_per_channel_axis</span></code></a></p></td>
<td><p>Given a Tensor quantized by linear (affine) per-channel quantization, returns the index of dimension on which per-channel quantization is applied.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.resize_"/><a class="reference internal" href="generated/torch.Tensor.resize_.html#torch.Tensor.resize_" title="torch.Tensor.resize_"><code class="xref py py-obj docutils literal notranslate"><span class="pre">resize_</span></code></a></p></td>
<td><p>Resizes <code class="xref py py-attr docutils literal notranslate"><span class="pre">self</span></code> tensor to the specified size.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.Tensor.sort"/><a class="reference internal" href="generated/torch.Tensor.sort.html#torch.Tensor.sort" title="torch.Tensor.sort"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sort</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.sort.html#torch.sort" title="torch.sort"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.sort()</span></code></a></p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.Tensor.topk"/><a class="reference internal" href="generated/torch.Tensor.topk.html#torch.Tensor.topk" title="torch.Tensor.topk"><code class="xref py py-obj docutils literal notranslate"><span class="pre">topk</span></code></a></p></td>
<td><p>See <a class="reference internal" href="generated/torch.topk.html#torch.topk" title="torch.topk"><code class="xref py py-func docutils literal notranslate"><span class="pre">torch.topk()</span></code></a></p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-observer">
<h2>torch.ao.quantization.observer<a class="headerlink" href="#torch-ao-quantization-observer" title="Permalink to this heading">¶</a></h2>
<p>This module contains observers which are used to collect statistics about
the values observed during calibration (PTQ) or training (QAT).</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.ObserverBase"/><a class="reference internal" href="generated/torch.ao.quantization.observer.ObserverBase.html#torch.ao.quantization.observer.ObserverBase" title="torch.ao.quantization.observer.ObserverBase"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ObserverBase</span></code></a></p></td>
<td><p>Base observer Module.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.MinMaxObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.MinMaxObserver.html#torch.ao.quantization.observer.MinMaxObserver" title="torch.ao.quantization.observer.MinMaxObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">MinMaxObserver</span></code></a></p></td>
<td><p>Observer module for computing the quantization parameters based on the running min and max values.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.MovingAverageMinMaxObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.MovingAverageMinMaxObserver.html#torch.ao.quantization.observer.MovingAverageMinMaxObserver" title="torch.ao.quantization.observer.MovingAverageMinMaxObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">MovingAverageMinMaxObserver</span></code></a></p></td>
<td><p>Observer module for computing the quantization parameters based on the moving average of the min and max values.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.PerChannelMinMaxObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.PerChannelMinMaxObserver.html#torch.ao.quantization.observer.PerChannelMinMaxObserver" title="torch.ao.quantization.observer.PerChannelMinMaxObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">PerChannelMinMaxObserver</span></code></a></p></td>
<td><p>Observer module for computing the quantization parameters based on the running per channel min and max values.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.MovingAveragePerChannelMinMaxObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.MovingAveragePerChannelMinMaxObserver.html#torch.ao.quantization.observer.MovingAveragePerChannelMinMaxObserver" title="torch.ao.quantization.observer.MovingAveragePerChannelMinMaxObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">MovingAveragePerChannelMinMaxObserver</span></code></a></p></td>
<td><p>Observer module for computing the quantization parameters based on the running per channel min and max values.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.HistogramObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.HistogramObserver.html#torch.ao.quantization.observer.HistogramObserver" title="torch.ao.quantization.observer.HistogramObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">HistogramObserver</span></code></a></p></td>
<td><p>The module records the running histogram of tensor values along with min/max values.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.PlaceholderObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.PlaceholderObserver.html#torch.ao.quantization.observer.PlaceholderObserver" title="torch.ao.quantization.observer.PlaceholderObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">PlaceholderObserver</span></code></a></p></td>
<td><p>Observer that doesn't do anything and just passes its configuration to the quantized module's <code class="docutils literal notranslate"><span class="pre">.from_float()</span></code>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.RecordingObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.RecordingObserver.html#torch.ao.quantization.observer.RecordingObserver" title="torch.ao.quantization.observer.RecordingObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">RecordingObserver</span></code></a></p></td>
<td><p>The module is mainly for debug and records the tensor values during runtime.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.NoopObserver"/><a class="reference internal" href="generated/torch.ao.quantization.observer.NoopObserver.html#torch.ao.quantization.observer.NoopObserver" title="torch.ao.quantization.observer.NoopObserver"><code class="xref py py-obj docutils literal notranslate"><span class="pre">NoopObserver</span></code></a></p></td>
<td><p>Observer that doesn't do anything and just passes its configuration to the quantized module's <code class="docutils literal notranslate"><span class="pre">.from_float()</span></code>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.get_observer_state_dict"/><a class="reference internal" href="generated/torch.ao.quantization.observer.get_observer_state_dict.html#torch.ao.quantization.observer.get_observer_state_dict" title="torch.ao.quantization.observer.get_observer_state_dict"><code class="xref py py-obj docutils literal notranslate"><span class="pre">get_observer_state_dict</span></code></a></p></td>
<td><p>Returns the state dict corresponding to the observer stats.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.load_observer_state_dict"/><a class="reference internal" href="generated/torch.ao.quantization.observer.load_observer_state_dict.html#torch.ao.quantization.observer.load_observer_state_dict" title="torch.ao.quantization.observer.load_observer_state_dict"><code class="xref py py-obj docutils literal notranslate"><span class="pre">load_observer_state_dict</span></code></a></p></td>
<td><p>Given input model and a state_dict containing model observer stats, load the stats back into the model.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.default_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_observer.html#torch.ao.quantization.observer.default_observer" title="torch.ao.quantization.observer.default_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_observer</span></code></a></p></td>
<td><p>Default observer for static quantization, usually used for debugging.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.default_placeholder_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_placeholder_observer.html#torch.ao.quantization.observer.default_placeholder_observer" title="torch.ao.quantization.observer.default_placeholder_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_placeholder_observer</span></code></a></p></td>
<td><p>Default placeholder observer, usually used for quantization to torch.float16.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.default_debug_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_debug_observer.html#torch.ao.quantization.observer.default_debug_observer" title="torch.ao.quantization.observer.default_debug_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_debug_observer</span></code></a></p></td>
<td><p>Default debug-only observer.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.default_weight_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_weight_observer.html#torch.ao.quantization.observer.default_weight_observer" title="torch.ao.quantization.observer.default_weight_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_weight_observer</span></code></a></p></td>
<td><p>Default weight observer.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.default_histogram_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_histogram_observer.html#torch.ao.quantization.observer.default_histogram_observer" title="torch.ao.quantization.observer.default_histogram_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_histogram_observer</span></code></a></p></td>
<td><p>Default histogram observer, usually used for PTQ.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.default_per_channel_weight_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_per_channel_weight_observer.html#torch.ao.quantization.observer.default_per_channel_weight_observer" title="torch.ao.quantization.observer.default_per_channel_weight_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_per_channel_weight_observer</span></code></a></p></td>
<td><p>Default per-channel weight observer, usually used on backends where per-channel weight quantization is supported, such as <cite>fbgemm</cite>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.observer.default_dynamic_quant_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_dynamic_quant_observer.html#torch.ao.quantization.observer.default_dynamic_quant_observer" title="torch.ao.quantization.observer.default_dynamic_quant_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_dynamic_quant_observer</span></code></a></p></td>
<td><p>Default observer for dynamic quantization.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.observer.default_float_qparams_observer"/><a class="reference internal" href="generated/torch.ao.quantization.observer.default_float_qparams_observer.html#torch.ao.quantization.observer.default_float_qparams_observer" title="torch.ao.quantization.observer.default_float_qparams_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_float_qparams_observer</span></code></a></p></td>
<td><p>Default observer for a floating point zero-point.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-fake-quantize">
<h2>torch.ao.quantization.fake_quantize<a class="headerlink" href="#torch-ao-quantization-fake-quantize" title="Permalink to this heading">¶</a></h2>
<p>This module implements modules which are used to perform fake quantization
during QAT.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.FakeQuantizeBase"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.FakeQuantizeBase.html#torch.ao.quantization.fake_quantize.FakeQuantizeBase" title="torch.ao.quantization.fake_quantize.FakeQuantizeBase"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FakeQuantizeBase</span></code></a></p></td>
<td><p>Base fake quantize module Any fake quantize implementation should derive from this class.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.FakeQuantize"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.FakeQuantize.html#torch.ao.quantization.fake_quantize.FakeQuantize" title="torch.ao.quantization.fake_quantize.FakeQuantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FakeQuantize</span></code></a></p></td>
<td><p>Simulate the quantize and dequantize operations in training time. The output of this module is given by::.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.FixedQParamsFakeQuantize"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.FixedQParamsFakeQuantize.html#torch.ao.quantization.fake_quantize.FixedQParamsFakeQuantize" title="torch.ao.quantization.fake_quantize.FixedQParamsFakeQuantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FixedQParamsFakeQuantize</span></code></a></p></td>
<td><p>Simulate quantize and dequantize with fixed quantization parameters in training time.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize.html#torch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize" title="torch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">FusedMovingAvgObsFakeQuantize</span></code></a></p></td>
<td><p>Fused module that is used to observe the input tensor (compute min/max), compute scale/zero_point and fake_quantize the tensor.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.default_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_fake_quant.html#torch.ao.quantization.fake_quantize.default_fake_quant" title="torch.ao.quantization.fake_quantize.default_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_fake_quant</span></code></a></p></td>
<td><p>Default fake_quant for activations.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.default_weight_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_weight_fake_quant.html#torch.ao.quantization.fake_quantize.default_weight_fake_quant" title="torch.ao.quantization.fake_quantize.default_weight_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_weight_fake_quant</span></code></a></p></td>
<td><p>Default fake_quant for weights.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant.html#torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant" title="torch.ao.quantization.fake_quantize.default_per_channel_weight_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_per_channel_weight_fake_quant</span></code></a></p></td>
<td><p>Default fake_quant for per-channel weights.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.default_histogram_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_histogram_fake_quant.html#torch.ao.quantization.fake_quantize.default_histogram_fake_quant" title="torch.ao.quantization.fake_quantize.default_histogram_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_histogram_fake_quant</span></code></a></p></td>
<td><p>Fake_quant for activations using a histogram..</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.default_fused_act_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_fused_act_fake_quant.html#torch.ao.quantization.fake_quantize.default_fused_act_fake_quant" title="torch.ao.quantization.fake_quantize.default_fused_act_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_fused_act_fake_quant</span></code></a></p></td>
<td><p>Fused version of <cite>default_fake_quant</cite>, with improved performance.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.default_fused_wt_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_fused_wt_fake_quant.html#torch.ao.quantization.fake_quantize.default_fused_wt_fake_quant" title="torch.ao.quantization.fake_quantize.default_fused_wt_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_fused_wt_fake_quant</span></code></a></p></td>
<td><p>Fused version of <cite>default_weight_fake_quant</cite>, with improved performance.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.default_fused_per_channel_wt_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.default_fused_per_channel_wt_fake_quant.html#torch.ao.quantization.fake_quantize.default_fused_per_channel_wt_fake_quant" title="torch.ao.quantization.fake_quantize.default_fused_per_channel_wt_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_fused_per_channel_wt_fake_quant</span></code></a></p></td>
<td><p>Fused version of <cite>default_per_channel_weight_fake_quant</cite>, with improved performance.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.disable_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.disable_fake_quant.html#torch.ao.quantization.fake_quantize.disable_fake_quant" title="torch.ao.quantization.fake_quantize.disable_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">disable_fake_quant</span></code></a></p></td>
<td><p>Disable fake quantization for this module, if applicable. Example usage::.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.enable_fake_quant"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.enable_fake_quant.html#torch.ao.quantization.fake_quantize.enable_fake_quant" title="torch.ao.quantization.fake_quantize.enable_fake_quant"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enable_fake_quant</span></code></a></p></td>
<td><p>Enable fake quantization for this module, if applicable. Example usage::.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.fake_quantize.disable_observer"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.disable_observer.html#torch.ao.quantization.fake_quantize.disable_observer" title="torch.ao.quantization.fake_quantize.disable_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">disable_observer</span></code></a></p></td>
<td><p>Disable observation for this module, if applicable. Example usage::.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.fake_quantize.enable_observer"/><a class="reference internal" href="generated/torch.ao.quantization.fake_quantize.enable_observer.html#torch.ao.quantization.fake_quantize.enable_observer" title="torch.ao.quantization.fake_quantize.enable_observer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enable_observer</span></code></a></p></td>
<td><p>Enable observation for this module, if applicable. Example usage::.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="torch-ao-quantization-qconfig">
<h2>torch.ao.quantization.qconfig<a class="headerlink" href="#torch-ao-quantization-qconfig" title="Permalink to this heading">¶</a></h2>
<p>This module defines <cite>QConfig</cite> objects which are used
to configure quantization settings for individual ops.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.QConfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.QConfig.html#torch.ao.quantization.qconfig.QConfig" title="torch.ao.quantization.qconfig.QConfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">QConfig</span></code></a></p></td>
<td><p>Describes how to quantize a layer or a part of the network by providing settings (observer classes) for activations and weights respectively.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.default_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_qconfig.html#torch.ao.quantization.qconfig.default_qconfig" title="torch.ao.quantization.qconfig.default_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_qconfig</span></code></a></p></td>
<td><p>Default qconfig configuration.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.default_debug_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_debug_qconfig.html#torch.ao.quantization.qconfig.default_debug_qconfig" title="torch.ao.quantization.qconfig.default_debug_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_debug_qconfig</span></code></a></p></td>
<td><p>Default qconfig configuration for debugging.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.default_per_channel_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_per_channel_qconfig.html#torch.ao.quantization.qconfig.default_per_channel_qconfig" title="torch.ao.quantization.qconfig.default_per_channel_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_per_channel_qconfig</span></code></a></p></td>
<td><p>Default qconfig configuration for per channel weight quantization.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.default_dynamic_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_dynamic_qconfig.html#torch.ao.quantization.qconfig.default_dynamic_qconfig" title="torch.ao.quantization.qconfig.default_dynamic_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_dynamic_qconfig</span></code></a></p></td>
<td><p>Default dynamic qconfig.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.float16_dynamic_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.float16_dynamic_qconfig.html#torch.ao.quantization.qconfig.float16_dynamic_qconfig" title="torch.ao.quantization.qconfig.float16_dynamic_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">float16_dynamic_qconfig</span></code></a></p></td>
<td><p>Dynamic qconfig with weights quantized to <cite>torch.float16</cite>.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.float16_static_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.float16_static_qconfig.html#torch.ao.quantization.qconfig.float16_static_qconfig" title="torch.ao.quantization.qconfig.float16_static_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">float16_static_qconfig</span></code></a></p></td>
<td><p>Dynamic qconfig with both activations and weights quantized to <cite>torch.float16</cite>.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.per_channel_dynamic_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.per_channel_dynamic_qconfig.html#torch.ao.quantization.qconfig.per_channel_dynamic_qconfig" title="torch.ao.quantization.qconfig.per_channel_dynamic_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">per_channel_dynamic_qconfig</span></code></a></p></td>
<td><p>Dynamic qconfig with weights quantized per channel.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.float_qparams_weight_only_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.float_qparams_weight_only_qconfig.html#torch.ao.quantization.qconfig.float_qparams_weight_only_qconfig" title="torch.ao.quantization.qconfig.float_qparams_weight_only_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">float_qparams_weight_only_qconfig</span></code></a></p></td>
<td><p>Dynamic qconfig with weights quantized with a floating point zero_point.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.default_qat_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_qat_qconfig.html#torch.ao.quantization.qconfig.default_qat_qconfig" title="torch.ao.quantization.qconfig.default_qat_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_qat_qconfig</span></code></a></p></td>
<td><p>Default qconfig for QAT.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.default_weight_only_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_weight_only_qconfig.html#torch.ao.quantization.qconfig.default_weight_only_qconfig" title="torch.ao.quantization.qconfig.default_weight_only_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_weight_only_qconfig</span></code></a></p></td>
<td><p>Default qconfig for quantizing weights only.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.quantization.qconfig.default_activation_only_qconfig"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_activation_only_qconfig.html#torch.ao.quantization.qconfig.default_activation_only_qconfig" title="torch.ao.quantization.qconfig.default_activation_only_qconfig"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_activation_only_qconfig</span></code></a></p></td>
<td><p>Default qconfig for quantizing activations only.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.quantization.qconfig.default_qat_qconfig_v2"/><a class="reference internal" href="generated/torch.ao.quantization.qconfig.default_qat_qconfig_v2.html#torch.ao.quantization.qconfig.default_qat_qconfig_v2" title="torch.ao.quantization.qconfig.default_qat_qconfig_v2"><code class="xref py py-obj docutils literal notranslate"><span class="pre">default_qat_qconfig_v2</span></code></a></p></td>
<td><p>Fused version of <cite>default_qat_config</cite>, has performance benefits.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="module-torch.ao.nn.intrinsic">
<span id="torch-ao-nn-intrinsic"></span><h2>torch.ao.nn.intrinsic<a class="headerlink" href="#module-torch.ao.nn.intrinsic" title="Permalink to this heading">¶</a></h2>
<span class="target" id="module-torch.ao.nn.intrinsic.modules"></span><p>This module implements the combined (fused) modules conv + relu which can
then be quantized.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.ConvReLU1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvReLU1d.html#torch.ao.nn.intrinsic.ConvReLU1d" title="torch.ao.nn.intrinsic.ConvReLU1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU1d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv1d and ReLU modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.ConvReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvReLU2d.html#torch.ao.nn.intrinsic.ConvReLU2d" title="torch.ao.nn.intrinsic.ConvReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU2d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv2d and ReLU modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.ConvReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvReLU3d.html#torch.ao.nn.intrinsic.ConvReLU3d" title="torch.ao.nn.intrinsic.ConvReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU3d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv3d and ReLU modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.LinearReLU"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.LinearReLU.html#torch.ao.nn.intrinsic.LinearReLU" title="torch.ao.nn.intrinsic.LinearReLU"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LinearReLU</span></code></a></p></td>
<td><p>This is a sequential container which calls the Linear and ReLU modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.ConvBn1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBn1d.html#torch.ao.nn.intrinsic.ConvBn1d" title="torch.ao.nn.intrinsic.ConvBn1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn1d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 1d and Batch Norm 1d modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.ConvBn2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBn2d.html#torch.ao.nn.intrinsic.ConvBn2d" title="torch.ao.nn.intrinsic.ConvBn2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn2d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 2d and Batch Norm 2d modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.ConvBn3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBn3d.html#torch.ao.nn.intrinsic.ConvBn3d" title="torch.ao.nn.intrinsic.ConvBn3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn3d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 3d and Batch Norm 3d modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.ConvBnReLU1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBnReLU1d.html#torch.ao.nn.intrinsic.ConvBnReLU1d" title="torch.ao.nn.intrinsic.ConvBnReLU1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU1d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 1d, Batch Norm 1d, and ReLU modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.ConvBnReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBnReLU2d.html#torch.ao.nn.intrinsic.ConvBnReLU2d" title="torch.ao.nn.intrinsic.ConvBnReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU2d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 2d, Batch Norm 2d, and ReLU modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.ConvBnReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.ConvBnReLU3d.html#torch.ao.nn.intrinsic.ConvBnReLU3d" title="torch.ao.nn.intrinsic.ConvBnReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU3d</span></code></a></p></td>
<td><p>This is a sequential container which calls the Conv 3d, Batch Norm 3d, and ReLU modules.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.BNReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.BNReLU2d.html#torch.ao.nn.intrinsic.BNReLU2d" title="torch.ao.nn.intrinsic.BNReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BNReLU2d</span></code></a></p></td>
<td><p>This is a sequential container which calls the BatchNorm 2d and ReLU modules.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.BNReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.BNReLU3d.html#torch.ao.nn.intrinsic.BNReLU3d" title="torch.ao.nn.intrinsic.BNReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BNReLU3d</span></code></a></p></td>
<td><p>This is a sequential container which calls the BatchNorm 3d and ReLU modules.</p></td>
</tr>
</tbody>
</table>
</section>
<section id="module-torch.ao.nn.intrinsic.qat">
<span id="torch-ao-nn-intrinsic-qat"></span><h2>torch.ao.nn.intrinsic.qat<a class="headerlink" href="#module-torch.ao.nn.intrinsic.qat" title="Permalink to this heading">¶</a></h2>
<span class="target" id="module-torch.ao.nn.intrinsic.qat.modules"></span><p>This module implements the versions of those fused operations needed for
quantization aware training.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.LinearReLU"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.LinearReLU.html#torch.ao.nn.intrinsic.qat.LinearReLU" title="torch.ao.nn.intrinsic.qat.LinearReLU"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LinearReLU</span></code></a></p></td>
<td><p>A LinearReLU module fused from Linear and ReLU modules, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBn1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBn1d.html#torch.ao.nn.intrinsic.qat.ConvBn1d" title="torch.ao.nn.intrinsic.qat.ConvBn1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn1d</span></code></a></p></td>
<td><p>A ConvBn1d module is a module fused from Conv1d and BatchNorm1d, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBnReLU1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBnReLU1d.html#torch.ao.nn.intrinsic.qat.ConvBnReLU1d" title="torch.ao.nn.intrinsic.qat.ConvBnReLU1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU1d</span></code></a></p></td>
<td><p>A ConvBnReLU1d module is a module fused from Conv1d, BatchNorm1d and ReLU, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBn2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBn2d.html#torch.ao.nn.intrinsic.qat.ConvBn2d" title="torch.ao.nn.intrinsic.qat.ConvBn2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn2d</span></code></a></p></td>
<td><p>A ConvBn2d module is a module fused from Conv2d and BatchNorm2d, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBnReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBnReLU2d.html#torch.ao.nn.intrinsic.qat.ConvBnReLU2d" title="torch.ao.nn.intrinsic.qat.ConvBnReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU2d</span></code></a></p></td>
<td><p>A ConvBnReLU2d module is a module fused from Conv2d, BatchNorm2d and ReLU, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvReLU2d.html#torch.ao.nn.intrinsic.qat.ConvReLU2d" title="torch.ao.nn.intrinsic.qat.ConvReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU2d</span></code></a></p></td>
<td><p>A ConvReLU2d module is a fused module of Conv2d and ReLU, attached with FakeQuantize modules for weight for quantization aware training.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBn3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBn3d.html#torch.ao.nn.intrinsic.qat.ConvBn3d" title="torch.ao.nn.intrinsic.qat.ConvBn3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBn3d</span></code></a></p></td>
<td><p>A ConvBn3d module is a module fused from Conv3d and BatchNorm3d, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvBnReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvBnReLU3d.html#torch.ao.nn.intrinsic.qat.ConvBnReLU3d" title="torch.ao.nn.intrinsic.qat.ConvBnReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvBnReLU3d</span></code></a></p></td>
<td><p>A ConvBnReLU3d module is a module fused from Conv3d, BatchNorm3d and ReLU, attached with FakeQuantize modules for weight, used in quantization aware training.</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.ConvReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.ConvReLU3d.html#torch.ao.nn.intrinsic.qat.ConvReLU3d" title="torch.ao.nn.intrinsic.qat.ConvReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU3d</span></code></a></p></td>
<td><p>A ConvReLU3d module is a fused module of Conv3d and ReLU, attached with FakeQuantize modules for weight for quantization aware training.</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.qat.update_bn_stats"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.update_bn_stats.html#torch.ao.nn.intrinsic.qat.update_bn_stats" title="torch.ao.nn.intrinsic.qat.update_bn_stats"><code class="xref py py-obj docutils literal notranslate"><span class="pre">update_bn_stats</span></code></a></p></td>
<td><p></p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.qat.freeze_bn_stats"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.qat.freeze_bn_stats.html#torch.ao.nn.intrinsic.qat.freeze_bn_stats" title="torch.ao.nn.intrinsic.qat.freeze_bn_stats"><code class="xref py py-obj docutils literal notranslate"><span class="pre">freeze_bn_stats</span></code></a></p></td>
<td><p></p></td>
</tr>
</tbody>
</table>
</section>
<section id="module-torch.ao.nn.intrinsic.quantized">
<span id="torch-ao-nn-intrinsic-quantized"></span><h2>torch.ao.nn.intrinsic.quantized<a class="headerlink" href="#module-torch.ao.nn.intrinsic.quantized" title="Permalink to this heading">¶</a></h2>
<span class="target" id="module-torch.ao.nn.intrinsic.quantized.modules"></span><p>This module implements the quantized implementations of fused operations
like conv + relu. No BatchNorm variants as it’s usually folded into convolution
for inference.</p>
<table class="autosummary longtable docutils align-default">
<tbody>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.quantized.BNReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.BNReLU2d.html#torch.ao.nn.intrinsic.quantized.BNReLU2d" title="torch.ao.nn.intrinsic.quantized.BNReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BNReLU2d</span></code></a></p></td>
<td><p>A BNReLU2d module is a fused module of BatchNorm2d and ReLU</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.quantized.BNReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.BNReLU3d.html#torch.ao.nn.intrinsic.quantized.BNReLU3d" title="torch.ao.nn.intrinsic.quantized.BNReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">BNReLU3d</span></code></a></p></td>
<td><p>A BNReLU3d module is a fused module of BatchNorm3d and ReLU</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.quantized.ConvReLU1d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.ConvReLU1d.html#torch.ao.nn.intrinsic.quantized.ConvReLU1d" title="torch.ao.nn.intrinsic.quantized.ConvReLU1d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU1d</span></code></a></p></td>
<td><p>A ConvReLU1d module is a fused module of Conv1d and ReLU</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.quantized.ConvReLU2d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.ConvReLU2d.html#torch.ao.nn.intrinsic.quantized.ConvReLU2d" title="torch.ao.nn.intrinsic.quantized.ConvReLU2d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU2d</span></code></a></p></td>
<td><p>A ConvReLU2d module is a fused module of Conv2d and ReLU</p></td>
</tr>
<tr class="row-odd"><td><p><p id="torch.ao.nn.intrinsic.quantized.ConvReLU3d"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.ConvReLU3d.html#torch.ao.nn.intrinsic.quantized.ConvReLU3d" title="torch.ao.nn.intrinsic.quantized.ConvReLU3d"><code class="xref py py-obj docutils literal notranslate"><span class="pre">ConvReLU3d</span></code></a></p></td>
<td><p>A ConvReLU3d module is a fused module of Conv3d and ReLU</p></td>
</tr>
<tr class="row-even"><td><p><p id="torch.ao.nn.intrinsic.quantized.LinearReLU"/><a class="reference internal" href="generated/torch.ao.nn.intrinsic.quantized.LinearReLU.html#torch.ao.nn.intrinsic.quantized.LinearReLU" title="torch.ao.nn.intrinsic.quantized.LinearReLU"><code class="xref py py-obj docutils literal notranslate"><span class="pre">LinearReLU</span></code></a></p></td>
<td><p>A LinearReLU module fused from Linear and ReLU modules</p></td>
</tr>
</tbody>
</table>
</section>