-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathoptimizer.html
1335 lines (1052 loc) · 114 KB
/
optimizer.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 name="robots" content="noindex">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>torch.optim.optimizer — PyTorch 2.0 documentation</title>
<link rel="canonical" href="https://pytorch.org/docs/stable/_modules/torch/optim/optimizer.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" />
<!-- 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>
<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"><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="../../index.html">Module code</a> ></li>
<li><a href="../../torch.html">torch</a> ></li>
<li>torch.optim.optimizer</li>
<li class="pytorch-breadcrumbs-aside">
</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">
<h1>Source code for torch.optim.optimizer</h1><div class="highlight"><pre>
<span></span><span class="kn">from</span> <span class="nn">collections</span> <span class="kn">import</span> <span class="n">OrderedDict</span><span class="p">,</span> <span class="n">defaultdict</span><span class="p">,</span> <span class="n">abc</span> <span class="k">as</span> <span class="n">container_abcs</span>
<span class="kn">import</span> <span class="nn">torch</span>
<span class="kn">from</span> <span class="nn">copy</span> <span class="kn">import</span> <span class="n">deepcopy</span>
<span class="kn">from</span> <span class="nn">itertools</span> <span class="kn">import</span> <span class="n">chain</span>
<span class="kn">import</span> <span class="nn">warnings</span>
<span class="kn">import</span> <span class="nn">functools</span>
<span class="kn">import</span> <span class="nn">math</span>
<span class="kn">from</span> <span class="nn">typing</span> <span class="kn">import</span> <span class="n">Callable</span><span class="p">,</span> <span class="n">Dict</span><span class="p">,</span> <span class="n">List</span><span class="p">,</span> <span class="n">Tuple</span>
<span class="kn">import</span> <span class="nn">torch.utils.hooks</span> <span class="k">as</span> <span class="nn">hooks</span>
<span class="kn">from</span> <span class="nn">torch.utils.hooks</span> <span class="kn">import</span> <span class="n">RemovableHandle</span>
<span class="kn">from</span> <span class="nn">torch._utils</span> <span class="kn">import</span> <span class="n">is_compiling</span>
<span class="n">__all__</span> <span class="o">=</span> <span class="p">[</span><span class="s1">'Optimizer'</span><span class="p">,</span> <span class="s1">'register_optimizer_step_pre_hook'</span><span class="p">,</span> <span class="s1">'register_optimizer_step_post_hook'</span><span class="p">]</span>
<span class="n">_global_optimizer_pre_hooks</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">int</span><span class="p">,</span> <span class="n">Callable</span><span class="p">]</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="n">_global_optimizer_post_hooks</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">int</span><span class="p">,</span> <span class="n">Callable</span><span class="p">]</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="n">_foreach_supported_types</span> <span class="o">=</span> <span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">nn</span><span class="o">.</span><span class="n">parameter</span><span class="o">.</span><span class="n">Parameter</span><span class="p">]</span>
<span class="k">class</span> <span class="nc">_RequiredParameter</span><span class="p">:</span>
<span class="sd">"""Singleton class representing a required parameter for an Optimizer."""</span>
<span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="s2">"<required parameter>"</span>
<span class="n">required</span> <span class="o">=</span> <span class="n">_RequiredParameter</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">_use_grad_for_differentiable</span><span class="p">(</span><span class="n">func</span><span class="p">):</span>
<span class="k">def</span> <span class="nf">_use_grad</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="n">prev_grad</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">is_grad_enabled</span><span class="p">()</span>
<span class="k">try</span><span class="p">:</span>
<span class="n">torch</span><span class="o">.</span><span class="n">set_grad_enabled</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="p">[</span><span class="s1">'differentiable'</span><span class="p">])</span>
<span class="n">ret</span> <span class="o">=</span> <span class="n">func</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
<span class="k">finally</span><span class="p">:</span>
<span class="n">torch</span><span class="o">.</span><span class="n">set_grad_enabled</span><span class="p">(</span><span class="n">prev_grad</span><span class="p">)</span>
<span class="k">return</span> <span class="n">ret</span>
<span class="k">return</span> <span class="n">_use_grad</span>
<span class="k">def</span> <span class="nf">_get_value</span><span class="p">(</span><span class="n">x</span><span class="p">):</span>
<span class="c1"># item is significantly faster than a cpu tensor in eager mode</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">is_scripting</span><span class="p">()</span> <span class="ow">and</span> <span class="n">is_compiling</span><span class="p">():</span>
<span class="k">return</span> <span class="n">x</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">x</span><span class="o">.</span><span class="n">item</span><span class="p">()</span>
<span class="k">def</span> <span class="nf">_stack_if_compiling</span><span class="p">(</span><span class="n">x</span><span class="p">):</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">is_scripting</span><span class="p">()</span> <span class="ow">and</span> <span class="n">is_compiling</span><span class="p">():</span>
<span class="k">return</span> <span class="n">torch</span><span class="o">.</span><span class="n">stack</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">x</span>
<span class="k">def</span> <span class="nf">_dispatch_sqrt</span><span class="p">(</span><span class="n">x</span><span class="p">:</span> <span class="nb">float</span><span class="p">):</span> <span class="c1"># float annotation is needed because of torchscript type inference</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">is_scripting</span><span class="p">()</span> <span class="ow">and</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">x</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
<span class="k">return</span> <span class="n">x</span><span class="o">.</span><span class="n">sqrt</span><span class="p">()</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">math</span><span class="o">.</span><span class="n">sqrt</span><span class="p">(</span><span class="n">x</span><span class="p">)</span>
<span class="c1"># For any optimizer with a faster implementation, we attempt to default to the</span>
<span class="c1"># fastest + stablest whenever possible. For foreach, the requirements are to have</span>
<span class="c1"># native params all on CUDA. For fused, there's currently the additional requirement</span>
<span class="c1"># that the tensors' dtypes must be floating point. Neither alternative supports</span>
<span class="c1"># torch.jit.script nor differentiable, so we fall back to the single tensor</span>
<span class="c1"># implementation in those cases.</span>
<span class="k">def</span> <span class="nf">_default_to_fused_or_foreach</span><span class="p">(</span><span class="n">params</span><span class="p">:</span> <span class="n">List</span><span class="p">[</span><span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">],</span>
<span class="n">differentiable</span><span class="p">:</span> <span class="nb">bool</span><span class="p">,</span>
<span class="n">use_fused</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">False</span><span class="p">)</span> <span class="o">-></span> <span class="n">Tuple</span><span class="p">[</span><span class="nb">bool</span><span class="p">,</span> <span class="nb">bool</span><span class="p">]:</span>
<span class="k">if</span> <span class="n">torch</span><span class="o">.</span><span class="n">jit</span><span class="o">.</span><span class="n">is_scripting</span><span class="p">()</span> <span class="ow">or</span> <span class="n">differentiable</span><span class="p">:</span>
<span class="k">return</span> <span class="kc">False</span><span class="p">,</span> <span class="kc">False</span>
<span class="n">fused</span> <span class="o">=</span> <span class="n">use_fused</span> <span class="ow">and</span> <span class="nb">all</span><span class="p">(</span>
<span class="n">p</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="p">(</span><span class="nb">type</span><span class="p">(</span><span class="n">p</span><span class="p">)</span> <span class="ow">in</span> <span class="n">_foreach_supported_types</span> <span class="ow">and</span> <span class="n">p</span><span class="o">.</span><span class="n">is_cuda</span> <span class="ow">and</span> <span class="n">torch</span><span class="o">.</span><span class="n">is_floating_point</span><span class="p">(</span><span class="n">p</span><span class="p">))</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">params</span>
<span class="p">)</span>
<span class="n">foreach</span> <span class="o">=</span> <span class="ow">not</span> <span class="n">fused</span> <span class="ow">and</span> <span class="nb">all</span><span class="p">(</span>
<span class="n">p</span> <span class="ow">is</span> <span class="kc">None</span> <span class="ow">or</span> <span class="p">(</span><span class="nb">type</span><span class="p">(</span><span class="n">p</span><span class="p">)</span> <span class="ow">in</span> <span class="n">_foreach_supported_types</span> <span class="ow">and</span> <span class="n">p</span><span class="o">.</span><span class="n">is_cuda</span><span class="p">)</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">params</span>
<span class="p">)</span>
<span class="k">return</span> <span class="n">fused</span><span class="p">,</span> <span class="n">foreach</span>
<span class="c1"># Common doc strings among optimizers</span>
<span class="n">_foreach_doc</span> <span class="o">=</span> <span class="sa">r</span><span class="s2">"""foreach (bool, optional): whether foreach implementation of optimizer</span>
<span class="s2"> is used. If unspecified by the user (so foreach is None), we will try to use</span>
<span class="s2"> foreach over the for-loop implementation on CUDA, since it is usually</span>
<span class="s2"> significantly more performant. (default: None)"""</span>
<span class="n">_fused_doc</span> <span class="o">=</span> <span class="sa">r</span><span class="s2">"""fused (bool, optional): whether the fused implementation (CUDA only) is used.</span>
<span class="s2"> Currently, `torch.float64`, `torch.float32`, `torch.float16`, and `torch.bfloat16`</span>
<span class="s2"> are supported. (default: None)</span>
<span class="s2"> .. note:: The foreach and fused implementations are typically faster than the for-loop,</span>
<span class="s2"> single-tensor implementation. Thus, if the user has not specified BOTH flags</span>
<span class="s2"> (i.e., when foreach = fused = None), we will attempt defaulting to the foreach</span>
<span class="s2"> implementation when the tensors are all on CUDA. For example, if the user specifies</span>
<span class="s2"> True for fused but nothing for foreach, we will run the fused implementation. If</span>
<span class="s2"> the user specifies False for foreach but nothing for fused (or False for fused but</span>
<span class="s2"> nothing for foreach), we will run the for-loop implementation. If the user specifies</span>
<span class="s2"> True for both foreach and fused, we will prioritize fused over foreach, as it is</span>
<span class="s2"> typically faster. We attempt to use the fastest, so the hierarchy goes fused -></span>
<span class="s2"> foreach -> for-loop. HOWEVER, since the fused implementation is relatively new,</span>
<span class="s2"> we want to give it sufficient bake-in time, so we default to foreach and NOT</span>
<span class="s2"> fused when the user has not specified either flag."""</span>
<span class="n">_capturable_doc</span> <span class="o">=</span> <span class="sa">r</span><span class="s2">"""capturable (bool, optional): whether this instance is safe to</span>
<span class="s2"> capture in a CUDA graph. Passing True can impair ungraphed performance,</span>
<span class="s2"> so if you don't intend to graph capture this instance, leave it False</span>
<span class="s2"> (default: False)"""</span>
<span class="n">_differentiable_doc</span> <span class="o">=</span> <span class="sa">r</span><span class="s2">"""differentiable (bool, optional): whether autograd should</span>
<span class="s2"> occur through the optimizer step in training. Otherwise, the step()</span>
<span class="s2"> function runs in a torch.no_grad() context. Setting to True can impair</span>
<span class="s2"> performance, so leave it False if you don't intend to run autograd</span>
<span class="s2"> through this instance (default: False)"""</span>
<span class="n">_maximize_doc</span> <span class="o">=</span> <span class="sa">r</span><span class="s2">"""maximize (bool, optional): maximize the params based on the</span>
<span class="s2"> objective, instead of minimizing (default: False)"""</span>
<span class="k">def</span> <span class="nf">register_optimizer_step_pre_hook</span><span class="p">(</span><span class="n">hook</span><span class="p">:</span> <span class="n">Callable</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="kc">None</span><span class="p">])</span> <span class="o">-></span> <span class="n">RemovableHandle</span><span class="p">:</span>
<span class="sa">r</span><span class="sd">"""Register a pre hook common to all optimizers. The hook should have the following</span>
<span class="sd"> signature::</span>
<span class="sd"> hook(optimizer, args, kwargs) -> None or modified args and kwargs</span>
<span class="sd"> Args:</span>
<span class="sd"> hook (Callable): A user defined hook which is registered on all optimizers.</span>
<span class="sd"> Returns:</span>
<span class="sd"> :class:`torch.utils.hooks.RemoveableHandle`:</span>
<span class="sd"> a handle that can be used to remove the added hook by calling</span>
<span class="sd"> ``handle.remove()``</span>
<span class="sd"> """</span>
<span class="n">handle</span> <span class="o">=</span> <span class="n">hooks</span><span class="o">.</span><span class="n">RemovableHandle</span><span class="p">(</span><span class="n">_global_optimizer_pre_hooks</span><span class="p">)</span>
<span class="n">_global_optimizer_pre_hooks</span><span class="p">[</span><span class="n">handle</span><span class="o">.</span><span class="n">id</span><span class="p">]</span> <span class="o">=</span> <span class="n">hook</span>
<span class="k">return</span> <span class="n">handle</span>
<span class="k">def</span> <span class="nf">register_optimizer_step_post_hook</span><span class="p">(</span><span class="n">hook</span><span class="p">:</span> <span class="n">Callable</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="kc">None</span><span class="p">])</span> <span class="o">-></span> <span class="n">RemovableHandle</span><span class="p">:</span>
<span class="sa">r</span><span class="sd">"""Register a post hook common to all optimizers. The hook should have the following</span>
<span class="sd"> signature::</span>
<span class="sd"> hook(optimizer, args, kwargs) -> None</span>
<span class="sd"> Args:</span>
<span class="sd"> hook (Callable): A user defined hook which is registered on all optimizers.</span>
<span class="sd"> Returns:</span>
<span class="sd"> :class:`torch.utils.hooks.RemoveableHandle`:</span>
<span class="sd"> a handle that can be used to remove the added hook by calling</span>
<span class="sd"> ``handle.remove()``</span>
<span class="sd"> """</span>
<span class="n">handle</span> <span class="o">=</span> <span class="n">hooks</span><span class="o">.</span><span class="n">RemovableHandle</span><span class="p">(</span><span class="n">_global_optimizer_post_hooks</span><span class="p">)</span>
<span class="n">_global_optimizer_post_hooks</span><span class="p">[</span><span class="n">handle</span><span class="o">.</span><span class="n">id</span><span class="p">]</span> <span class="o">=</span> <span class="n">hook</span>
<span class="k">return</span> <span class="n">handle</span>
<div class="viewcode-block" id="Optimizer"><a class="viewcode-back" href="../../../optim.html#torch.optim.Optimizer">[docs]</a><span class="k">class</span> <span class="nc">Optimizer</span><span class="p">:</span>
<span class="sa">r</span><span class="sd">"""Base class for all optimizers.</span>
<span class="sd"> .. warning::</span>
<span class="sd"> Parameters need to be specified as collections that have a deterministic</span>
<span class="sd"> ordering that is consistent between runs. Examples of objects that don't</span>
<span class="sd"> satisfy those properties are sets and iterators over values of dictionaries.</span>
<span class="sd"> Args:</span>
<span class="sd"> params (iterable): an iterable of :class:`torch.Tensor` s or</span>
<span class="sd"> :class:`dict` s. Specifies what Tensors should be optimized.</span>
<span class="sd"> defaults: (dict): a dict containing default values of optimization</span>
<span class="sd"> options (used when a parameter group doesn't specify them).</span>
<span class="sd"> """</span>
<span class="k">def</span> <span class="fm">__init__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">params</span><span class="p">,</span> <span class="n">defaults</span><span class="p">):</span>
<span class="n">torch</span><span class="o">.</span><span class="n">_C</span><span class="o">.</span><span class="n">_log_api_usage_once</span><span class="p">(</span><span class="s2">"python.optimizer"</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">defaults</span> <span class="o">=</span> <span class="n">defaults</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_pre_hooks</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">int</span><span class="p">,</span> <span class="n">Callable</span><span class="p">]</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_post_hooks</span><span class="p">:</span> <span class="n">Dict</span><span class="p">[</span><span class="nb">int</span><span class="p">,</span> <span class="n">Callable</span><span class="p">]</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_patch_step_function</span><span class="p">()</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">params</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">TypeError</span><span class="p">(</span><span class="s2">"params argument given to the optimizer should be "</span>
<span class="s2">"an iterable of Tensors or dicts, but got "</span> <span class="o">+</span>
<span class="n">torch</span><span class="o">.</span><span class="n">typename</span><span class="p">(</span><span class="n">params</span><span class="p">))</span>
<span class="bp">self</span><span class="o">.</span><span class="n">state</span> <span class="o">=</span> <span class="n">defaultdict</span><span class="p">(</span><span class="nb">dict</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span> <span class="o">=</span> <span class="p">[]</span>
<span class="n">param_groups</span> <span class="o">=</span> <span class="nb">list</span><span class="p">(</span><span class="n">params</span><span class="p">)</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">param_groups</span><span class="p">)</span> <span class="o">==</span> <span class="mi">0</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"optimizer got an empty parameter list"</span><span class="p">)</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">param_groups</span><span class="p">[</span><span class="mi">0</span><span class="p">],</span> <span class="nb">dict</span><span class="p">):</span>
<span class="n">param_groups</span> <span class="o">=</span> <span class="p">[{</span><span class="s1">'params'</span><span class="p">:</span> <span class="n">param_groups</span><span class="p">}]</span>
<span class="k">for</span> <span class="n">param_group</span> <span class="ow">in</span> <span class="n">param_groups</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">add_param_group</span><span class="p">(</span><span class="n">param_group</span><span class="p">)</span>
<span class="c1"># Allows _cuda_graph_capture_health_check to rig a poor man's TORCH_WARN_ONCE in python,</span>
<span class="c1"># which I don't think exists</span>
<span class="c1"># https://github.com/pytorch/pytorch/issues/72948</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_warned_capturable_if_run_uncaptured</span> <span class="o">=</span> <span class="kc">True</span>
<span class="k">def</span> <span class="nf">__getstate__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">return</span> <span class="p">{</span>
<span class="s1">'defaults'</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="p">,</span>
<span class="s1">'state'</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">state</span><span class="p">,</span>
<span class="s1">'param_groups'</span><span class="p">:</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">,</span>
<span class="p">}</span>
<span class="k">def</span> <span class="nf">__setstate__</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">state</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="vm">__dict__</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="n">state</span><span class="p">)</span>
<span class="k">if</span> <span class="s1">'_optimizer_step_pre_hooks'</span> <span class="ow">not</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="vm">__dict__</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_pre_hooks</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="k">if</span> <span class="s1">'_optimizer_step_post_hooks'</span> <span class="ow">not</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="vm">__dict__</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_post_hooks</span> <span class="o">=</span> <span class="n">OrderedDict</span><span class="p">()</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_patch_step_function</span><span class="p">()</span> <span class="c1"># To support multiprocessing pickle/unpickle</span>
<span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="s1">'differentiable'</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span>
<span class="k">def</span> <span class="fm">__repr__</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="n">format_string</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span> <span class="o">+</span> <span class="s1">' ('</span>
<span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">group</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">):</span>
<span class="n">format_string</span> <span class="o">+=</span> <span class="s1">'</span><span class="se">\n</span><span class="s1">'</span>
<span class="n">format_string</span> <span class="o">+=</span> <span class="s1">'Parameter Group </span><span class="si">{0}</span><span class="se">\n</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">i</span><span class="p">)</span>
<span class="k">for</span> <span class="n">key</span> <span class="ow">in</span> <span class="nb">sorted</span><span class="p">(</span><span class="n">group</span><span class="o">.</span><span class="n">keys</span><span class="p">()):</span>
<span class="k">if</span> <span class="n">key</span> <span class="o">!=</span> <span class="s1">'params'</span><span class="p">:</span>
<span class="n">format_string</span> <span class="o">+=</span> <span class="s1">' </span><span class="si">{0}</span><span class="s1">: </span><span class="si">{1}</span><span class="se">\n</span><span class="s1">'</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="n">key</span><span class="p">,</span> <span class="n">group</span><span class="p">[</span><span class="n">key</span><span class="p">])</span>
<span class="n">format_string</span> <span class="o">+=</span> <span class="s1">')'</span>
<span class="k">return</span> <span class="n">format_string</span>
<span class="c1"># Currently needed by Adam and AdamW</span>
<span class="k">def</span> <span class="nf">_cuda_graph_capture_health_check</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="k">if</span> <span class="n">torch</span><span class="o">.</span><span class="n">has_cuda</span> <span class="ow">and</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">is_available</span><span class="p">():</span>
<span class="n">capturing</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">cuda</span><span class="o">.</span><span class="n">is_current_stream_capturing</span><span class="p">()</span>
<span class="k">if</span> <span class="n">capturing</span> <span class="ow">and</span> <span class="ow">not</span> <span class="nb">all</span><span class="p">(</span><span class="n">group</span><span class="p">[</span><span class="s1">'capturable'</span><span class="p">]</span> <span class="k">for</span> <span class="n">group</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="s2">"Attempting CUDA graph capture of step() for an instance of "</span> <span class="o">+</span>
<span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span> <span class="o">+</span>
<span class="s2">" but param_groups' capturable is False."</span><span class="p">)</span>
<span class="k">if</span> <span class="p">(</span>
<span class="p">(</span><span class="ow">not</span> <span class="nb">getattr</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="s2">"_warned_capturable_if_run_uncaptured"</span><span class="p">,</span> <span class="kc">False</span><span class="p">))</span>
<span class="ow">and</span> <span class="nb">all</span><span class="p">(</span><span class="n">group</span><span class="p">[</span><span class="s1">'capturable'</span><span class="p">]</span> <span class="k">for</span> <span class="n">group</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">)</span>
<span class="ow">and</span> <span class="p">(</span><span class="ow">not</span> <span class="n">capturing</span><span class="p">)</span>
<span class="p">):</span>
<span class="n">warnings</span><span class="o">.</span><span class="n">warn</span><span class="p">(</span>
<span class="s2">"This instance was constructed with capturable=True or some of all the param_groups came with capturable=True, "</span>
<span class="s2">"but step() is running without CUDA graph capture. If you never intend to graph-capture this "</span>
<span class="s2">"instance, capturable=True can impair performance, and you should set capturable=False."</span>
<span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_warned_capturable_if_run_uncaptured</span> <span class="o">=</span> <span class="kc">True</span>
<span class="k">def</span> <span class="nf">_optimizer_step_code</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="sd">"""Entry point for `torch.profile.profiler`.</span>
<span class="sd"> When python tracing is enabled the profiler will hook into this</span>
<span class="sd"> function at the CPython level to inspect the optimizer's parameters and</span>
<span class="sd"> param groups. It is called it after `step()` since many optimizers</span>
<span class="sd"> lazily initialize state.</span>
<span class="sd"> This is a workaround due to lack of a proper step hook on the optimizer,</span>
<span class="sd"> and will be removed if it exists.</span>
<span class="sd"> """</span>
<span class="k">pass</span>
<span class="nd">@staticmethod</span>
<span class="k">def</span> <span class="nf">profile_hook_step</span><span class="p">(</span><span class="n">func</span><span class="p">):</span>
<span class="nd">@functools</span><span class="o">.</span><span class="n">wraps</span><span class="p">(</span><span class="n">func</span><span class="p">)</span>
<span class="k">def</span> <span class="nf">wrapper</span><span class="p">(</span><span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">):</span>
<span class="bp">self</span><span class="p">,</span> <span class="o">*</span><span class="n">_</span> <span class="o">=</span> <span class="n">args</span>
<span class="n">profile_name</span> <span class="o">=</span> <span class="s2">"Optimizer.step#</span><span class="si">{}</span><span class="s2">.step"</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span><span class="p">)</span>
<span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">autograd</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">record_function</span><span class="p">(</span><span class="n">profile_name</span><span class="p">):</span>
<span class="c1"># call optimizer step pre hooks</span>
<span class="k">for</span> <span class="n">pre_hook</span> <span class="ow">in</span> <span class="n">chain</span><span class="p">(</span><span class="n">_global_optimizer_pre_hooks</span><span class="o">.</span><span class="n">values</span><span class="p">(),</span> <span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_pre_hooks</span><span class="o">.</span><span class="n">values</span><span class="p">()):</span>
<span class="n">result</span> <span class="o">=</span> <span class="n">pre_hook</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">args</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">)</span>
<span class="k">if</span> <span class="n">result</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">result</span><span class="p">,</span> <span class="nb">tuple</span><span class="p">)</span> <span class="ow">and</span> <span class="nb">len</span><span class="p">(</span><span class="n">result</span><span class="p">)</span> <span class="o">==</span> <span class="mi">2</span><span class="p">:</span>
<span class="n">args</span><span class="p">,</span> <span class="n">kwargs</span> <span class="o">=</span> <span class="n">result</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">RuntimeError</span><span class="p">(</span><span class="sa">f</span><span class="s2">"</span><span class="si">{</span><span class="n">func</span><span class="si">}</span><span class="s2"> must return None or a tuple of (new_args, new_kwargs),"</span>
<span class="sa">f</span><span class="s2">"but got </span><span class="si">{</span><span class="n">result</span><span class="si">}</span><span class="s2">."</span><span class="p">)</span>
<span class="n">out</span> <span class="o">=</span> <span class="n">func</span><span class="p">(</span><span class="o">*</span><span class="n">args</span><span class="p">,</span> <span class="o">**</span><span class="n">kwargs</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_code</span><span class="p">()</span>
<span class="c1"># call optimizer step post hooks</span>
<span class="k">for</span> <span class="n">post_hook</span> <span class="ow">in</span> <span class="n">chain</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_post_hooks</span><span class="o">.</span><span class="n">values</span><span class="p">(),</span> <span class="n">_global_optimizer_post_hooks</span><span class="o">.</span><span class="n">values</span><span class="p">()):</span>
<span class="n">post_hook</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">args</span><span class="p">,</span> <span class="n">kwargs</span><span class="p">)</span>
<span class="k">return</span> <span class="n">out</span>
<span class="k">return</span> <span class="n">wrapper</span>
<span class="k">def</span> <span class="nf">_patch_step_function</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_zero_grad_profile_name</span> <span class="o">=</span> <span class="s2">"Optimizer.zero_grad#</span><span class="si">{}</span><span class="s2">.zero_grad"</span><span class="o">.</span><span class="n">format</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="vm">__name__</span><span class="p">)</span>
<span class="n">hooked</span> <span class="o">=</span> <span class="nb">getattr</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="n">step</span><span class="p">,</span> <span class="s2">"hooked"</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">hooked</span><span class="p">:</span>
<span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="n">step</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">profile_hook_step</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="n">step</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="vm">__class__</span><span class="o">.</span><span class="n">step</span><span class="o">.</span><span class="n">hooked</span> <span class="o">=</span> <span class="kc">True</span>
<span class="k">def</span> <span class="nf">register_step_pre_hook</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">hook</span><span class="p">:</span> <span class="n">Callable</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="kc">None</span><span class="p">])</span> <span class="o">-></span> <span class="n">RemovableHandle</span><span class="p">:</span>
<span class="sa">r</span><span class="sd">"""Register an optimizer step pre hook which will be called before</span>
<span class="sd"> optimizer step. It should have the following signature::</span>
<span class="sd"> hook(optimizer, args, kwargs) -> None or modified args and kwargs</span>
<span class="sd"> The ``optimizer`` argument is the optimizer instance being used. If</span>
<span class="sd"> args and kwargs are modified by the pre-hook, then the transformed</span>
<span class="sd"> values are returned as a tuple containing the new_args and new_kwargs.</span>
<span class="sd"> Args:</span>
<span class="sd"> hook (Callable): The user defined hook to be registered.</span>
<span class="sd"> Returns:</span>
<span class="sd"> :class:`torch.utils.hooks.RemoveableHandle`:</span>
<span class="sd"> a handle that can be used to remove the added hook by calling</span>
<span class="sd"> ``handle.remove()``</span>
<span class="sd"> """</span>
<span class="n">handle</span> <span class="o">=</span> <span class="n">hooks</span><span class="o">.</span><span class="n">RemovableHandle</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_pre_hooks</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_pre_hooks</span><span class="p">[</span><span class="n">handle</span><span class="o">.</span><span class="n">id</span><span class="p">]</span> <span class="o">=</span> <span class="n">hook</span>
<span class="k">return</span> <span class="n">handle</span>
<span class="k">def</span> <span class="nf">register_step_post_hook</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">hook</span><span class="p">:</span> <span class="n">Callable</span><span class="p">[</span><span class="o">...</span><span class="p">,</span> <span class="kc">None</span><span class="p">])</span> <span class="o">-></span> <span class="n">RemovableHandle</span><span class="p">:</span>
<span class="sa">r</span><span class="sd">"""Register an optimizer step post hook which will be called after optimizer step.</span>
<span class="sd"> It should have the following signature::</span>
<span class="sd"> hook(optimizer, args, kwargs) -> None</span>
<span class="sd"> The ``optimizer`` argument is the optimizer instance being used.</span>
<span class="sd"> Args:</span>
<span class="sd"> hook (Callable): The user defined hook to be registered.</span>
<span class="sd"> Returns:</span>
<span class="sd"> :class:`torch.utils.hooks.RemoveableHandle`:</span>
<span class="sd"> a handle that can be used to remove the added hook by calling</span>
<span class="sd"> ``handle.remove()``</span>
<span class="sd"> """</span>
<span class="n">handle</span> <span class="o">=</span> <span class="n">hooks</span><span class="o">.</span><span class="n">RemovableHandle</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_post_hooks</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_optimizer_step_post_hooks</span><span class="p">[</span><span class="n">handle</span><span class="o">.</span><span class="n">id</span><span class="p">]</span> <span class="o">=</span> <span class="n">hook</span>
<span class="k">return</span> <span class="n">handle</span>
<div class="viewcode-block" id="Optimizer.state_dict"><a class="viewcode-back" href="../../../generated/torch.optim.Optimizer.state_dict.html#torch.optim.Optimizer.state_dict">[docs]</a> <span class="k">def</span> <span class="nf">state_dict</span><span class="p">(</span><span class="bp">self</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Returns the state of the optimizer as a :class:`dict`.</span>
<span class="sd"> It contains two entries:</span>
<span class="sd"> * state - a dict holding current optimization state. Its content</span>
<span class="sd"> differs between optimizer classes.</span>
<span class="sd"> * param_groups - a list containing all parameter groups where each</span>
<span class="sd"> parameter group is a dict</span>
<span class="sd"> """</span>
<span class="c1"># Save order indices instead of Tensors</span>
<span class="n">param_mappings</span> <span class="o">=</span> <span class="p">{}</span>
<span class="n">start_index</span> <span class="o">=</span> <span class="mi">0</span>
<span class="k">def</span> <span class="nf">pack_group</span><span class="p">(</span><span class="n">group</span><span class="p">):</span>
<span class="k">nonlocal</span> <span class="n">start_index</span>
<span class="n">packed</span> <span class="o">=</span> <span class="p">{</span><span class="n">k</span><span class="p">:</span> <span class="n">v</span> <span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">group</span><span class="o">.</span><span class="n">items</span><span class="p">()</span> <span class="k">if</span> <span class="n">k</span> <span class="o">!=</span> <span class="s1">'params'</span><span class="p">}</span>
<span class="n">param_mappings</span><span class="o">.</span><span class="n">update</span><span class="p">({</span><span class="nb">id</span><span class="p">(</span><span class="n">p</span><span class="p">):</span> <span class="n">i</span> <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="nb">enumerate</span><span class="p">(</span><span class="n">group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">],</span> <span class="n">start_index</span><span class="p">)</span>
<span class="k">if</span> <span class="nb">id</span><span class="p">(</span><span class="n">p</span><span class="p">)</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">param_mappings</span><span class="p">})</span>
<span class="n">packed</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="o">=</span> <span class="p">[</span><span class="n">param_mappings</span><span class="p">[</span><span class="nb">id</span><span class="p">(</span><span class="n">p</span><span class="p">)]</span> <span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]]</span>
<span class="n">start_index</span> <span class="o">+=</span> <span class="nb">len</span><span class="p">(</span><span class="n">packed</span><span class="p">[</span><span class="s1">'params'</span><span class="p">])</span>
<span class="k">return</span> <span class="n">packed</span>
<span class="n">param_groups</span> <span class="o">=</span> <span class="p">[</span><span class="n">pack_group</span><span class="p">(</span><span class="n">g</span><span class="p">)</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">]</span>
<span class="c1"># Remap state to use order indices as keys</span>
<span class="n">packed_state</span> <span class="o">=</span> <span class="p">{(</span><span class="n">param_mappings</span><span class="p">[</span><span class="nb">id</span><span class="p">(</span><span class="n">k</span><span class="p">)]</span> <span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">k</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">)</span> <span class="k">else</span> <span class="n">k</span><span class="p">):</span> <span class="n">v</span>
<span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">state</span><span class="o">.</span><span class="n">items</span><span class="p">()}</span>
<span class="k">return</span> <span class="p">{</span>
<span class="s1">'state'</span><span class="p">:</span> <span class="n">packed_state</span><span class="p">,</span>
<span class="s1">'param_groups'</span><span class="p">:</span> <span class="n">param_groups</span><span class="p">,</span>
<span class="p">}</span></div>
<div class="viewcode-block" id="Optimizer.load_state_dict"><a class="viewcode-back" href="../../../generated/torch.optim.Optimizer.load_state_dict.html#torch.optim.Optimizer.load_state_dict">[docs]</a> <span class="k">def</span> <span class="nf">load_state_dict</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">state_dict</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Loads the optimizer state.</span>
<span class="sd"> Args:</span>
<span class="sd"> state_dict (dict): optimizer state. Should be an object returned</span>
<span class="sd"> from a call to :meth:`state_dict`.</span>
<span class="sd"> """</span>
<span class="c1"># deepcopy, to be consistent with module API</span>
<span class="n">state_dict</span> <span class="o">=</span> <span class="n">deepcopy</span><span class="p">(</span><span class="n">state_dict</span><span class="p">)</span>
<span class="c1"># Validate the state_dict</span>
<span class="n">groups</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span>
<span class="n">saved_groups</span> <span class="o">=</span> <span class="n">state_dict</span><span class="p">[</span><span class="s1">'param_groups'</span><span class="p">]</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">groups</span><span class="p">)</span> <span class="o">!=</span> <span class="nb">len</span><span class="p">(</span><span class="n">saved_groups</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"loaded state dict has a different number of "</span>
<span class="s2">"parameter groups"</span><span class="p">)</span>
<span class="n">param_lens</span> <span class="o">=</span> <span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">g</span><span class="p">[</span><span class="s1">'params'</span><span class="p">])</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="n">groups</span><span class="p">)</span>
<span class="n">saved_lens</span> <span class="o">=</span> <span class="p">(</span><span class="nb">len</span><span class="p">(</span><span class="n">g</span><span class="p">[</span><span class="s1">'params'</span><span class="p">])</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="n">saved_groups</span><span class="p">)</span>
<span class="k">if</span> <span class="nb">any</span><span class="p">(</span><span class="n">p_len</span> <span class="o">!=</span> <span class="n">s_len</span> <span class="k">for</span> <span class="n">p_len</span><span class="p">,</span> <span class="n">s_len</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">param_lens</span><span class="p">,</span> <span class="n">saved_lens</span><span class="p">)):</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"loaded state dict contains a parameter group "</span>
<span class="s2">"that doesn't match the size of optimizer's group"</span><span class="p">)</span>
<span class="c1"># Update the state</span>
<span class="n">id_map</span> <span class="o">=</span> <span class="p">{</span><span class="n">old_id</span><span class="p">:</span> <span class="n">p</span> <span class="k">for</span> <span class="n">old_id</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span>
<span class="nb">zip</span><span class="p">(</span><span class="n">chain</span><span class="o">.</span><span class="n">from_iterable</span><span class="p">((</span><span class="n">g</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="n">saved_groups</span><span class="p">)),</span>
<span class="n">chain</span><span class="o">.</span><span class="n">from_iterable</span><span class="p">((</span><span class="n">g</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="k">for</span> <span class="n">g</span> <span class="ow">in</span> <span class="n">groups</span><span class="p">)))}</span>
<span class="k">def</span> <span class="nf">cast</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">value</span><span class="p">,</span> <span class="n">key</span><span class="o">=</span><span class="kc">None</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Make a deep copy of value, casting all tensors to device of param."""</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">value</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
<span class="c1"># Floating-point types are a bit special here. They are the only ones</span>
<span class="c1"># that are assumed to always match the type of params.</span>
<span class="c1"># Make sure state['step'] is not casted https://github.com/pytorch/pytorch/issues/74424</span>
<span class="k">if</span> <span class="p">(</span><span class="n">key</span> <span class="o">!=</span> <span class="s2">"step"</span><span class="p">):</span>
<span class="k">if</span> <span class="n">param</span><span class="o">.</span><span class="n">is_floating_point</span><span class="p">():</span>
<span class="n">value</span> <span class="o">=</span> <span class="n">value</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">param</span><span class="o">.</span><span class="n">dtype</span><span class="p">)</span>
<span class="n">value</span> <span class="o">=</span> <span class="n">value</span><span class="o">.</span><span class="n">to</span><span class="p">(</span><span class="n">param</span><span class="o">.</span><span class="n">device</span><span class="p">)</span>
<span class="k">return</span> <span class="n">value</span>
<span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">value</span><span class="p">,</span> <span class="nb">dict</span><span class="p">):</span>
<span class="k">return</span> <span class="p">{</span><span class="n">k</span><span class="p">:</span> <span class="n">cast</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">v</span><span class="p">,</span> <span class="n">key</span><span class="o">=</span><span class="n">k</span><span class="p">)</span> <span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">value</span><span class="o">.</span><span class="n">items</span><span class="p">()}</span>
<span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">value</span><span class="p">,</span> <span class="n">container_abcs</span><span class="o">.</span><span class="n">Iterable</span><span class="p">):</span>
<span class="k">return</span> <span class="nb">type</span><span class="p">(</span><span class="n">value</span><span class="p">)(</span><span class="n">cast</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span> <span class="k">for</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">value</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">return</span> <span class="n">value</span>
<span class="c1"># Copy state assigned to params (and cast tensors to appropriate types).</span>
<span class="c1"># State that is not assigned to params is copied as is (needed for</span>
<span class="c1"># backward compatibility).</span>
<span class="n">state</span> <span class="o">=</span> <span class="n">defaultdict</span><span class="p">(</span><span class="nb">dict</span><span class="p">)</span>
<span class="k">for</span> <span class="n">k</span><span class="p">,</span> <span class="n">v</span> <span class="ow">in</span> <span class="n">state_dict</span><span class="p">[</span><span class="s1">'state'</span><span class="p">]</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
<span class="k">if</span> <span class="n">k</span> <span class="ow">in</span> <span class="n">id_map</span><span class="p">:</span>
<span class="n">param</span> <span class="o">=</span> <span class="n">id_map</span><span class="p">[</span><span class="n">k</span><span class="p">]</span>
<span class="n">state</span><span class="p">[</span><span class="n">param</span><span class="p">]</span> <span class="o">=</span> <span class="n">cast</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">v</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">state</span><span class="p">[</span><span class="n">k</span><span class="p">]</span> <span class="o">=</span> <span class="n">v</span>
<span class="c1"># Update parameter groups, setting their 'params' value</span>
<span class="k">def</span> <span class="nf">update_group</span><span class="p">(</span><span class="n">group</span><span class="p">,</span> <span class="n">new_group</span><span class="p">):</span>
<span class="n">new_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="o">=</span> <span class="n">group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span>
<span class="k">return</span> <span class="n">new_group</span>
<span class="n">param_groups</span> <span class="o">=</span> <span class="p">[</span>
<span class="n">update_group</span><span class="p">(</span><span class="n">g</span><span class="p">,</span> <span class="n">ng</span><span class="p">)</span> <span class="k">for</span> <span class="n">g</span><span class="p">,</span> <span class="n">ng</span> <span class="ow">in</span> <span class="nb">zip</span><span class="p">(</span><span class="n">groups</span><span class="p">,</span> <span class="n">saved_groups</span><span class="p">)]</span>
<span class="bp">self</span><span class="o">.</span><span class="n">__setstate__</span><span class="p">({</span><span class="s1">'state'</span><span class="p">:</span> <span class="n">state</span><span class="p">,</span> <span class="s1">'param_groups'</span><span class="p">:</span> <span class="n">param_groups</span><span class="p">})</span></div>
<div class="viewcode-block" id="Optimizer.zero_grad"><a class="viewcode-back" href="../../../generated/torch.optim.Optimizer.zero_grad.html#torch.optim.Optimizer.zero_grad">[docs]</a> <span class="k">def</span> <span class="nf">zero_grad</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">set_to_none</span><span class="p">:</span> <span class="nb">bool</span> <span class="o">=</span> <span class="kc">True</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Sets the gradients of all optimized :class:`torch.Tensor` s to zero.</span>
<span class="sd"> Args:</span>
<span class="sd"> set_to_none (bool): instead of setting to zero, set the grads to None.</span>
<span class="sd"> This will in general have lower memory footprint, and can modestly improve performance.</span>
<span class="sd"> However, it changes certain behaviors. For example:</span>
<span class="sd"> 1. When the user tries to access a gradient and perform manual ops on it,</span>
<span class="sd"> a None attribute or a Tensor full of 0s will behave differently.</span>
<span class="sd"> 2. If the user requests ``zero_grad(set_to_none=True)`` followed by a backward pass, ``.grad``\ s</span>
<span class="sd"> are guaranteed to be None for params that did not receive a gradient.</span>
<span class="sd"> 3. ``torch.optim`` optimizers have a different behavior if the gradient is 0 or None</span>
<span class="sd"> (in one case it does the step with a gradient of 0 and in the other it skips</span>
<span class="sd"> the step altogether).</span>
<span class="sd"> """</span>
<span class="n">foreach</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s1">'foreach'</span><span class="p">,</span> <span class="kc">False</span><span class="p">)</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">hasattr</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="s2">"_zero_grad_profile_name"</span><span class="p">):</span>
<span class="bp">self</span><span class="o">.</span><span class="n">_patch_step_function</span><span class="p">()</span>
<span class="k">if</span> <span class="n">foreach</span><span class="p">:</span>
<span class="n">per_device_and_dtype_grads</span> <span class="o">=</span> <span class="n">defaultdict</span><span class="p">(</span><span class="k">lambda</span><span class="p">:</span> <span class="n">defaultdict</span><span class="p">(</span><span class="nb">list</span><span class="p">))</span>
<span class="k">with</span> <span class="n">torch</span><span class="o">.</span><span class="n">autograd</span><span class="o">.</span><span class="n">profiler</span><span class="o">.</span><span class="n">record_function</span><span class="p">(</span><span class="bp">self</span><span class="o">.</span><span class="n">_zero_grad_profile_name</span><span class="p">):</span>
<span class="k">for</span> <span class="n">group</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">:</span>
<span class="k">for</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]:</span>
<span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">grad</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="k">if</span> <span class="n">set_to_none</span><span class="p">:</span>
<span class="n">p</span><span class="o">.</span><span class="n">grad</span> <span class="o">=</span> <span class="kc">None</span>
<span class="k">else</span><span class="p">:</span>
<span class="k">if</span> <span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">grad_fn</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span><span class="p">:</span>
<span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">detach_</span><span class="p">()</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">requires_grad_</span><span class="p">(</span><span class="kc">False</span><span class="p">)</span>
<span class="k">if</span> <span class="p">(</span><span class="ow">not</span> <span class="n">foreach</span> <span class="ow">or</span> <span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">is_sparse</span><span class="p">):</span>
<span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">zero_</span><span class="p">()</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">per_device_and_dtype_grads</span><span class="p">[</span><span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">device</span><span class="p">][</span><span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="o">.</span><span class="n">dtype</span><span class="p">]</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">p</span><span class="o">.</span><span class="n">grad</span><span class="p">)</span>
<span class="k">if</span> <span class="n">foreach</span><span class="p">:</span>
<span class="k">for</span> <span class="n">_</span><span class="p">,</span> <span class="n">per_dtype_grads</span> <span class="ow">in</span> <span class="n">per_device_and_dtype_grads</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
<span class="k">for</span> <span class="n">grads</span> <span class="ow">in</span> <span class="n">per_dtype_grads</span><span class="o">.</span><span class="n">values</span><span class="p">():</span>
<span class="n">torch</span><span class="o">.</span><span class="n">_foreach_zero_</span><span class="p">(</span><span class="n">grads</span><span class="p">)</span></div>
<div class="viewcode-block" id="Optimizer.step"><a class="viewcode-back" href="../../../generated/torch.optim.Optimizer.step.html#torch.optim.Optimizer.step">[docs]</a> <span class="k">def</span> <span class="nf">step</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">closure</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Performs a single optimization step (parameter update).</span>
<span class="sd"> Args:</span>
<span class="sd"> closure (Callable): A closure that reevaluates the model and</span>
<span class="sd"> returns the loss. Optional for most optimizers.</span>
<span class="sd"> .. note::</span>
<span class="sd"> Unless otherwise specified, this function should not modify the</span>
<span class="sd"> ``.grad`` field of the parameters.</span>
<span class="sd"> """</span>
<span class="k">raise</span> <span class="ne">NotImplementedError</span></div>
<div class="viewcode-block" id="Optimizer.add_param_group"><a class="viewcode-back" href="../../../generated/torch.optim.Optimizer.add_param_group.html#torch.optim.Optimizer.add_param_group">[docs]</a> <span class="k">def</span> <span class="nf">add_param_group</span><span class="p">(</span><span class="bp">self</span><span class="p">,</span> <span class="n">param_group</span><span class="p">):</span>
<span class="sa">r</span><span class="sd">"""Add a param group to the :class:`Optimizer` s `param_groups`.</span>
<span class="sd"> This can be useful when fine tuning a pre-trained network as frozen layers can be made</span>
<span class="sd"> trainable and added to the :class:`Optimizer` as training progresses.</span>
<span class="sd"> Args:</span>
<span class="sd"> param_group (dict): Specifies what Tensors should be optimized along with group</span>
<span class="sd"> specific optimization options.</span>
<span class="sd"> """</span>
<span class="k">assert</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">param_group</span><span class="p">,</span> <span class="nb">dict</span><span class="p">),</span> <span class="s2">"param group must be a dict"</span>
<span class="n">params</span> <span class="o">=</span> <span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span>
<span class="k">if</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">params</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
<span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="o">=</span> <span class="p">[</span><span class="n">params</span><span class="p">]</span>
<span class="k">elif</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">params</span><span class="p">,</span> <span class="nb">set</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">TypeError</span><span class="p">(</span><span class="s1">'optimizer parameters need to be organized in ordered collections, but '</span>
<span class="s1">'the ordering of tensors in sets will change between runs. Please use a list instead.'</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span> <span class="o">=</span> <span class="nb">list</span><span class="p">(</span><span class="n">params</span><span class="p">)</span>
<span class="k">for</span> <span class="n">param</span> <span class="ow">in</span> <span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]:</span>
<span class="k">if</span> <span class="ow">not</span> <span class="nb">isinstance</span><span class="p">(</span><span class="n">param</span><span class="p">,</span> <span class="n">torch</span><span class="o">.</span><span class="n">Tensor</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">TypeError</span><span class="p">(</span><span class="s2">"optimizer can only optimize Tensors, "</span>
<span class="s2">"but one of the params is "</span> <span class="o">+</span> <span class="n">torch</span><span class="o">.</span><span class="n">typename</span><span class="p">(</span><span class="n">param</span><span class="p">))</span>
<span class="k">if</span> <span class="ow">not</span> <span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="s1">'differentiable'</span><span class="p">,</span> <span class="kc">None</span><span class="p">)</span> <span class="ow">and</span> <span class="ow">not</span> <span class="p">(</span><span class="n">param</span><span class="o">.</span><span class="n">is_leaf</span> <span class="ow">or</span> <span class="n">param</span><span class="o">.</span><span class="n">retains_grad</span><span class="p">):</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"can't optimize a non-leaf Tensor"</span><span class="p">)</span>
<span class="k">for</span> <span class="n">name</span><span class="p">,</span> <span class="n">default</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">defaults</span><span class="o">.</span><span class="n">items</span><span class="p">():</span>
<span class="k">if</span> <span class="n">default</span> <span class="ow">is</span> <span class="n">required</span> <span class="ow">and</span> <span class="n">name</span> <span class="ow">not</span> <span class="ow">in</span> <span class="n">param_group</span><span class="p">:</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"parameter group didn't specify a value of required optimization parameter "</span> <span class="o">+</span>
<span class="n">name</span><span class="p">)</span>
<span class="k">else</span><span class="p">:</span>
<span class="n">param_group</span><span class="o">.</span><span class="n">setdefault</span><span class="p">(</span><span class="n">name</span><span class="p">,</span> <span class="n">default</span><span class="p">)</span>
<span class="n">params</span> <span class="o">=</span> <span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]</span>
<span class="k">if</span> <span class="nb">len</span><span class="p">(</span><span class="n">params</span><span class="p">)</span> <span class="o">!=</span> <span class="nb">len</span><span class="p">(</span><span class="nb">set</span><span class="p">(</span><span class="n">params</span><span class="p">)):</span>
<span class="n">warnings</span><span class="o">.</span><span class="n">warn</span><span class="p">(</span><span class="s2">"optimizer contains a parameter group with duplicate parameters; "</span>
<span class="s2">"in future, this will cause an error; "</span>
<span class="s2">"see github.com/pytorch/pytorch/issues/40967 for more information"</span><span class="p">,</span> <span class="n">stacklevel</span><span class="o">=</span><span class="mi">3</span><span class="p">)</span>
<span class="n">param_set</span> <span class="o">=</span> <span class="nb">set</span><span class="p">()</span>
<span class="k">for</span> <span class="n">group</span> <span class="ow">in</span> <span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="p">:</span>
<span class="n">param_set</span><span class="o">.</span><span class="n">update</span><span class="p">(</span><span class="nb">set</span><span class="p">(</span><span class="n">group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">]))</span>
<span class="k">if</span> <span class="ow">not</span> <span class="n">param_set</span><span class="o">.</span><span class="n">isdisjoint</span><span class="p">(</span><span class="nb">set</span><span class="p">(</span><span class="n">param_group</span><span class="p">[</span><span class="s1">'params'</span><span class="p">])):</span>
<span class="k">raise</span> <span class="ne">ValueError</span><span class="p">(</span><span class="s2">"some parameters appear in more than one parameter group"</span><span class="p">)</span>
<span class="bp">self</span><span class="o">.</span><span class="n">param_groups</span><span class="o">.</span><span class="n">append</span><span class="p">(</span><span class="n">param_group</span><span class="p">)</span></div></div>
</pre></div>
</article>
</div>
<footer>